Ex vivo modeling of morphological, molecular and pharmacological tumor heterogeneity of metastatic colorectal cancer

Ex vivo modeling of morphological, molecular and pharmacological tumor heterogeneity of metastatic colorectal cancer | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Resource Ex vivo modeling of morphological, molecular and pharmacological tumor heterogeneity of metastatic colorectal cancer Ragnhild Lothe, Kushtrim Kryeziu, Solveig Klokkerud, Max Totland, and 13 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6507406/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract This study reports the establishment and pharmacogenomics analyses of tumor heterogeneity in a living biobank of tumor organoids of 213 liver metastases from 102 patients with metastatic colorectal cancer. Successful organoid culturing reflected poorer chemosensitivity and patient survival. Molecular fidelity was demonstrated in tumor-organoid sample pairs, and multi-modal phenotypes were proposed based on organoid morphologies. Cystic morphology was associated with intestinal stem cell markers and higher drug sensitivities, and solid morphology with markers of cancer cell plasticity and aggressiveness. Potential to identify treatments with less vulnerability to tumor heterogeneity was supported by multi-lesion analyses in 65 patients. Complexity of clinical translation was illustrated by two prospective cases of pharmacogenomics-guided treatment, including successful chemotherapy rechallenge and targeted therapy resistance in cancers with low and high tumor heterogeneity, respectively. All pharmacogenomics data are available as a functional oncology resource and serve as reference for an ongoing intervention trial, supporting the interpretation of ex vivo drug sensitivities into prospective clinical “actionability”. Biological sciences/Cancer/Cancer models Biological sciences/Cancer/Tumour heterogeneity Biological sciences/Cancer/Gastrointestinal cancer/Colorectal cancer Biological sciences/Cancer/Cancer therapy Biological sciences/Cancer/Metastasis Figures Figure 1 Figure 2 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Functional oncology based on ex vivo drug sensitivity testing of patient-derived cancer cells can complement genomics-guided approaches to precision cancer medicine 1 . Prospective treatment selection guided by functional assays has shown clinical benefit in patients with hematologic cancers but has proven more difficult in patients with solid tumors 2 – 4 . Tumor cells can be cultured as self-organizing three-dimensional organoid structures that resemble the original tissue, and the use of organoid models in cancer research has increased exponentially over the past decade. Colorectal cancer (CRC) was the first cancer type to be modeled with patient-derived organoids (PDOs) 5 , and both the tissue architecture and molecular profiles of CRCs can be recapitulated with this model system 6 , 7 . Drug sensitivity testing of CRC PDOs can also reflect the clinical responses of the original tumors to standard chemotherapy and targeted agents 8 – 10 . However, successful examples of functional precision oncology and prospective treatment selection in patients using this approach remain few or anecdotal 10 – 12 . Tumor heterogeneity is a major cause of treatment resistance and poor patient outcome 13 , 14 . Metastatic heterogeneity of CRC has been extensively studied at the molecular level, highlighting early seeding, polyclonal seeding, seeding among metastatic deposits, and phenotypic plasticity in development of metastatic disease, potentially augmented by treatment exposure 15 – 19 . Furthermore, heterogeneity of response to standard treatment is common among metastatic lesions and can have a negative impact on patient outcomes 20 – 22 . However, the largest living biobanks of CRC do not model tumor heterogeneity and typically include a single PDO per patient 23 , 24 . PDOs derived from multiple single cells or regions of primary tumors have suggested marked intra-tumor heterogeneity of drug sensitivities 25 , 26 , while paired PDOs of the primary tumor and liver metastasis have shown similar ex vivo sensitivity to standard chemotherapies 27 , 28 . Here, we expand on the largest published study of ex vivo pharmacogenomic heterogeneity in metastatic CRC 29 . The corresponding pharmaco-omics data serve as reference for prospective drug nomination in an ongoing intervention trial of metastatic CRC (ClinicalTrials.gov identifier NCT05725200), and we report from the observational and translational pilot phase of this functional oncology study. Results A living biobank of metastatic CRC Fresh tissue samples of 346 colorectal liver metastases (CRLMs) from 132 patients treated by hepatectomy for metastatic CRC were included for ex vivo culturing (Supplementary Table 1 and Supplementary Fig. 1). PDOs were successfully established for 213 lesions (62%) from 102 patients (77%), including multiple (up to 6) spatially separated lesions from each of 65 patients (64%), as well as recurrent CRLMs sampled at re-resections of six patients (Fig. 1 a-d). The median time from sampling to a completed drug screen and cryopreservation of PDOs for biobanking was 8.4 weeks (10th–90th percentile range: 4–21 weeks). There was no difference in the culture success rate according to tumor location (liver segment; Fig. 1 a), the number of lesions attempted to culture per patient (Fig. 1 c), the type of surgery, or any of the clinicopathological or molecular characteristics listed in Supplementary Table 1. Samples with low tumor cell content (n = 3) or necrosis (n = 5) failed to grow, but the majority (90%) of growth failures could not be attributed to histopathological or molecular markers (Fig. 1 b, Supplementary Table 1, Extended Data Fig. 1 ). There was also no growth selection based on neoadjuvant chemotherapy exposure (p > 0.9 by Pearson's Chi-squared test of treated versus chemo-naïve lesions), but lesions that did not propagate ex vivo had a stronger response to neoadjuvant treatment (Fig. 1 e). Ex vivo drug sensitivities in the successfully established PDOs did not reflect clinical tumor responses to the corresponding neoadjuvant treatment, and this might also reflect selection of propagated cells based on treatment response (Supplementary Fig. 2). Patients with at least one successfully established PDO had a significantly shorter median overall survival time after liver resection (38 months) than patients with no PDOs established (median not reached at four years follow-up; Fig. 1 f), supporting the hypothesis that successful ex vivo growth reflects aggressive disease and/or poorer response to chemotherapy. An overview of the living biobank and pharmaco-omics data is shown in Fig. 2 , including drug sensitivity profiles of 41 or 47 drugs (illustrated by 24 overlapping drugs between the two custom libraries), gene expression profiles (illustrated by classification according to the intrinsic consensus molecular subtypes, iCMS 30 ), mutations of 20 CRC-relevant genes (illustrated by KRAS / NRAS [ RAS ], BRAF and TP53 ), multiplex fluorescent immunohistochemistry of 12 proteins (illustrated by CDX2, KRT20 and KI67), and morphological phenotypes based on hematoxylin and eosin (HE) stains. All PDOs were microsatellite stable (MSS). PDOs are faithful models of the biological and molecular diversity of CRLMs Paired HE stains of PDOs and corresponding tumor tissue samples (n = 46 lesions from 26 patients) indicated that 85% of PDOs retained the tumor morphologies for histopathological features such as acinar, cystic and solid structures (Fig. 3 a, Extended Data Fig. 2 and Supplementary Table 2). PDOs with dissimilar morphologies were not characterized by distinct growth rates or any specific molecular marker (Supplementary Table 3). Gene expression profiling indicated high similarity scores of PDOs (n = 211) for primary CRC tissue according to the CancerCellNet approach 31 (Extended Data Fig. 3 a). Notably, PDOs had higher similarity scores than their paired tissue samples (n = 50 sample pairs evaluated; p = 7×10 − 4 by paired t-test; Extended Data Fig. 3 b), likely reflecting influence from the liver-specific tumor microenvironment in the comparison of CRLMs with primary tumors. Gene set enrichment analysis of the PDO-CRLM sample pairs confirmed enrichment exclusively with tumor microenvironment signals in tissue samples relative to PDOs (Fig. 3 b). PDOs showed enrichment with signatures of metabolism and the cell cycle, likely reflecting higher proliferative activity and ample access to nutrients and oxygen in the cell cultures. The first component from principal component analysis across both PDOs and CRLMs was primarily driven by the difference in tumor microenvironment signals (Extended Data Fig. 3 c-e) and showed no correlation between the matched sample pairs (Fig. 3 c). In contrast, the second component was driven by cancer cell-intrinsic features (illustrated with a BRAF mutant-like expression signature 32 ) and was strongly correlated between the PDOs and paired tumor tissue, indicating that PDOs recapitulated the transcriptomic profile of their original tumor. Mutation frequencies of the ten most frequently mutated genes (analyzed across a single PDO from each of the 102 patients) were similar to two previous studies of metastatic CRC 33 , 34 (p = 0.98 by chi-squared test; Fig. 3 d). Matched PDO-CRLM sample pairs (n = 74 lesions from 56 patients) also showed highly correspondent mutation profiles, with concordant status for 93.7% of the 491 detected mutations (Supplementary Fig. 3a). Discordances were primarily due to subclonal mutations with low mutant allele fractions (p = 0.002 by Wilcoxon’s test of the allelic fraction of discordant and concordant mutations), frequently involving a second hit in APC or TP53 mutations (Supplementary Fig. 3a and 3c). Pharmacological associations of molecular markers Known pharmacogenomic associations were confirmed in the living biobank, including resistance to EGFR inhibitors in RAS and BRAF V600E mutated PDOs (Fig. 4 a and Extended data Fig. 4 a-b), sensitivity to the MDM2-TP53 inhibitor idasanutlin in PDOs with wild-type TP53 and/or high TP53 mRNA expression (Extended Data Fig. 4 c-e), as well as outlier sensitivity to sotorasib in the two PDOs with KRAS G 12 C mutation (from a single patient; Extended Data Fig. 4 f). Notably, the majority of BRAF V600E mutated PDOs were resistant to single-agent treatment with the mutation-specific inhibitor encorafenib (PDOs from 4 of 5 patients; Extended Data Fig. 4 g). A combination of encorafenib with afatinib (EGFR inhibitor) and trametinib (MEK inhibitor) showed increased activity (Extended Data Fig. 4 h). However, this effect was not specific to BRAF V600E mutated PDOs (Extended Data Fig. 4 i), suggesting that distinction between the synergistic and additive effects of the drug combination requires further analyses of complete dose-response matrices of the drug pairs. An extended analysis of drug sensitivities (n = 51 unique drugs across both libraries; Supplementary Table 4) and mutations (n = 8 genes with at least 10 mutated and wild-type PDOs) showed 12 significant associations, all involving APC, KRAS, NRAS, BRAF , or TP53 (Wilcoxon tests with Benjamini-Hochberg correction; Fig. 4 a). Beyond the well-known pharmacogenomic associations, KRAS mutations were associated with sensitivity to the multi-kinase inhibitor regorafenib (p = 0.045). There was no difference according to mutation hotspots (Extended Data Fig. 5 a), and the association was stronger when including BRAF V600E (p = 7.7×10 − 3 ). BRAF V600E mutated PDOs showed low sensitivity to the IGF1R inhibitor BMS-754807 (p = 0.01). This association was specific to BRAF V600E and not found for any of the RAS hotspots (Extended Data Fig. 5 b). Furthermore, TP53 mutated and APC mutated PDOs were less sensitive to the standard chemotherapies 5-FU (p = 0.03) and SN-38 (active metabolite of irinotecan; p = 0.01) than the corresponding wild-type PDOs, respectively (Fig. 4 a). A systematic evaluation of co-occurring mutations in pairs of genes (Extended Data Fig. 6 a) showed that TP53 and APC co-mutated PDOs had lower drug sensitivities in general (Supplementary Table 5) and a lower mean drug sensitivity score (DSS) across the drug libraries (p = 0.02). This was also the case for the poor-prognostic subgroup defined by co-occurring mutations of RAS / BRAF V600E and TP53 (Fig. 4 b). The strongest positive association to drug sensitivity was observed with co-occurring RAS and PIK3CA mutations, and these PDOs had strong sensitivity to the combination chemotherapy FLIRI (5-FU, leucovorin and SN-38; Fig. 4 c), although not to any of the corresponding single drugs (Supplementary Table 5). Comparisons of DSS values according to the transcriptomic iCMS classification showed higher activity of EGFR inhibitors in iCMS2 PDOs, attributed to enrichment with RAS/BRAF V600E mutations in the iCMS3 group (odds ratio 32, p = 3.5×10 − 13 by Fisher’s exact test). However, higher activity of gemcitabine (chemotherapy) and alisertib (Aurora kinase A inhibitor) in iCMS3 PDOs was independent of RAS/BRAF V600E mutation status (Fig. 4 d; Extended Data Fig. 5 c). Multiplex fluorescent immunohistochemistry of selected diagnostic markers, proteins involved in drug resistance, and other proteins of interest (n = 12) showed that the activity of 78% of drugs (n = 28 of 33 single drugs) was significantly correlated to the expression of at least one protein among PDOs (n = 77–134 PDOs from 37–67 patients; Fig. 4 e). The strongest drug-protein interaction was a negative correlation between expression of the ABC transporter ABCB1 and activity of the PI3K/AKT pathway inhibitor gedatolisib (Spearman´s rho=-0.5, p = 9.3x10 − 10 ). In general, proteins known to be involved in drug resistance and stress response, including ABCB1, ABCG2, UGT1A, and HSF1, were negatively correlated with the activity of several drugs. The strongest positive correlations were found between CDX2 expression and the standard chemotherapies 5-FU and TAS-102 (trifluridine/tipiracil). 5-FU sensitivity was also correlated with RCC2 expression, a protein suggested to be associated with benefit from adjuvant chemotherapy in locoregional CRC 35 . High expression of either CDX2 or RCC2 (> 70th percentiles) accounted for the majority of 5-FU-sensitive PDOs (DSS > 14.3; odds ratio 2.7, p = 0.01; Extended Data Fig. 5 d). Furthermore, TP53 expression was negatively correlated with sensitivity to the PARP inhibitor olaparib. High TP53 expression was found exclusively in PDOs with missense TP53 mutations (Extended Data Fig. 4 e), consistent with our previous study suggesting that wild-type TP53 activity is needed for PARP inhibitor sensitivity in CRC cells 36 . Additionally, RIPK1 expression was positively correlated to the activity of the SMAC mimetic LCL161, which targets inhibitor of apoptosis proteins 37 . Clinicopathological and pharmacological associations Patients with right-sided primary tumors have limited benefit from anti-EGFR treatment 38 , and the living biobank confirmed sensitivity to EGFR inhibitors only in RAS / BRAF wild-type PDOs derived from patients with left-sided or rectal primary tumors (p = 0.02 by Wilcoxon test; Extended Data Fig. 6 b). In contrast, right-sided primary tumor location was associated with higher sensitivity to the standard chemotherapies SN-38 and TAS-102, independent of RAS / BRAF mutation status (Extended Data Fig. 6 b and Supplementary Table 6). Patient sex was not associated with sensitivity to any drug, including 5-FU (p = 0.7 by Wilcoxon test), while high patient age was weakly correlated to 5-FU sensitivity (Spearman rho = 0.28, p = 0.02; Supplementary Table 7). Diagnosis with multiple CRLMs on radiological imaging before liver surgery was associated with low drug sensitivity in general, although a statistically significant correlation was found for SN-38 and idasanutlin only (Supplementary Table 8). Cox proportional hazards analysis indicated that high ex vivo drug sensitivities were associated with better overall survival among the patients (mean DSS across 24 drugs as predictor; hazard ratio 0.91, 95% confidence interval 0.82-1.0, p = 0.07), and the prognostic association was significant for four drugs (atorvastatin, bemcentinib, and two EGFR inhibitors; Supplementary Table 9). Sensitivity to the cholesterol-lowering agent atorvastatin and the AXL inhibitor bemcentinib maintained significant prognostic associations in multivariable models with clinicopathological variables and RAS / BRAF mutation status (multivariable hazard ratio 0.81 and 0.79, 95% confidence interval 0.71–0.93 and 0.64–0.96, respectively; Extended Data Fig. 7 ). PDOs of CRLMs represent two main morphological phenotypes PDOs were classified into two main morphological phenotypes based on HE stains of paraffin embedded samples (n = 213; Fig. 5 a, Extended Data Fig. 2 ). A cystic phenotype was identified in 28% of PDOs and was characterized by organized cuboidal and columnar cells forming cystic/glandular structures with a well-defined central lumen. A solid phenotype was found in 38% of PDOs and these lacked a central lumen and had predominantly solid growth patterns, including cytoplasmic vacuoles or individually disaggregated cells. The remaining PDOs (34%) had traits of both the cystic and solid phenotypes and were classified as mixed (less than 50% of the individual structures showed clear characteristics of either phenotype). Multiplex fluorescent immunohistochemistry showed differential expression of the CRC diagnostic markers KRT20 and CDX2 according to the morphological phenotypes (Fig. 5 b and Extended Data Fig. 8a-b). CDX2 had highest expression in cystic PDOs, KRT20 in solid PDOs, and the mixed phenotype showed intermediate expression levels of both proteins. The solid phenotype was also enriched with RAS and BRAF V600E mutations (Fig. 5 c), and 92% of BRAF V600E PDOs had a solid phenotype (Extended Data Fig. 8c). Consistently, enrichment with a BRAF mutant-like expression signature in solid PDOs was also the strongest distinction between the subtypes on the transcriptomic level (Fig. 5 e, Extended Data Fig. 8e). Solid PDOs were also more MSI-like and enriched with signatures of apoptosis and inflammation, including TNF-α signaling and IFN-α response, as well as the iCMS3 class (Fig. 5 d). In contrast, cystic PDOs were MSS-like and enriched with cancer stem cell signatures and iCMS2. Cystic PDOs were also generally more sensitive to drugs, estimated as a higher mean sensitivity across the 24 drugs evaluated in both libraries (Fig. 5 f and Supplementary Table 10). This included higher sensitivity to standard treatments for CRC such as 5-FU and EGFR inhibitors (Fig. 5 f), reflecting higher expression of CDX2 and less frequent mutations of RAS / BRAF V600E (Extended Data Fig. 8d). Modeling metastatic heterogeneity with multi-lesion PDOs Morphological and pharmacological tumor heterogeneity was analyzed among PDOs of distinct metastatic lesions from each of 66 patients (n = 175 PDOs; Fig. 6 a). Most patients had PDOs of the same morphological phenotype (n = 40, 61%). However, the mixed group represents an inherently heterogeneous phenotype, and beyond the 18% of patients (n = 12) with exclusively mixed PDOs, another 23% (n = 15) had inter-metastatic heterogeneity with a combination of mixed and either cystic or solid PDOs. Furthermore, heterogeneity of the cystic and solid phenotypes was observed in 12% of patients (n = 8), and a morphological phenotype switch after recurrence was observed in one of six evaluable patients (Fig. 6 a). Heterogeneity of drug sensitivities was less prominent in intra-patient than inter-patient comparisons, evaluated as lower pharmacological heterogeneity scores between patient-matched sample pairs and estimated based on Euclidean distances of DSS values (n = 24 overlapping drugs; Fig. 6 b; p < 0.0001 by one-way ANOVA). This was supported by hierarchical clustering analysis, showing clustering of PDOs from the same patient (Supplementary Fig. 4). There was no difference in the level of spatial heterogeneity among lesions from the same resection and longitudinal heterogeneity among recurrent lesions sampled at re-resections (Fig. 6 b). The intra-patient pharmacological heterogeneity score was not associated with any clinicopathological variable or patient survival (Supplementary Table 11). The solid PDOs had lower mean DSS values than the cystic in all patients with morphological heterogeneity except one (p = 0.027, Wilcoxon paired test; Fig. 6 c), indicating that morphological heterogeneity promoted pharmacological heterogeneity. This was supported by higher pharmacological heterogeneity scores among PDOs in patients with morphological heterogeneity (p = 0.017 by Welch’s t-test, Fig. 6 d). However, the highest pharmacological heterogeneity score was found in a patient with three cystic PDOs and heterogeneous APC R213X mutation status (Fig. 6 e; Extended Data Fig. 9a). The mutated PDO was less sensitive to several chemotherapies (5-FU, methotrexate, gemcitabine) and more sensitive to the SMAC mimetic LCL161 (Extended Data Fig. 9b). Notably, all three PDOs were highly sensitive to the EGFR inhibitor afatinib (within the top 3rd -11th percentiles among all PDOs) and moderately sensitive to combination chemotherapy with FLIRI (25th -55th percentiles), and this corresponded with similar clinical responses to neoadjuvant FOLFIRI plus an anti-EGFR antibody in all three lesions (partial response with 35–41% reduction of tumor sizes; Extended Data Fig. 9a). There were no clear differences in the level of intra-patient pharmacological heterogeneity among drugs or classes of drugs, estimated as the maximum DSS difference for each drug between any pair of PDOs per patient (Extended Data Fig. 9c; the low heterogeneity of encorafenib and idasanutlin was likely due to low drug activity in general). The highest heterogeneity estimates (> 95th percentile) were evenly distributed across drugs (median of 3 high scores per drug, 95% confidence interval 2–5), supporting that pharmacological heterogeneity is more dependent on PDO biology than on the specific drug. Prospective treatment selection based on ex vivo pharmacogenomics Two patients with promising ex vivo data for standard or well-tested treatments for metastatic CRC were treated according to their pharmacogenomics profile. Drug nominations were made relative to a reference dataset of the PDOs screened with the largest drug library (n = 114 PDOs from 55 patients screened with lib2 of 47 drugs). The first patient (Pt137) was a man with known history of ulcerative colitis who was diagnosed with rectal cancer and synchronous CRLMs. He had stable disease after first-line treatment with FOLFIRI plus the antiangiogenic agent bevacizumab, and mixed response among metastatic lesions to subsequent neoadjuvant treatment with FOLFOXIRI plus bevacizumab (Fig. 7 a). PDO culturing was attempted for five resected CRLMs and the primary tumor, but the two liver lesions with response to neoadjuvant treatment were necrotic and did not propagate ex vivo (Extended Data Fig. 1 a). The dual EGFR-ERBB2 inhibitor lapatinib was the most active drug in the four established PDOs (highest mean DSS), although with considerable inter-tumor heterogeneity (ranked in the top 2nd -26th percentiles relative to the reference; Extended Data Fig. 10a). Molecular profiling confirmed ERBB2 amplification and RAS wild-type status in all tissue samples and PDOs analyzed, but in situ expression of ERBB2 was heterogeneous and corresponded with the heterogeneous ex vivo sensitivity to lapatinib (lowest in the primary tumor and highest in the T5 CRLM). The patient had recurrence in the liver after resection, and recurrent CRLMs progressed on 3rd line treatment with FOLFIRI plus the anti-EGFR antibody cetuximab. Based on the pharmacogenomics data and previous clinical studies 39 , 40 , the patient was treated with a combination of the two anti-ERBB2 antibodies trastuzumab and pertuzumab in the 4th line. However, the CRLMs progressed on the first evaluation after four treatment cycles, likely attributed to the heterogeneity of ERBB2 expression and/or an ERBB2 S310F mutation later detected in all the PDOs and corresponding tumor tissue samples and potentially associated with pertuzumab resistance 41 , 42 . The second patient (Pt160) was a man diagnosed with sigmoidal colon cancer and synchronous CRLMs ( KRAS G 12 A mutated). He underwent simultaneous surgery of the primary tumor and metastases after four cycles of neoadjuvant treatment with FOLFOX (Fig. 7 b). The patient continued FOLFOX after surgery, until recurrence with non-resectable liver and lymph node metastases. The treatment was switched to FOLFIRI plus bevacizumab (totally 16 cycles over 273 days), which resulted in initial disease stabilization in the liver, except one small new metastasis (< 5 mm) detected after nine cycles (remained stable for three additional cycles). Following a brief treatment break, four additional cycles of FOLFIRI plus bevacizumab were given due to progression in the lungs and adrenal gland. Magnetic resonance imaging confirmed stable disease in the liver and progression in the lungs after completing the 16 treatment cycles. The patient subsequently progressed also on 3rd line chemotherapy and experimental treatment with immune checkpoint inhibitors. PDOs were successfully established from biopsies of two CRLMs taken after 3rd line chemotherapy but before immunotherapy. Both PDOs had strong ex vivo sensitivity to SN-38 (top 1st -3rd percentile relative to reference; Extended Data Fig. 10b) and combinations of SN-38 with 5-FU and leucovorin (preclinical FLIRI; top 1st -6th percentile) and the PLK1 inhibitor volasertib (top 1st -2nd percentile). FOLFIRI was re-administered in the 5th line based on the ex vivo drug sensitivity data, 300 days after completing FOLFIRI and bevacizumab in the 2nd line (Fig. 7 b). The evaluation after 47 days and 3 cycles of FOLFIRI showed partial response of the two largest liver lesions and a reduction of most lung metastases (50% reduction of the total disease burden). Treatment with FOLFIRI was continued for an additional 60 days until progression of the lung metastases. Discussion Patients with CRC have limited benefit from precision medicine guided by cancer genome sequencing 43 , 44 . This study reports the preclinical development of a functional oncology platform that is the basis for an ongoing intervention trial of metastatic CRC (NCT05725200). Clinical translation was illustrated by successful rechallenge with combination chemotherapy guided by ex vivo drug sensitivity testing. Rechallenge with first-line therapies can provide a survival benefit over other late-line treatment options 45 , 46 . However, studies of anti-EGFR therapies highlight the need for treatment guidance based on monitoring of resistance clones in liquid biopsies 47 , and there are currently no biomarkers to guide chemotherapy rechallenge in a similar manner. This study proposed drug sensitivity testing of PDOs as a potential approach. The turnaround time from sampling to a completed drug screen is a concern with diagnostic use of PDOs. In this respect, it is worth noting that rechallenge in the 5th line was successfully guided by biopsies taken before start of 4th line treatment. However, 4th line treatment had no effect in this patient, and dedicated longitudinal studies are needed to address whether response and/or exposure to intervening treatment has a potentially modifying effect on the predictive power of PDOs for subsequent treatment lines. The other example of ex vivo pharmacogenomics-guided treatment highlighted the impact of tumor heterogeneity. This patient showed heterogeneous ex vivo sensitivity to ERBB2 inhibition in a genetic background of ERBB2 amplified tumors and derived no clinical benefit from anti-ERBB2 treatment. The ex vivo drug sensitivities reflected closely the heterogeneous expression levels of ERBB2, and the clinical data were consistent with a previous study showing that ERBB2 amplification does not confer treatment response without protein overexpression 39 . Notably, the patient received targeted treatment for recurrent metastases without analyses of the recurrent lesions, and the relative impact of protein expression levels and co-occurring genetic aberrations ( ERBB2 amplification and S310 mutation) on the treatment outcome was not clear. Across the living biobank, patients showed large variation in the level of inter-metastatic heterogeneity of ex vivo drug sensitivities, and we found no strong clinicopathological or molecular correlates to potentially predict this heterogeneity. Furthermore, heterogeneity was observed for all drugs and with no apparent proclivity for heterogeneity of specific drug classes. However, no patient had heterogeneous sensitivity to all drugs, indicating potential to bypass vulnerability to heterogeneity by selecting treatments with similar activity across lesions. This was illustrated with the most heterogeneous patient, who showed large variation in ex vivo sensitivities to several drugs but similar sensitivity to standard of care agents, consistent with clinical response to the corresponding treatment in all lesions. These data highlight the importance of performing functional precision oncology in the context of tumor heterogeneity, although profiling of multiple samples collected at a single timepoint cannot account for tumor dynamics and potential heterogeneity of adaptive treatment responses and acquired resistance. Multi-level data analyses suggested associations of PDO morphologies, molecular profiles and drug sensitivities, supporting the presence of multimodal phenotypes of CRCs. Specifically, PDOs with a cystic morphology had higher drug sensitivities in general and expressed the intestinal lineage marker CDX2 48 , as well as stem cell signatures of the bottom of the colonic crypt 30 . Positive correlation of CDX2 expression with sensitivity to several drugs, including standard chemotherapies, is consistent with benefit from 1st line chemotherapy in CDX2-positive metastatic CRCs 49 , 50 . However, studies of adjuvant treatment in stage II-III CRCs have suggested that loss of CDX2 is associated with chemotherapy benefit 51 , 52 . This inconsistency is not likely attributable to marker heterogeneity, since CDX2 expression is consistent between primary and metastatic tumors 53 , but context dependency can be rationalized by an indirect effect of CDX2 on chemosensitivity by regulating genes involved in drug efflux and metabolism 54 . Compared to cystic PDOs, the solid morphology showed several features of cancer aggressiveness and expressed differentiation markers associated with the top of colonic crypts, including KRT20 55,56 . KRT20 might mark a plastic cellular phenotype with capacity for de-differentiation into stem-like cellular states, which is a hallmark of cancer development and therapy resistance 57 , 58 . Notably, the PDO culturing conditions induce stem cell behavior and might have impacted this distinction between the two PDO phenotypes 19 , although approximately two thirds of the PDOs recapitulated the histological morphologies of the original tumors. We did not evaluate the cystic and solid PDO phenotypes in relation to the consensus histopathological growth patterns of CRLMs, since the desmoplastic and replacement growth patterns are primarily defined at the interface between the tumors and liver parenchyma 59 . Spatial transcriptomics has suggested that the growth patterns have distinct expression markers in the cancer cell compartment 60 , but we found no consistent distinctions of these markers between the cystic and solid PDOs (data not shown). Additional studies are needed to potentially consolidate the PDO-derived and tumor-derived histopathological phenotypes, including a potential association between the favorable prognosis of the desmoplastic growth pattern 61 and the stronger drug sensitivities of cystic PDOs. The living biobank supported a favorable prognosis of patients with strong ex vivo drug sensitivities. Comprehensive analyses of sample pairs indicated strong molecular correspondence of PDOs and their original CRLMs, but also highlighted known limitations of cancer stem cell-derived cultures. Beyond the failure to model the tumor microenvironment and its influence on cancer drug activity 62 , the PDOs showed signs of adaptation to culturing conditions, with higher metabolic and proliferative activity. It has been shown that the cellular states of PDOs can to some extent be controlled by manipulation of growth factors in the culture media 19 . Careful adaptation of culturing conditions for features such as physiological hypoxia and niche factors can increase the culture success rate and even improve the fidelity of organoids as cancer models 7 , 63 . In this study, standardization of protocols was favored over sample-wise optimization, as standardization is an important consideration in clinical translation of functional precision oncology 64 . Furthermore, a large proportion of patients received chemotherapy prior to resection and sampling, and successful ex vivo culturing appeared to be dependent on poor chemosensitivity. This resulted in selection of cells and patients in the living biobank, including selection for poor-prognosis patients, but also reflected a clinically relevant situation where functional precision oncology is used to identify new treatment options after development of chemoresistance. A limitation of the study was the lack of co-clinical evaluation of subsequent treatments given to the patients. Radiological images were analyzed for the sampled specimens in a lesion-wise manner, but longitudinal radiological data of recurrent lesions were not available to evaluate the prospective predictive value of PDOs in an observational setting. Ultimately, the benefit and representativeness of PDOs as pharmacogenomic models need validation in prospective intervention trials 64 . In conclusion, this study reports the establishment of a living PDO biobank and analyses of multi-modal pharmaco-omics data of resected CRLMs in the context of tumor heterogeneity. The pharmaco-omics data represent a preclinical resource for functional precision oncology and have been made available to the scientific community. The resource also serves as reference for an ongoing intervention trial, to support the interpretation of ex vivo drug sensitivities into clinical “actionability”. As illustrated with examples of prospective treatment selection, top-scoring drugs are nominated only if the ex vivo activity is high in the prospective patient relative to patients in the reference. Furthermore, the study suggests that ex vivo pharmacogenomics of multiple metastatic lesions from each patient can aid the selection of treatments with lower vulnerability to tumor heterogeneity. Materials and Methods Patients and specimens Patients were included from an observational study of liver resection for metastatic CRC at Oslo University Hospital, involving prospective biobanking of metastatic lesions larger than 5 mm 29 , 65 . Patients in this study (n = 132; Supplementary Table 1) were treated between December 2017 and 2021 and selected to increase the number with multiple CRLMs (85% of the included patients). Metastatic lesions were considered distinct when clearly separated from the neighboring lesions in the same or distant liver segments by radiological evaluation and visual examination after surgery. Longitudinal samples of recurrent CRLMs were collected from eight patients treated by one (n = 6) or two (n = 2) hepatic re-resections. Tumor tissue samples were collected immediately after surgery, and parallel samples were either frozen in liquid nitrogen and stored at − 80°C until processing for molecular profiling, or stored and transported in ice-cold basal media for organoid culture within 24 h. Clinicopathological data were prospectively collected from hospital medical records. Patients received systemic treatment according to standard national protocols, and evaluation of radiological response to neoadjuvant treatment was performed from magnetic resonance imaging (MRI) and/or computed tomography images in a lesion-wise manner. The last examination before start of chemotherapy was used as baseline and compared to the first examination after finishing each line of chemotherapy. The same imaging modality was used for pre- and post-chemotherapy comparisons if available, and MRI was preferred if both modalities were available. Data registration was finished at the time of resection for collection of PDO. The CRLMs sampled to establish PDOs were identified on images based on the segment localization according to the surgical or radiological description at the time of tissue specimen collection. If there was more than one CRLM in one segment, the largest and/or most central lesion in the segment was considered. The CRLM identification/selection and all measurements of lesion size was performed by one radiologist (KKL), and difficult cases reviewed by another radiologist (TS) and consensus reached. Each tumor was measured in its largest diameter on axial images in mm and response expressed lesion-wise as % size change (in mm) relative to baseline. The study was conducted in accordance with the Declaration of Helsinki and all patients provided written informed consent. The study has been approved by the Norwegian Data Protection Authority and Regional Committee for Medical and Health Research Ethics, South-Eastern Norway (REC numbers 1.2005.1629, 2010/1805, 2017/780). Patient-derived tumor organoids PDOs were established as previously described 29 . In short, tumor specimens were minced with a scalpel, filtered with a 70 µm pore mesh, washed with ice-cold culture media, and collected by centrifugation at 400g, 4°C for 5 min. The tissue pellet was resuspended in Growth Factor Reduced Matrigel (Corning, NY, USA) and dispensed as 25 µl drops in 6-well tissue culture plates, fed with 3 ml organoid growth media supplemented with 10 µM Y-27632 (Merk, Darmstadt, Germany), and incubated at 37°C in a humidified atmosphere with 5% CO 2 . The composition of the organoid growth media was based on a previous study 7 and consisted of advanced DMEM/F-12 (Gibco, MA, USA) supplemented with 10 mM HEPES (Gibco), 2 mM GlutaMAX (Gibco), 50 U/ml penicillin/streptomycin (Gibco), 1x B27 Supplement (Gibco), 10 nM [Leu15]-Gastrin I (Merk), 1 mM N-acetyl-l-cysteine (Merk), 50 ng/mL EGF (Gibco), 100 ng/mL Noggin (Preprotech, NJ, UDA), 500 nM A83-01 (Tocris, Bristol, UK) and 10 µM SB202190 (Merk). Organoids were passaged by enzymatic digestion with TrypLE Express (Gibco) for 5 min − 10 min at 37°C. Organoid growth media was refreshed every two to four days without Y-27632. Each PDO was propagated with more than 1x10 6 organoids for cryopreservation, drug sensitivity assays, immunohistochemistry and sampling for DNA/RNA extraction. MycoAlert Mycoplasma Detection Assay was used to ensure that PDOs were not contaminated (Lonza, Bazel, Switzerland). PDOs were authenticated by comparisons with the original tumor tissue using the AmpFLSTR Identifiler PCR Amplification Kit (Thermo Fisher Scientific, MA, USA). Drug sensitivity screening A medium-throughput drug screen of two drug libraries comprising totally 55 small molecule inhibitors at seven (lib2) or nine (lib1) different concentrations, as well as cetuximab at two (lib1) and five (lib2) concentrations and six drug combinations in seven concentrations was performed in PDOs (Supplementary Table 4, Supplementary Fig. 5). Lib2 (n = 47) is an updated version of lib1 (n = 41), in which inactive drugs and noninformative concentrations have been replaced with new drugs or combinations. The libraries were customized for CRC, and include all agents and combinations used in standard of care, as well as drugs approved for other cancer types and drugs in pre-clinical and clinical testing for CRC or other cancer types. Lib1 was screened in 117 PDOs from 59 patients, and lib2 in 116 PDOs from 55 patients, including 22 PDOs from 12 patients screened with both libraries. Drug dilutions and printing for sensitivity screens were carried out at the FIMM High Throughput Biomedicine Unit, which is hosted by the University of Helsinki and supported by HiLIFE and Biocenter Finland. Drugs were preprinted in 384-well tissue culture plates with the liquid acoustic dispensing technology Echo 550 (Labcyte Inc). The monoclonal antibody cetuximab was added manually right before screening. Each plate included 100 µM benzethonium chloride and 0.1% DMSO as positive and negative controls, respectively. Each PDO was screened in two parallel technical replicates. The drug screens were performed by adding 10 µl of 50% Matrigel in growth media and 30 µl of a 3% Matrigel suspension of 450–600 organoids filtered with 70 µm mesh size to each well of a 384-well plate. PDOs were incubated for 96 h at 37°C in a humidified 5% CO 2 atmosphere. Viability was measured using the CellTiter-Glo 3D Cell Viability Assay (Promega, WI, USA) according to the manufacturer's instructions and luminescence readouts on a Victor 3 microplate reader (Perkin Elmer, Waltham, MA, USA). Drug sensitivity estimates were obtained as previously described 29 . Briefly, CellTiter-Glo luminescence readouts were converted to relative viability estimates by normalization based on the median of the negative and positive control wells. Drug-response curves were modeled using logistic regression function logLogisticRegression in the R package PharmacoGx (v3.0.2) on the viability data points, and drug sensitivity scores (DSS) were calculated from the curves using the R package “DSS” 66 . PDO growth speed was determined based on the CTG 3D assay readouts (n = 121) or mean diameter (measured from micrographs, n = 69) of PDOs after four days of growth in 0.1% DMSO, divided by the baseline viability or mean PDO diameter at the time of seeding, respectively. Drugs with technical variation (variability between drug lots, technical replicates and/or visual inspection of the dose response curves) and drugs with no activity (maxDSS 0.50) were excluded from downstream analysis (Supplementary Table 4). Correlation analysis of DSS values indicated similar activity of drugs with the same molecular target (three EGFR inhibitors, two MEK inhibitors, three PI3K/MTOR inhibitors), mode of action (three inhibitors of ATR, WEE2 or CHK1), as well as drug combinations and their respective single agents (combinations with 5-FU, combinations with SN-38; Supplementary Fig. 6). There was no correlation between the number of lesions analyzed and DSS heterogeneity based on median pair-wise Euclidean distances (Supplementary Fig. 8e-f). DNA/RNA extraction PDOs were collected from Matrigel and dissociated into single cells with TrypLE. RNA and DNA were isolated from both the PDOs and their corresponding fresh frozen tumor tissue samples using the Allprep DNA/RNA/miRNA Universal Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The quantity of nucleic acids was determined using a NanoDrop ND-1000 Spectrophotometer, and the quality of the RNA was assessed using a 2100 Bioanalyzer (RNA 6000 Nano kit; Agilent Technologies, Santa Clara, CA, USA). All tumor samples were found to be microsatellite stable (MSS) by the PCR-based MSI Analysis System, Version 1.2 (Promega #MD1641). Mutation analyses Sequencing was performed with a custom gene panel (Twist Bioscience, San Francisco, CA, USA) targeting all coding regions of APC, TP53, KRAS, NRAS, BRAF, ARID1A, PIK3CA, PTEN, EGFR, ERBB2, FBXW7, JAK1, MYC, NF1, CTNNB1, SMAD4 , SMAD2, CCND2 , and MDM2 , as well as selected pathogenic exonuclease domain mutations of POLE (n = 11; Supplementary Table 12). Multiple PDOs from patients with homogeneous and heterogenous mutations are shown in Supplementary Fig. 8a. Sequencing libraries were prepared from 50ng genomic DNA using the Library Preparation Enzymatic Fragmentation (EF) Kit and target enrichment workflow with DNA purification beads according to the manufacturer's instructions (Twist Bioscience). Sequencing was done on the Illumina MiniSeq system in a 2 x 73 base-pair paired-end mode using the MiniSeq High Output Kit (150-cycle; Illumina, San Diego, CA, USA). Raw sequencing reads were assessed using FastQC (v.0.11.8), before further preprocessing. Alignment to the GRCh38 human reference genome was performed using BWA (v.0.7.17), and file format conversion and refinement of sequencing reads was performed using Picard (2.19.0) and GATK (v4.1.2). Somatic variants were called with MuTect2 and annotated by ANNOVAR (version 2016Feb01). Variant calling was done in either tumor-normal or tumor-only mode depending on the availability of a matched normal tissue sample. In tumor-normal mode, only candidate somatic non-synonymous single nucleotide variants and insertion-deletions annotated as “PASS” or “clustered events” were kept and further filtered to include only loci with a minimum variant allele frequency of 5% and more than 5 reads in the tumor sample. The coverage threshold at each locus was set to a minimum of 15 reads, and a variant allele frequency less than 1.5% was accepted in the normal tissue. In tumor-only mode (98% of the PDOs and 89% of the CRLMs), additional filtering was performed to discard germline variants 67 . The median depth of coverage across the 20 genes was 570X (range 229 to 1033). Gene expression analyses PDOs and matching tumor tissue samples were analyzed for gene expression on Affymetrix Human Transcriptome 2.0 arrays (n = 119 PDOs and 50 matching tissue) or by RNA sequencing (n = 92 PDOs, 5 overlapping with arrays). Seven PDOs were not analyzed due to low RNA yields. Microarray experiments were performed with 100 ng of total RNA as input and following the manufacturer’s protocol (Thermo Fisher Scientific, Waltham, MA, USA). RNA sequencing was performed in 2×101 base-pair paired-end mode on the Illumina NovaSeq 6000 platform (Illumina) at the Oslo University Hospital Genomics Core Facility to a median depth of 71.6 million uniquely mapped read pairs per sample (10-90th percentile 60.6–83.1 million reads). Sample preparation was performed by ribosomal RNA depletion using the Ribo-Zero Gold rRNA Depletion kit and sequence library generation with the TruSeq Stranded Total RNA Library Prep Gold kit (Illumina). Raw intensity CEL-files from microarray experiments were processed in two runs (PDO samples only or PDOs and matching tumor tissue together) according to the robust multi-array average approach 68 , using the function justRMA in the R package affy (v1.80.0) 69 and custom Entrez CDF file from Brainarray (hta20hsentrezgcdf_25.0.0) 70 . Gene annotations according to the GRCh38 genome assembly were retrieved using the function getBM in the R package biomaRt (v2.58.0) 71 .Protein-coding genes annotated with HGNC symbols were retained. Raw RNA sequencing reads were processed by adapter trimming with Trimmomatic (v.0.38), read alignment to the human reference genome GRCh38.p13 (v.41) using STAR (v.2.7.6a) with 2-pass mapping and the feature annotation file gencode.v41.annotation.gtf, as well as quantification of reads mapping to Ensembl gene ids using the HTSeq-count tool (v.2.0.2) 72 . Annotations were converted to HGNC symbols with biomaRt. Count data were normalized as transcripts per kilobase million (TPM). For differential gene expression analyses, count data were normalized with trimmed mean of M-values (TMM) and summarized as counts per million (CPM) using the R package edgeR. Normalized gene expression estimates were log2-transformed (after adding a constant of 0.1 to TPM values). Similarity of PDOs to CRC tissue on the transcriptomic level was evaluated with the R package CancerCellNet (v.0.2.0) using the function broadClass_predict and a classifier trained with CRC tissue samples from TCGA, obtained with the function broadClass_train and default settings 31 . Classification according to iCMS was performed using the approach and gene template described in the original publication 73 . Most PDOs (94%) derived from distinct lesions of the same patient were classified in the same iCMS group (Supplementary Fig. 8b). Principal components analysis was performed using the R package FactoMineR (v2.9) on genes (n = 3,000) with the highest cross-sample 10-90th percentile range. Gene set enrichment analysis of a custom gene set collection (n = 106) was performed with the R package GSA (v1.03.0) 74 for sample group comparisons, including false discovery rate adjustment of the p-value, and the R package GSVA (v1.50.0) 75 for single-sample scoring by the gene set variation analysis approach. Histopathology and multiplex immunohistochemistry PDOs were fixed with 4% paraformaldehyde in PBS for 20 minutes and embedded in the protective gel from Shandon Cytoblock Cell Block Preparation System (Thermo Scientific). Fixed cells were dehydrated, infused with paraffin wax, and embedded in paraffin to create histology blocks. The blocks were cut into 4 µm sections for staining with hematoxylin and eosin (HE) and antibodies. Morphological resemblance between tumor tissues and their corresponding PDOs was evaluated by a pathologist on HE stained sections from 46 paired samples from 26 patients. The comparisons were based on structural patterns of the epithelial cell compartment, including small/acinar and dilated/cystic gland-like structures, cribriform/complex structures, solid growth, cytoplasmatic vacuoles, and presence of single cells. Separately, HE stained sections of all 213-paraffin embedded PDOs were used to determine the morphological cystic, solid and mixed phenotypes as described in the results. Fluorescence-based multiplex immunohistochemistry and digital image analyses were used to analyze in situ expression of fourteen proteins in 136 PDOs and two corresponding tumor tissue samples from 67 patients (Supplementary Fig. 7a-d). Fluorescence staining was based on Opal kits (NEL810001KT (includes fluorophores Opal 520, 570 and 690, DAPI, antibody diluent and anti-mouse/rabbit HRP secondary antibodies) and FP1495001KT (Opal 620) from Akoya Biosciences (Marlborough, MA, USA) and reagents were used according to the manufacturer's recommendations (Akoya Biosciences), unless otherwise noted below. Five multiplex stains were developed using primary antibodies against the following targets (see Supplementary Table 13 for staining sequence and pairing/dilution of fluorophores): CDX2 (1:400, clone EPR2764Y, Cell Marque, CA, USA), KRT20 (1:400 and 1:1000, clone Ks20.8, Agilent Dako, Glostrup, Denmark), CDH1 (1:10.000, clone 36, BD Biosciences, NJ, USA), KRT7 (1:400, clone OV-TL 12/30, Agilent Dako), KI67 (undiluted, clone MIB-1, DAKO/Agilent), HSF1 (1:100 clone D3L8I, cell signaling technology, MA, USA), UGT1A (1:300 and 1:500, clone B-4, Santa Cruz, CA, USA), RIPK1 (1:50 clone E8S7U, cell signaling technology), TP53 (1:12, clone DO-7, Agilent Dako), RCC2 (1:100 clone D14F3, cell signaling technology), ABCG2 (1:50, clone D5V2K, cell signaling technology), CFTR (1:2000, clone 24 − 1, R&D systems, MN, USA), ERBB2 (1:500, polyclonal, catalogue number AO485, DAKO/Agilent), and ABCB1 (1:100, clone E1Y7S, cell signaling technology). Deparaffinization and the initial antigen retrieval were performed in a PT-link module (DAKO/Agilent) at 97 o C for 20 min, using 3-in-1 high-pH buffer (catalogue number K8004, DAKO/Agilent). The following rounds of heat treatment for antibody stripping were also performed in the PT-link module, however instead using high/low pH buffers from Akoya (catalogue numbers AR9001KT & AR6001KT, respectively), specified in Supplementary Table 13. All primary antibodies were incubated for 30 min at room temperature. Cell nuclei were stained with DAPI prior to mounting with Prolong Diamond Antifade Mountant (Life Technologies/Thermo Fisher Scientific). The sections were scanned at 10x magnification, and multispectral images were acquired at 20x magnification using the Vectra 3 Automated Quantitative Pathology Imaging System (Akoya Biosciences). Image analysis was performed using inForm Image Analysis Software (Akoya Biosciences), which used a supervised machine learning algorithm to accurately segment the cells and nuclei. All images were manually checked after segmentation and poor-quality regions of the samples were excluded (e.g. due to sample folds). Mean relative expression of each protein was quantified and normalized to the total cellular content in each PDO image. KRT7 and ERBB2 were excluded from analysis due to no or very low expression levels, respectively. Protein expression heterogeneity was less pronounced in intra-patient than inter-patient comparisons, as indicated by lower heterogeneity scores between patient-matched sample pairs, calculated using Euclidean distances of protein expression profiles (Supplementary Fig. 8c). There was no correlation between the number of lesions analyzed and protein expression heterogeneity based on median pair-wise Euclidean distances (Supplementary Fig. 8d). ERBB2 protein expression (Fig. 7 a) was also visualized using the DAB chromogen on the Autostainer Link 48 system. Deparaffinization and antigen retrieval were performed in the PT-link module, as stated above, except that low-pH buffer was used (catalogue number K8005, DAKO/Agilent). Primary antibody against ERBB2 (1:50, clone CB11, Leica Biosystems, Nussloch, Germany) was incubated for 30 min at room temperature. Anti-mouse EnVision + System HRP labelled polymer (catalogue number K4001, DAKO/Agilent) was incubated for 30 min prior to incubation with DAB chromogen (catalogue number K3468, DAKO/Agilent) for 10 min. Hematoxylin (catalogue number 01800, Histolab) was used as counterstain and incubated for 30 sec. Slides were rinsed in water before dehydrating and mounting. Statistics Statistical analyses and data visualization were performed with R software version 4.2.1, Graph Pad Prism version 10.2.2 or SPSS version 29. Clinicopathological variables were summarized and compared using the "gtsummary" v1.7.0 R package. Chi-squared tests and visualization were performed with the "ggstatsplot" v0.11.0 package in R. Spearman and Pearson correlation analyses were performed using the "cor" function. Wilcoxon tests were performed using the "wilcox_test" function from the "rstatix" v0.7.2 R package. Kruskal–Wallis tests of three groups of continuous variables were performed using the “ggpubr” v0.6.0 R package. Euclidean distances were calculated using the dist function in R. Heatmaps were generated using the "pheatmap" v1.0.12 R package with the "complete" clustering method. Survival analyses were conducted using the "survival" v3.3-5 and "survminer" v0.4.9 R packages. Overall survival was estimated from the date of liver surgery and using death from any cause as events. Patients still alive at the end of the study were censored at the last follow-up or four years after liver surgery. Cox proportional hazards models were used to evaluate prognostic associations of clinicopathological variables. P-values lower than 0.05 were considered statistically significant. All statistical tests were two-sided. Declarations Data availability All data generated and analyzed in this study are available in public repositories or as Data Sets in the manuscript. Mutation data (n=20 genes in 150 PDOs and 74 tumor tissue samples) are available as Data Set 1. The microarray gene expression data (n=119 PDOs and 50 tumor tissue samples) have been deposited to the NCBI’s Gene Expression Omnibus under accession code GSE294511 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE294511]. The raw RNA sequencing data (n=92 PDOs) are considered patient identifiable and subject to secure storage regulations in accordance with Norwegian legislation and the ethical approval of the study by the Regional Committee for Medical and Health Research Ethics, South Eastern Norway, but count data of protein-coding genes are available as Data Set 2. Protein expression data from multiplex immunohistochemistry (n=14 proteins in 136 PDOs from 67 patients) are available as Data Set 3. Drug sensitivity scores (library 1: n=41 drugs in 117 PDOs; and library 2: n=47 drugs in 116 PDOs) are available as Data Set 4. Code availability All data processing and analyses were performed with published software packages and computer code and have been described and cited in the Results and/or Methods sections. No custom code was developed in the study. Acknowledgments We would like to thank the study nurses Magdalena Maria Kowalewska-Harbiyeli and Vlora Krasniqi Hulaj for their help in collecting patient samples and clinical data. We also appreciate the technical support provided by the principal engineer Mette Eknæs and head engineer Merete Hektoen. Additionally, we acknowledge the contributions of former lab members Barbara Niederdorfer, Jonas Langerud, Jarle Bruun, Peter W. Eide, Christer A. Andreassen and Kaja C. G. Berg in the initial stages of this project. The study was funded by grants from the Norwegian Cancer Society (project numbers 182759 and 223319 to RAL, project numbers 208336 and 246954 to AS, project number 297971 to KK), the South-Eastern Norway Regional Health Authority (project number 2023101 to AS and 2024108 to RAL), the Research Council of Norway (project number 287899 to AS), and the Oslo University Hospital (“Strategic research area, 2019-2024” to RAL and AS). References van Renterghem, A.W.J., van de Haar, J. & Voest, E.E. Functional precision oncology using patient-derived assays: bridging genotype and phenotype. Nat Rev Clin Oncol 20 , 305-317 (2023). Kornauth, C. et al. Functional Precision Medicine Provides Clinical Benefit in Advanced Aggressive Hematologic Cancers and Identifies Exceptional Responders. Cancer Discov 12 , 372-387 (2022). Malani, D. et al. Implementing a Functional Precision Medicine Tumor Board for Acute Myeloid Leukemia. Cancer Discov 12 , 388-401 (2022). Ooft, S.N. et al. Prospective experimental treatment of colorectal cancer patients based on organoid drug responses. ESMO Open 6 , 100103 (2021). Sato, T. et al. Long-term expansion of epithelial organoids from human colon, adenoma, adenocarcinoma, and Barrett's epithelium. Gastroenterology 141 , 1762-72 (2011). van de Wetering, M. et al. Prospective derivation of a living organoid biobank of colorectal cancer patients. Cell 161 , 933-45 (2015). Fujii, M. et al. A Colorectal Tumor Organoid Library Demonstrates Progressive Loss of Niche Factor Requirements during Tumorigenesis. Cell Stem Cell 18 , 827-38 (2016). Ooft, S.N. et al. Patient-derived organoids can predict response to chemotherapy in metastatic colorectal cancer patients. Sci Transl Med 11 (2019). Vlachogiannis, G. et al. Patient-derived organoids model treatment response of metastatic gastrointestinal cancers. Science 359 , 920-926 (2018). Narasimhan, V. et al. Medium-throughput Drug Screening of Patient-derived Organoids from Colorectal Peritoneal Metastases to Direct Personalized Therapy. Clin Cancer Res 26 , 3662-3670 (2020). Letai, A. Functional Precision Medicine: Putting Drugs on Patient Cancer Cells and Seeing What Happens. Cancer Discov 12 , 290-292 (2022). Jensen, L.H. et al. Precision medicine applied to metastatic colorectal cancer using tumor-derived organoids and in-vitro sensitivity testing: a phase 2, single-center, open-label, and non-comparative study. J Exp Clin Cancer Res 42 , 115 (2023). McGranahan, N. & Swanton, C. Clonal Heterogeneity and Tumor Evolution: Past, Present, and the Future. Cell 168 , 613-628 (2017). Dagogo-Jack, I. & Shaw, A.T. Tumour heterogeneity and resistance to cancer therapies. Nat Rev Clin Oncol 15 , 81-94 (2018). Hu, Z. et al. Quantitative evidence for early metastatic seeding in colorectal cancer. Nat Genet 51 , 1113-1122 (2019). Dang, H.X. et al. The clonal evolution of metastatic colorectal cancer. Sci Adv 6 , eaay9691 (2020). Cai, J. et al. Single-cell exome sequencing reveals polyclonal seeding and TRPS1 mutations in colon cancer metastasis. Signal Transduct Target Ther 9 , 247 (2024). Langerud, J. et al. Multiregional transcriptomics identifies congruent consensus subtypes with prognostic value beyond tumor heterogeneity of colorectal cancer. Nat Commun 15 , 4342 (2024). Moorman, A. et al. Progressive plasticity during colorectal cancer metastasis. Nature 637 , 947-954 (2025). Brunsell, T.H. et al. Heterogeneous radiological response to neoadjuvant therapy is associated with poor prognosis after resection of colorectal liver metastases. Eur J Surg Oncol 45 , 2340-2346 (2019). Zhou, J. et al. Mapping lesion-specific response and progression dynamics and inter-organ variability in metastatic colorectal cancer. Nat Commun 14 , 417 (2023). Ou, F.S. et al. Evaluation of Intratumoral Response Heterogeneity in Metastatic Colorectal Cancer and Its Impact on Patient Overall Survival: Findings from 10,551 Patients in the ARCAD Database. Cancers (Basel) 15 (2023). Geevimaan, K. et al. Patient-Derived Organoid Serves as a Platform for Personalized Chemotherapy in Advanced Colorectal Cancer Patients. Front Oncol 12 , 883437 (2022). Schumacher, D. et al. Heterogeneous pathway activation and drug response modelled in colorectal-tumor-derived 3D cultures. PLoS Genet 15 , e1008076 (2019). Roerink, S.F. et al. Intra-tumour diversification in colorectal cancer at the single-cell level. Nature 556 , 457-462 (2018). Kim, S.C. et al. Multifocal Organoid Capturing of Colon Cancer Reveals Pervasive Intratumoral Heterogenous Drug Responses. Adv Sci (Weinh) , e2103360 (2021). Mo, S. et al. Patient-Derived Organoids from Colorectal Cancer with Paired Liver Metastasis Reveal Tumor Heterogeneity and Predict Response to Chemotherapy. Adv Sci (Weinh) 9 , e2204097 (2022). Thng, D.K.H. et al. A functional personalised oncology approach against metastatic colorectal cancer in matched patient derived organoids. NPJ Precis Oncol 8 , 52 (2024). Bruun, J. et al. Patient-Derived Organoids from Multiple Colorectal Cancer Liver Metastases Reveal Moderate Intra-patient Pharmacotranscriptomic Heterogeneity. Clin Cancer Res 26 , 4107-4119 (2020). Joanito, I. et al. Single-cell and bulk transcriptome sequencing identifies two epithelial tumor cell states and refines the consensus molecular classification of colorectal cancer. Nat Genet 54 , 963-975 (2022). Peng, D. et al. Evaluating the transcriptional fidelity of cancer models. Genome Med 13 , 73 (2021). Vecchione, L. et al. A Vulnerability of a Subset of Colon Cancers with Potential Clinical Utility. Cell 165 , 317-30 (2016). Yaeger, R. et al. Clinical Sequencing Defines the Genomic Landscape of Metastatic Colorectal Cancer. Cancer Cell 33 , 125-136 e3 (2018). Mendelaar, P.A.J. et al. Whole genome sequencing of metastatic colorectal cancer reveals prior treatment effects and specific metastasis features. Nat Commun 12 , 574 (2021). Bergsland, C.H. et al. Prediction of relapse-free survival according to adjuvant chemotherapy and regulator of chromosome condensation 2 (RCC2) expression in colorectal cancer. ESMO Open 5 , e001040 (2020). Smeby, J. et al. Molecular correlates of sensitivity to PARP inhibition beyond homologous recombination deficiency in pre-clinical models of colorectal cancer point to wild-type TP53 activity. EBioMedicine 59 , 102923 (2020). Kryeziu, K. et al. Increased sensitivity to SMAC mimetic LCL161 identified by longitudinal ex vivo pharmacogenomics of recurrent, KRAS mutated rectal cancer liver metastases. J Transl Med 19 , 384 (2021). Arnold, D. et al. Prognostic and predictive value of primary tumour side in patients with RAS wild-type metastatic colorectal cancer treated with chemotherapy and EGFR directed antibodies in six randomized trials. Ann Oncol 28 , 1713-1729 (2017). Meric-Bernstam, F. et al. Pertuzumab plus trastuzumab for HER2-amplified metastatic colorectal cancer (MyPathway): an updated report from a multicentre, open-label, phase 2a, multiple basket study. Lancet Oncol 20 , 518-530 (2019). Gupta, R. et al. Pertuzumab Plus Trastuzumab in Patients With Colorectal Cancer With ERBB2 Amplification or ERBB2/3 Mutations: Results From the TAPUR Study. JCO Precis Oncol 6 , e2200306 (2022). Zhang, Y. et al. Identification of an Activating Mutation in the Extracellular Domain of HER2 Conferring Resistance to Pertuzumab. Onco Targets Ther 12 , 11597-11608 (2019). Diwanji, D. et al. Structures of the HER2-HER3-NRG1beta complex reveal a dynamic dimer interface. Nature 600 , 339-343 (2021). Di Nicolantonio, F. et al. Precision oncology in metastatic colorectal cancer - from biology to medicine. Nat Rev Clin Oncol 18 , 506-525 (2021). Sveen, A., Kopetz, S. & Lothe, R.A. Biomarker-guided therapy for colorectal cancer: strength in complexity. Nat Rev Clin Oncol 17 , 11-32 (2020). Bazarbashi, S. et al. Efficacy of Chemotherapy Rechallenge Versus Regorafenib or Trifluridine/Tipiracil in Third-Line Setting of Metastatic Colorectal Cancer: A Multicenter Retrospective Comparative Study. JCO Glob Oncol 10 , e2300461 (2024). Sartore-Bianchi, A. et al. Circulating tumor DNA to guide rechallenge with panitumumab in metastatic colorectal cancer: the phase 2 CHRONOS trial. Nat Med 28 , 1612-1618 (2022). Ciardiello, D. et al. The role of anti-EGFR rechallenge in metastatic colorectal cancer, from available data to future developments: A systematic review. Cancer Treat Rev 124 , 102683 (2024). Badia-Ramentol, J. et al. The prognostic potential of CDX2 in colorectal cancer: Harmonizing biology and clinical practice. Cancer Treat Rev 121 , 102643 (2023). Zhang, B.Y. et al. Lack of Caudal-Type Homeobox Transcription Factor 2 Expression as a Prognostic Biomarker in Metastatic Colorectal Cancer. Clin Colorectal Cancer 16 , 124-128 (2017). Aasebo, K. et al. CDX2: A Prognostic Marker in Metastatic Colorectal Cancer Defining a Better BRAF Mutated and a Worse KRAS Mutated Subgroup. Front Oncol 10 , 8 (2020). Dalerba, P. et al. CDX2 as a Prognostic Biomarker in Stage II and Stage III Colon Cancer. N Engl J Med 374 , 211-22 (2016). Bruun, J. et al. Prognostic, predictive, and pharmacogenomic assessments of CDX2 refine stratification of colorectal cancer. Mol Oncol 12 , 1639-1655 (2018). Shigematsu, Y. et al. CDX2 expression is concordant between primary colorectal cancer lesions and corresponding liver metastases independent of chemotherapy: a single-center retrospective study in Japan. Oncotarget 9 , 17056-17065 (2018). Delhorme, J.B. et al. CDX2 controls genes involved in the metabolism of 5-fluorouracil and is associated with reduced efficacy of chemotherapy in colorectal cancer. Biomed Pharmacother 147 , 112630 (2022). Chan, C.W. et al. Gastrointestinal differentiation marker Cytokeratin 20 is regulated by homeobox gene CDX1. Proc Natl Acad Sci U S A 106 , 1936-41 (2009). Lugli, A., Tzankov, A., Zlobec, I. & Terracciano, L.M. Differential diagnostic and functional role of the multi-marker phenotype CDX2/CK20/CK7 in colorectal cancer stratified by mismatch repair status. Mod Pathol 21 , 1403-12 (2008). Shimokawa, M. et al. Visualization and targeting of LGR5(+) human colon cancer stem cells. Nature 545 , 187-192 (2017). Mehta, A. & Stanger, B.Z. Lineage Plasticity: The New Cancer Hallmark on the Block. Cancer Res 84 , 184-191 (2024). Latacz, E. et al. Histopathological growth patterns of liver metastasis: updated consensus guidelines for pattern scoring, perspectives and recent mechanistic insights. Br J Cancer 127 , 988-1013 (2022). Fleischer, J.R. et al. Molecular differences of angiogenic versus vessel co-opting colorectal cancer liver metastases at single-cell resolution. Mol Cancer 22 , 17 (2023). Stremitzer, S. et al. Immune phenotype and histopathological growth pattern in patients with colorectal liver metastases. Br J Cancer 122 , 1518-1524 (2020). Jin, M.Z. & Jin, W.L. The updated landscape of tumor microenvironment and drug repurposing. Signal Transduct Target Ther 5 , 166 (2020). Walaas, G.A. et al. Physiological hypoxia improves growth and functional differentiation of human intestinal epithelial organoids. Front Immunol 14 , 1095812 (2023). Bose, S. et al. A path to translation: How 3D patient tumor avatars enable next generation precision oncology. Cancer Cell 40 , 1448-1453 (2022). Brunsell, T.H. et al. High Concordance and Negative Prognostic Impact of RAS/BRAF/PIK3CA Mutations in Multiple Resected Colorectal Liver Metastases. Clin Colorectal Cancer 19 , e26-e47 (2020). Yadav, B. et al. Quantitative scoring of differential drug sensitivity for individually optimized anticancer therapies. Sci Rep 4 , 5193 (2014). Moosavi, S.H. et al. Molecular prognostic factors for liver transplantation of unresectable metastatic colorectal cancer. Br J Cancer (2025). Irizarry, R.A. et al. Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 4 , 249-264 (2003). Gautier, L., Cope, L., Bolstad, B.M. & Irizarry, R.A. affy--analysis of Affymetrix GeneChip data at the probe level. Bioinformatics 20 , 307-15 (2004). Dai, M. et al. Evolving gene/transcript definitions significantly alter the interpretation of GeneChip data. Nucleic Acids Res 33 , e175 (2005). Durinck, S., Spellman, P.T., Birney, E. & Huber, W. Mapping identifiers for the integration of genomic datasets with the R/Bioconductor package biomaRt. Nat. Protoc. 4 , 1184-1191 (2009). Eilertsen, I. et al. Technical differences between sequencing and microarray platforms impact transcriptomic subtyping of colorectal cancer. Cancer Lett. 469 , 246-255 (2020). Joanito, I. et al. Single-cell and bulk transcriptome sequencing identifies two epithelial tumor cell states and refines the consensus molecular classification of colorectal cancer. Nat. Genet. 54 , 963-975 (2022). Efron, B. & Tibshirani, R. On testing the significance of sets of genes. Ann. Appl. Stat. 1 , 107-129 (2007). Hanzelmann, S., Castelo, R. & Guinney, J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics 14 , 7 (2013). Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryTablesNATCANCER.pdf Supplementary Table 1 to Table 13 DataSet1Mutations150PDOs74CRLMs.xlsx Data Set 1 DataSet2PDO92HTSeqcountproteincoding.xlsx Data Set 2 DataSet3ProteinsmIHC136PDOs.xlsx Data Set 3 DataSet4DSRTRAW211PDOs.xlsx Data Set 4 SupplFigsNATCANCER.pdf Supplementary Figure 1 to Figure 8 ExtendedDataFigureLegends.docx Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6507406","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Resource","associatedPublications":[],"authors":[{"id":451035180,"identity":"d6b9031b-1031-4c7e-8c09-286cc6c54377","order_by":0,"name":"Ragnhild Lothe","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7UlEQVRIie3RsWoCMRzH8R8EvOVfXO+o+Ax/cRAnXyV5gJMDl5uOg0IdXQ/ah6gc6GgkQ5fzAUSHm1zqkL0O5aybBK9bh3wIGUK+hCSA5/1XNSAR5AAxqFnQDxMJliD95ySU+N3/yCgXNWR6mXajL1Wfk6w3Ag20xbHvSnq6w5AVz6K3uBy8s6FxTrwtcBq6khBgqFdWH4d49Uysr8MQjMqdSWCvyWa/W38TZ20Sup0SPq0EsWiRCEq0rIaqqOIyau7CppNsCzbuuwTzZW3TvlrMd0t7vmQT/nwprU2N88Ug7j5ONBM7A8/zPK+FH4hkU4V2P/tdAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0002-6633-4818","institution":"Oslo University Hospital / Institute for Cancer Research","correspondingAuthor":true,"prefix":"","firstName":"Ragnhild","middleName":"","lastName":"Lothe","suffix":""},{"id":451035181,"identity":"de206dd9-f5ba-426d-93f0-b656432b434e","order_by":1,"name":"Kushtrim Kryeziu","email":"","orcid":"","institution":"Oslo University Hospital / Institute for Cancer Research","correspondingAuthor":false,"prefix":"","firstName":"Kushtrim","middleName":"","lastName":"Kryeziu","suffix":""},{"id":451035182,"identity":"584bb18e-022f-4e75-a647-96dfaae9e379","order_by":2,"name":"Solveig Klokkerud","email":"","orcid":"","institution":"Oslo University Hospital / Institute for Cancer Research","correspondingAuthor":false,"prefix":"","firstName":"Solveig","middleName":"","lastName":"Klokkerud","suffix":""},{"id":451035183,"identity":"93c5a54c-b0d6-4598-a7d5-6ee0dc1e05bb","order_by":3,"name":"Max Totland","email":"","orcid":"","institution":"Oslo University Hospital / Institute for Cancer Research","correspondingAuthor":false,"prefix":"","firstName":"Max","middleName":"","lastName":"Totland","suffix":""},{"id":451035184,"identity":"88181798-b08a-4b0c-a984-8e537970d737","order_by":4,"name":"Seyed Moosavi","email":"","orcid":"","institution":"Oslo University Hospital / Institute for Cancer Research","correspondingAuthor":false,"prefix":"","firstName":"Seyed","middleName":"","lastName":"Moosavi","suffix":""},{"id":451035185,"identity":"b9b38d85-badc-456a-85b0-eb825af476e8","order_by":5,"name":"Henrik Reims","email":"","orcid":"","institution":"Oslo University Hospital / Department of Pathology","correspondingAuthor":false,"prefix":"","firstName":"Henrik","middleName":"","lastName":"Reims","suffix":""},{"id":451035186,"identity":"35f17a47-2fa6-4215-864d-63e18d188ab9","order_by":6,"name":"Christian Bergsland","email":"","orcid":"","institution":"Oslo University Hospital / Institute for Cancer Research","correspondingAuthor":false,"prefix":"","firstName":"Christian","middleName":"","lastName":"Bergsland","suffix":""},{"id":451035187,"identity":"282e8bd1-dd37-4fcf-90a1-45fe0f100e01","order_by":7,"name":"Ina Eilertsen","email":"","orcid":"","institution":"Oslo University Hospital / Institute for Cancer Research","correspondingAuthor":false,"prefix":"","firstName":"Ina","middleName":"","lastName":"Eilertsen","suffix":""},{"id":451035188,"identity":"6e4af1e1-b043-4930-a120-52f0f7f05f77","order_by":8,"name":"Jørgen Smeby","email":"","orcid":"","institution":"Oslo University Hospital / Department of Oncology","correspondingAuthor":false,"prefix":"","firstName":"Jørgen","middleName":"","lastName":"Smeby","suffix":""},{"id":451035189,"identity":"191fef68-3829-4b8f-9774-c9ddc1614a0d","order_by":9,"name":"Kristoffer Lassen","email":"","orcid":"","institution":"Oslo University Hospital / Department of Hepato-Pancreato-Biliary Surgery","correspondingAuthor":false,"prefix":"","firstName":"Kristoffer","middleName":"","lastName":"Lassen","suffix":""},{"id":451035190,"identity":"86b843af-358e-4278-99df-b7f93c4a0a08","order_by":10,"name":"Kristoffer Lundesgaard","email":"","orcid":"","institution":"Oslo University Hospital / Department of Radiology and Nuclear Medicine","correspondingAuthor":false,"prefix":"","firstName":"Kristoffer","middleName":"","lastName":"Lundesgaard","suffix":""},{"id":451035191,"identity":"857dc0ee-56bf-4623-a549-c7dda16fc6e5","order_by":11,"name":"Trygve Syversveen","email":"","orcid":"","institution":"Oslo University Hospital / Department of Radiology and Nuclear Medicine","correspondingAuthor":false,"prefix":"","firstName":"Trygve","middleName":"","lastName":"Syversveen","suffix":""},{"id":451035192,"identity":"628122a7-0de9-4abe-bc2f-f8436b273d85","order_by":12,"name":"Morten Brændengen","email":"","orcid":"","institution":"Diakonhjemmet Hospital / Center for Cancer Medicine","correspondingAuthor":false,"prefix":"","firstName":"Morten","middleName":"","lastName":"Brændengen","suffix":""},{"id":451035193,"identity":"e004c8d0-6177-428a-a73b-dc9b9fd73f32","order_by":13,"name":"Arild Nesbakken","email":"","orcid":"","institution":"Oslo University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Arild","middleName":"","lastName":"Nesbakken","suffix":""},{"id":451035194,"identity":"78cf992a-f8c5-4d10-a974-e09b8f367c0e","order_by":14,"name":"Tormod Guren","email":"","orcid":"","institution":"Oslo University Hospital / Department of Oncology","correspondingAuthor":false,"prefix":"","firstName":"Tormod","middleName":"","lastName":"Guren","suffix":""},{"id":451035195,"identity":"a4efc31f-b61f-4686-9a33-4b26323798bd","order_by":15,"name":"Sheraz Yaqub","email":"","orcid":"","institution":"Oslo University Hospital / Department of Hepato-Pancreato-Biliary Surgery","correspondingAuthor":false,"prefix":"","firstName":"Sheraz","middleName":"","lastName":"Yaqub","suffix":""},{"id":451035196,"identity":"88284e4b-d16d-46f4-b650-e2f2a90e5d87","order_by":16,"name":"Anita Sveen","email":"","orcid":"","institution":"Oslo University Hospital / Institute for Cancer Research","correspondingAuthor":false,"prefix":"","firstName":"Anita","middleName":"","lastName":"Sveen","suffix":""}],"badges":[],"createdAt":"2025-04-22 21:55:42","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6507406/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6507406/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":82348396,"identity":"c7e47024-c16a-4094-897c-4198046217a8","added_by":"auto","created_at":"2025-05-09 10:41:26","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":612216,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLiving biobank of PDOs of resected CRLMs.\u003c/strong\u003e \u003cstrong\u003ea.\u003c/strong\u003e Proportion of sampled lesions with successfully established PDOs per liver segment. The size of the pie charts is proportional to the number of sampled CRLMs per segment. \u003cstrong\u003eb.\u003c/strong\u003e Reasons for \u003cem\u003eex vivo\u003c/em\u003e growth failure of CRLMs (n=133). \u003cstrong\u003ec.\u003c/strong\u003eNumber of successfully established and failed PDOs summarized according to the number of CRLMs attempted to culture per patient. \u003cstrong\u003ed.\u003c/strong\u003e Number of successfully established and failed PDOs per patient, ranked in consecutive order of patient inclusion. Bars separated by black lines indicate CRLMs from one or two hepatic re-resections in eight patients, with the first resection at the bottom. Asterisk marks a CRLM with two multiregional PDOs established. \u003cstrong\u003ee.\u003c/strong\u003eRadiological tumor response to neoadjuvant systemic therapy plotted according to failed or successful PDO establishment of the corresponding lesion sampled at resection. \u003cstrong\u003ef.\u003c/strong\u003e Overall survival of patients (n=132) measured from time of liver surgery and grouped according to failed or successful PDO establishment. Hazard ratio (HR) and p-value is from Cox proportional hazards analysis.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/13a29607b335eb37d11ef4ef.png"},{"id":82349890,"identity":"c707ee50-6298-4e19-a1e4-6fae05ca1b47","added_by":"auto","created_at":"2025-05-09 10:49:26","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1969872,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCircos plot of PDOs and data types analyzed. \u003c/strong\u003eEach PDO (n=213) is plotted and grouped according to the patient (n=102) the original tumor was sampled from. Patients are separated by grey-shaded bars in the outermost and innermost tracks, and plotted clockwise in order of decreasing number of PDOs. The cDSS track shows box plots of DSS values for 24 drugs (centered by the mean and scaled by the standard deviation). Color codes reflect the median of the cDSS per PDO, grouped as drug resistant (median cDSS \u0026lt; 0.15), intermediate (median cDSS -0.15 – 0.15) or sensitive (median cDSS \u0026gt; 0.15). The morphology track indicates the morphological phenotype of each PDO. The growth speed track indicates three groups of PDOs based on viability (see Methods). The protein expression track shows the relative expression of CDX2, KRT20 and KI67 scaled according to the median across all PDOs (dashed lines indicate the median expression levels). The iCMS track shows the transcriptomic iCMS classification. The three innermost tracks indicate the mutation status of \u003cem\u003eRAS\u003c/em\u003e (\u003cem\u003eKRAS\u003c/em\u003eor \u003cem\u003eNRAS\u003c/em\u003e), \u003cem\u003eBRAF\u003c/em\u003e and \u003cem\u003eTP53\u003c/em\u003e. Empty cells indicate that the data type is unavailable (not generated).\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/2fcfa198c726ab3cab8692bd.png"},{"id":82348408,"identity":"db2a672f-954c-4d0a-a0a0-39ee99fc90d9","added_by":"auto","created_at":"2025-05-09 10:41:26","extension":"pdf","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":3157534,"visible":true,"origin":"","legend":" Proportion of sampled lesions with successfully established PDOs per liver segment. The size of the pie charts is proportional to the number of sampled CRLMs per segment. Reasons for growth failure of CRLMs (n\u0026thinsp;=\u0026thinsp;133). Number of successfully established and failed PDOs summarized according to the number of CRLMs attempted to culture per patient. Number of successfully established and failed PDOs per patient, ranked in consecutive order of patient inclusion. Bars separated by black lines indicate CRLMs from one or two hepatic re-resections in eight patients, with the first resection at the bottom. Asterisk marks a CRLM with two multiregional PDOs established. Radiological tumor response to neoadjuvant systemic therapy plotted according to failed or successful PDO establishment of the corresponding lesion sampled at resection. Overall survival of patients (n\u0026thinsp;=\u0026thinsp;132) measured from time of liver surgery and grouped according to failed or successful PDO establishment. Hazard ratio (HR) and p-value is from Cox proportional hazards analysis.","description":"","filename":"MainfiguresNATCANCER.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/d819c50a6a54abda63905452.pdf"},{"id":82348404,"identity":"effe0ba3-71b4-4555-be11-3f3eed850fe4","added_by":"auto","created_at":"2025-05-09 10:41:26","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":2032672,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePDOs are faithful models of the corresponding CRLMs. a. \u003c/strong\u003eRepresentative micrographs of four PDOs from two patients shown in phase contrast and HE stains of parallel samples (fixed and embedded) and corresponding CRLMs. Scale bar 100\u0026nbsp;µm. \u003cstrong\u003eb.\u003c/strong\u003e Volcano plot from gene expression enrichment analyses of a custom gene set collection (n=106) between corresponding CRLMs and PDOs (n=50 sample pairs analyzed in paired mode). The gene sets with lowest p-value (adjusted for false discovery rate, FDR) for each sample type are highlighted. The dashed line indicates the significance threshold (FDR-adjusted p=0.05). \u003cstrong\u003ec.\u003c/strong\u003e Heat map (left) of Spearman correlation coefficients for each of the gene expression-based principal components 1 to 3 between PDOs and corresponding CRLMs (n=50 sample pairs), and scatter plot (right) for the second principal component (PC2) in the paired samples. Sample pairs are colored according to single-sample enrichment scores of the \u003cem\u003eBRAF\u003c/em\u003e-mutant-like gene set in the PDOs. \u003cstrong\u003ed.\u003c/strong\u003e Frequency of the 10 most frequently mutated genes among patients in this study (n=102; analyzed in a single PDO per patient), compared to metastatic CRCs from previous studies published by Mendelaar et al\u003csup\u003e34\u003c/sup\u003e and Yaeger et al\u003csup\u003e33\u003c/sup\u003e. Pie charts indicate the frequency of hotspot mutations for \u003cem\u003eKRAS, NRAS\u003c/em\u003e and \u003cem\u003eBRAF\u003c/em\u003e among patients in this study.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/208abb3d7eb58b0364c164ec.png"},{"id":82348405,"identity":"5495fdab-49cd-4c6d-9552-e3f0de6c68c5","added_by":"auto","created_at":"2025-05-09 10:41:26","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":775539,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMolecular associations of drug sensitivities. a. \u003c/strong\u003eScatter plot of associations between mutations (n=18 genes) and drug sensitivities (n=51 drugs) according to significance level from Wilcoxon tests with Bonferroni-Hochberg correction versus the number of mutated PDOs per gene (total n=151 PDOs).\u003cstrong\u003e \u003c/strong\u003eDot sizes reflect the difference in DSS (ΔDSS) between the mutated and wild-type groups, and significant associations are indicated with colors according to sensitivity in the mutated group. \u003cstrong\u003eb.\u003c/strong\u003e Box plot of the mean DSS of 24 drugs according to co-mutation status of \u003cem\u003eRAS\u003c/em\u003e/\u003cem\u003eBRAF\u003c/em\u003e\u003csup\u003eV600E\u003c/sup\u003e and \u003cem\u003eTP53\u003c/em\u003e in PDOs. \u003cstrong\u003ec.\u003c/strong\u003e Box plot of DSS of the drug combination FLIRI (5-FU, SN-38 and leucovorin) according to co-mutation status of \u003cem\u003eRAS \u003c/em\u003eand\u003cem\u003e PIK3CA\u003c/em\u003e.\u003cstrong\u003e d. \u003c/strong\u003eBox plots of DSS for gemcitabine and the aurora kinase inhibitor alisertib according to the transcriptomic iCMS classes. ΔDSS is the difference of the median DSS between the groups, and p-values are from Wilcoxon tests. \u003cstrong\u003ee.\u003c/strong\u003e Spearman correlation coefficient (indicated by the color scale) and p-value (indicated by the dot size) between the expression level of proteins (n=12; plotted along the y-axis) and DSS values of drugs (n=28; plotted along the x-axis). Drugs are grouped according to class as indicated by the colored annotation bar.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/c10ef253362d50a7353cfb5f.png"},{"id":82348414,"identity":"84f2b8bb-0847-4edb-9bdd-70ea4c463269","added_by":"auto","created_at":"2025-05-09 10:41:26","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":2853978,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMorphological phenotypes of PDOs.\u003c/strong\u003e \u003cstrong\u003ea.\u003c/strong\u003e Examples of PDOs of each morphological phenotype, illustrated by HE stains (top) of two PDOs from distinct metastatic lesions (T1 and T2) from each of three patients, and multiplex fluorescent immunohistochemistry stains (mIHC; bottom) of CDX2, KRT20 and nuclei in the same PDOs. Scale bar 100\u0026nbsp;µm. \u003cstrong\u003eb.\u003c/strong\u003e Violin plots of CDX2 and KRT20 protein expression according to the three phenotypes of PDOs. P-values are from Kruskal-Wallis test across the three groups and Wilcoxon tests between two groups as indicated. \u003cstrong\u003ec.\u003c/strong\u003e Stacked bar plots of the proportion of PDOs with \u003cem\u003eRAS\u003c/em\u003e and \u003cem\u003eBRAF\u003c/em\u003e\u003csup\u003e\u003cem\u003eV600E\u003c/em\u003e\u003c/sup\u003e mutations and \u003cstrong\u003e(d)\u003c/strong\u003e the transcriptomic iCMS groups according to morphological phenotypes.\u003cstrong\u003e \u003c/strong\u003eP-values are from Pearson’s chi-square tests. \u003cstrong\u003ee. \u003c/strong\u003eVolcano plots from gene expression enrichment analyses of a custom gene set collection (n=106) between PDOs with the cystic and solid phenotypes, analyzed separately for the Human Transcriptome 2.0 Arrays (HTA; n=69 PDOs) and RNA sequencing data sets (n=70 PDOs). \u003cstrong\u003ef.\u003c/strong\u003e Box plots of the mean DSS across 24 drugs (left), DSS for 5-FU (middle), and DSS for afatinib (left) in PDOs according to the morphological phenotypes. P-values are from Kruskal-Wallis tests across three groups and Wilcoxon tests between two groups.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/762a1235cd6a8f3f27f280d6.png"},{"id":82348403,"identity":"e4d7dc59-cc47-4adc-a102-1cd0755592b2","added_by":"auto","created_at":"2025-05-09 10:41:26","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":627420,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eHeterogeneity of multi-lesion PDOs. a. \u003c/strong\u003eBar plot of the number of PDOs per patient with multiple PDOs (n=175 PDOs from 66 patients), colored and sorted according to morphological phenotype and tumor heterogeneity. Bars separated by black lines indicate PDOs from different hepatic resections in six patients (the order of resections is indicated with numbers). Asterisk indicates two multiregional PDOs from a single metastasis, remaining PDOs were from different metastatic lesions. \u003cstrong\u003eb.\u003c/strong\u003e Distribution of\u003cstrong\u003e \u003c/strong\u003epharmacological heterogeneity scores estimated as pair-wise Euclidean distances of drug sensitivities (n=24 drugs) in intra-patient comparisons of PDOs from the same liver resection (spatial) and recurrent lesions from re-resections (temporal), or inter-patient comparisons. The groups were compared with one-way ANOVA and p-values adjusted with Tukey’s multiple comparison test (****p\u0026lt;0.0001). \u003cstrong\u003ec\u003c/strong\u003e. Mean DSS values across 24 drugs in patients (n=9) with PDOs of both the cystic and solid morphological phenotypes. The darker colored dots indicate PDOs sampled at two resections of one patient. \u003cstrong\u003ed\u003c/strong\u003e. Boxplots of intra-patient inter-metastatic pharmacological heterogeneity scores according to homogeneity or heterogeneity of the morphological phenotypes of the PDOs. \u003cstrong\u003ee.\u003c/strong\u003e Box plots of intra-patient inter-metastatic pharmacological heterogeneity scores for all pairs of PDOs per patient (PDOs from the same resection in n=64 patients). Patients with two PDOs have a single pair-wise comparison and are plotted with a single dot. Patients are ranked according to increasing median pharmacological heterogeneity among sample pairs. The annotation panels below indicate the morphological phenotypes, transcriptomic iCMS classification and mutations per patient, and multicolor boxes with black outlines indicate intra-patient heterogeneity. Mutations were analyzed in a single PDO from most patients, and only the 6 genes with detected mutation heterogeneity in at least one patient are considered. NC, not classified.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/c191cd1e4fb55461cf616798.png"},{"id":82349894,"identity":"7ab66757-4f09-4b48-b6cd-f0013afb28d4","added_by":"auto","created_at":"2025-05-09 10:49:26","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":1686046,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTwo patients with prospective treatment selection according to pharmacogenomic profiles. a. \u003c/strong\u003eOrder of oncological treatment, radiological response evaluations and \u003cem\u003eex vivo\u003c/em\u003e molecular and functional analyses of patient Pt137 (from left to right in the top panel). Selected gene mutations are specified (bottom left). Immunohistochemistry stains show in situ ERBB2 expression in the primary tumor, two CRLMs (T4 and T5), and the corresponding PDOs. Scale bar 50 µm. Dose response curves show PDO viability according to treatment doses of lapatinib (ERBB2/EGFR inhibitor), and PDOs from patient Pt137 are highlighted by colors. Radiological images (bottom right) of the abdomen before and after treatment with anti-ERBB2 antibodies. \u003cstrong\u003eb.\u003c/strong\u003e Oncological treatment, radiological response evaluations and \u003cem\u003eex vivo\u003c/em\u003e molecular and functional analysis of patient Pt160. \u003cem\u003eEx vivo\u003c/em\u003edose response curves of FOLFIRI (5-FU and SN-38 at 1:1 ratio) indicate a strong and homogeneous sensitivity in PDOs derived from two distinct CRLMs. CT - computed tomography, MR - magnetic resonance imaging, PD – progressive disease, PR – partial response, SD – stable disease.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/427c45682cda70b7c0b8dc16.png"},{"id":82353107,"identity":"5d69f848-962d-4066-83c5-304fa925c8a0","added_by":"auto","created_at":"2025-05-09 11:05:32","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":11619760,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/be989794-da11-4cca-9f01-396fa2c4c794.pdf"},{"id":82348400,"identity":"b7521501-4ef9-4908-9a19-834d7d65a196","added_by":"auto","created_at":"2025-05-09 10:41:26","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":267792,"visible":true,"origin":"","legend":"Supplementary Table 1 to Table 13","description":"","filename":"SupplementaryTablesNATCANCER.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/c83e55bf7ad6e3e9acc7e7a1.pdf"},{"id":82348398,"identity":"6528f078-eeeb-43a7-a57e-7b28011062cc","added_by":"auto","created_at":"2025-05-09 10:41:26","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":75938,"visible":true,"origin":"","legend":"\u003cp\u003eData Set 1\u003c/p\u003e","description":"","filename":"DataSet1Mutations150PDOs74CRLMs.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/7c95686a8fac63706106f68a.xlsx"},{"id":82351496,"identity":"3ed2cb11-e77f-4190-8787-513872331dfd","added_by":"auto","created_at":"2025-05-09 10:57:26","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":10189391,"visible":true,"origin":"","legend":"Data Set 2","description":"","filename":"DataSet2PDO92HTSeqcountproteincoding.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/265bb7b73e45a6fde1d41455.xlsx"},{"id":82348402,"identity":"444bc76a-1af1-4c51-b879-dba6837625a3","added_by":"auto","created_at":"2025-05-09 10:41:26","extension":"xlsx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":119316,"visible":true,"origin":"","legend":"Data Set 3","description":"","filename":"DataSet3ProteinsmIHC136PDOs.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/681ba6a97e3322c915181c26.xlsx"},{"id":82349895,"identity":"624d4865-9bac-41c4-99e7-3a8d54142912","added_by":"auto","created_at":"2025-05-09 10:49:26","extension":"xlsx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":13772499,"visible":true,"origin":"","legend":"Data Set 4","description":"","filename":"DataSet4DSRTRAW211PDOs.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/0a19ee8091ef3e70130cb9c4.xlsx"},{"id":82348407,"identity":"5aaffbd6-6310-4cc2-af02-85aa1ee73b84","added_by":"auto","created_at":"2025-05-09 10:41:26","extension":"pdf","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":1155343,"visible":true,"origin":"","legend":"Supplementary Figure 1 to Figure 8","description":"","filename":"SupplFigsNATCANCER.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/4ee85a7510ee17501ea3a2a5.pdf"},{"id":82349893,"identity":"85529a3b-39b9-435d-a462-9a3a7dfd7cee","added_by":"auto","created_at":"2025-05-09 10:49:26","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":19488,"visible":true,"origin":"","legend":"","description":"","filename":"ExtendedDataFigureLegends.docx","url":"https://assets-eu.researchsquare.com/files/rs-6507406/v1/5d74f5431f1147a91f80dd38.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Ex vivo modeling of morphological, molecular and pharmacological tumor heterogeneity of metastatic colorectal cancer","fulltext":[{"header":"Introduction","content":"\u003cp\u003eFunctional oncology based on \u003cem\u003eex vivo\u003c/em\u003e drug sensitivity testing of patient-derived cancer cells can complement genomics-guided approaches to precision cancer medicine\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Prospective treatment selection guided by functional assays has shown clinical benefit in patients with hematologic cancers but has proven more difficult in patients with solid tumors\u003csup\u003e\u003cspan additionalcitationids=\"CR3\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. Tumor cells can be cultured as self-organizing three-dimensional organoid structures that resemble the original tissue, and the use of organoid models in cancer research has increased exponentially over the past decade. Colorectal cancer (CRC) was the first cancer type to be modeled with patient-derived organoids (PDOs)\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e, and both the tissue architecture and molecular profiles of CRCs can be recapitulated with this model system\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. Drug sensitivity testing of CRC PDOs can also reflect the clinical responses of the original tumors to standard chemotherapy and targeted agents\u003csup\u003e\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. However, successful examples of functional precision oncology and prospective treatment selection in patients using this approach remain few or anecdotal\u003csup\u003e\u003cspan additionalcitationids=\"CR11\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTumor heterogeneity is a major cause of treatment resistance and poor patient outcome\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e,\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. Metastatic heterogeneity of CRC has been extensively studied at the molecular level, highlighting early seeding, polyclonal seeding, seeding among metastatic deposits, and phenotypic plasticity in development of metastatic disease, potentially augmented by treatment exposure\u003csup\u003e\u003cspan additionalcitationids=\"CR16 CR17 CR18\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. Furthermore, heterogeneity of response to standard treatment is common among metastatic lesions and can have a negative impact on patient outcomes\u003csup\u003e\u003cspan additionalcitationids=\"CR21\" citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. However, the largest living biobanks of CRC do not model tumor heterogeneity and typically include a single PDO per patient\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e,\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. PDOs derived from multiple single cells or regions of primary tumors have suggested marked intra-tumor heterogeneity of drug sensitivities\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e,\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e, while paired PDOs of the primary tumor and liver metastasis have shown similar \u003cem\u003eex vivo\u003c/em\u003e sensitivity to standard chemotherapies\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e. Here, we expand on the largest published study of \u003cem\u003eex vivo\u003c/em\u003e pharmacogenomic heterogeneity in metastatic CRC\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. The corresponding pharmaco-omics data serve as reference for prospective drug nomination in an ongoing intervention trial of metastatic CRC (ClinicalTrials.gov identifier NCT05725200), and we report from the observational and translational pilot phase of this functional oncology study.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eA living biobank of metastatic CRC\u003c/h2\u003e \u003cp\u003eFresh tissue samples of 346 colorectal liver metastases (CRLMs) from 132 patients treated by hepatectomy for metastatic CRC were included for \u003cem\u003eex vivo\u003c/em\u003e culturing (Supplementary Table\u0026nbsp;1 and Supplementary Fig.\u0026nbsp;1). PDOs were successfully established for 213 lesions (62%) from 102 patients (77%), including multiple (up to 6) spatially separated lesions from each of 65 patients (64%), as well as recurrent CRLMs sampled at re-resections of six patients (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea-d). The median time from sampling to a completed drug screen and cryopreservation of PDOs for biobanking was 8.4 weeks (10th\u0026ndash;90th percentile range: 4\u0026ndash;21 weeks). There was no difference in the culture success rate according to tumor location (liver segment; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea), the number of lesions attempted to culture per patient (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ec), the type of surgery, or any of the clinicopathological or molecular characteristics listed in Supplementary Table\u0026nbsp;1. Samples with low tumor cell content (n\u0026thinsp;=\u0026thinsp;3) or necrosis (n\u0026thinsp;=\u0026thinsp;5) failed to grow, but the majority (90%) of growth failures could not be attributed to histopathological or molecular markers (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb, Supplementary Table\u0026nbsp;1, Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). There was also no growth selection based on neoadjuvant chemotherapy exposure (p\u0026thinsp;\u0026gt;\u0026thinsp;0.9 by Pearson's Chi-squared test of treated versus chemo-na\u0026iuml;ve lesions), but lesions that did not propagate \u003cem\u003eex vivo\u003c/em\u003e had a stronger response to neoadjuvant treatment (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ee). \u003cem\u003eEx vivo\u003c/em\u003e drug sensitivities in the successfully established PDOs did not reflect clinical tumor responses to the corresponding neoadjuvant treatment, and this might also reflect selection of propagated cells based on treatment response (Supplementary Fig.\u0026nbsp;2). Patients with at least one successfully established PDO had a significantly shorter median overall survival time after liver resection (38 months) than patients with no PDOs established (median not reached at four years follow-up; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ef), supporting the hypothesis that successful \u003cem\u003eex vivo\u003c/em\u003e growth reflects aggressive disease and/or poorer response to chemotherapy.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAn overview of the living biobank and pharmaco-omics data is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, including drug sensitivity profiles of 41 or 47 drugs (illustrated by 24 overlapping drugs between the two custom libraries), gene expression profiles (illustrated by classification according to the intrinsic consensus molecular subtypes, iCMS\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e), mutations of 20 CRC-relevant genes (illustrated by \u003cem\u003eKRAS\u003c/em\u003e/\u003cem\u003eNRAS\u003c/em\u003e [\u003cem\u003eRAS\u003c/em\u003e], \u003cem\u003eBRAF\u003c/em\u003e and \u003cem\u003eTP53\u003c/em\u003e), multiplex fluorescent immunohistochemistry of 12 proteins (illustrated by CDX2, KRT20 and KI67), and morphological phenotypes based on hematoxylin and eosin (HE) stains. All PDOs were microsatellite stable (MSS).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003ePDOs are faithful models of the biological and molecular diversity of CRLMs\u003c/h3\u003e\n\u003cp\u003ePaired HE stains of PDOs and corresponding tumor tissue samples (n\u0026thinsp;=\u0026thinsp;46 lesions from 26 patients) indicated that 85% of PDOs retained the tumor morphologies for histopathological features such as acinar, cystic and solid structures (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea, Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Supplementary Table\u0026nbsp;2). PDOs with dissimilar morphologies were not characterized by distinct growth rates or any specific molecular marker (Supplementary Table\u0026nbsp;3).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eGene expression profiling indicated high similarity scores of PDOs (n\u0026thinsp;=\u0026thinsp;211) for primary CRC tissue according to the CancerCellNet approach\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea). Notably, PDOs had higher similarity scores than their paired tissue samples (n\u0026thinsp;=\u0026thinsp;50 sample pairs evaluated; p\u0026thinsp;=\u0026thinsp;7\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e by paired t-test; Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb), likely reflecting influence from the liver-specific tumor microenvironment in the comparison of CRLMs with primary tumors. Gene set enrichment analysis of the PDO-CRLM sample pairs confirmed enrichment exclusively with tumor microenvironment signals in tissue samples relative to PDOs (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb). PDOs showed enrichment with signatures of metabolism and the cell cycle, likely reflecting higher proliferative activity and ample access to nutrients and oxygen in the cell cultures. The first component from principal component analysis across both PDOs and CRLMs was primarily driven by the difference in tumor microenvironment signals (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec-e) and showed no correlation between the matched sample pairs (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec). In contrast, the second component was driven by cancer cell-intrinsic features (illustrated with a \u003cem\u003eBRAF\u003c/em\u003e mutant-like expression signature\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e) and was strongly correlated between the PDOs and paired tumor tissue, indicating that PDOs recapitulated the transcriptomic profile of their original tumor.\u003c/p\u003e \u003cp\u003eMutation frequencies of the ten most frequently mutated genes (analyzed across a single PDO from each of the 102 patients) were similar to two previous studies of metastatic CRC\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e (p\u0026thinsp;=\u0026thinsp;0.98 by chi-squared test; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed). Matched PDO-CRLM sample pairs (n\u0026thinsp;=\u0026thinsp;74 lesions from 56 patients) also showed highly correspondent mutation profiles, with concordant status for 93.7% of the 491 detected mutations (Supplementary Fig.\u0026nbsp;3a). Discordances were primarily due to subclonal mutations with low mutant allele fractions (p\u0026thinsp;=\u0026thinsp;0.002 by Wilcoxon\u0026rsquo;s test of the allelic fraction of discordant and concordant mutations), frequently involving a second hit in \u003cem\u003eAPC\u003c/em\u003e or \u003cem\u003eTP53\u003c/em\u003e mutations (Supplementary Fig.\u0026nbsp;3a and 3c).\u003c/p\u003e\n\u003ch3\u003ePharmacological associations of molecular markers\u003c/h3\u003e\n\u003cp\u003eKnown pharmacogenomic associations were confirmed in the living biobank, including resistance to EGFR inhibitors in \u003cem\u003eRAS and BRAF\u003c/em\u003e\u003csup\u003e\u003cem\u003eV600E\u003c/em\u003e\u003c/sup\u003e mutated PDOs (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea and Extended data Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea-b), sensitivity to the MDM2-TP53 inhibitor idasanutlin in PDOs with wild-type \u003cem\u003eTP53\u003c/em\u003e and/or high \u003cem\u003eTP53\u003c/em\u003e mRNA expression (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec-e), as well as outlier sensitivity to sotorasib in the two PDOs with \u003cem\u003eKRAS\u003c/em\u003e\u003csup\u003e\u003cem\u003eG\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003eC\u003c/em\u003e\u003c/sup\u003e mutation (from a single patient; Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ef). Notably, the majority of \u003cem\u003eBRAF\u003c/em\u003e\u003csup\u003e\u003cem\u003eV600E\u003c/em\u003e\u003c/sup\u003e mutated PDOs were resistant to single-agent treatment with the mutation-specific inhibitor encorafenib (PDOs from 4 of 5 patients; Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eg). A combination of encorafenib with afatinib (EGFR inhibitor) and trametinib (MEK inhibitor) showed increased activity (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eh). However, this effect was not specific to \u003cem\u003eBRAF\u003c/em\u003e\u003csup\u003e\u003cem\u003eV600E\u003c/em\u003e\u003c/sup\u003e mutated PDOs (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ei), suggesting that distinction between the synergistic and additive effects of the drug combination requires further analyses of complete dose-response matrices of the drug pairs.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAn extended analysis of drug sensitivities (n\u0026thinsp;=\u0026thinsp;51 unique drugs across both libraries; Supplementary Table\u0026nbsp;4) and mutations (n\u0026thinsp;=\u0026thinsp;8 genes with at least 10 mutated and wild-type PDOs) showed 12 significant associations, all involving \u003cem\u003eAPC, KRAS, NRAS, BRAF\u003c/em\u003e, or \u003cem\u003eTP53\u003c/em\u003e (Wilcoxon tests with Benjamini-Hochberg correction; Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). Beyond the well-known pharmacogenomic associations, \u003cem\u003eKRAS\u003c/em\u003e mutations were associated with sensitivity to the multi-kinase inhibitor regorafenib (p\u0026thinsp;=\u0026thinsp;0.045). There was no difference according to mutation hotspots (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea), and the association was stronger when including \u003cem\u003eBRAF\u003c/em\u003e\u003csup\u003e\u003cem\u003eV600E\u003c/em\u003e\u003c/sup\u003e (p\u0026thinsp;=\u0026thinsp;7.7\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e). \u003cem\u003eBRAF\u003c/em\u003e\u003csup\u003e\u003cem\u003eV600E\u003c/em\u003e\u003c/sup\u003e mutated PDOs showed low sensitivity to the IGF1R inhibitor BMS-754807 (p\u0026thinsp;=\u0026thinsp;0.01). This association was specific to \u003cem\u003eBRAF\u003c/em\u003e\u003csup\u003e\u003cem\u003eV600E\u003c/em\u003e\u003c/sup\u003e and not found for any of the \u003cem\u003eRAS\u003c/em\u003e hotspots (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eb). Furthermore, \u003cem\u003eTP53\u003c/em\u003e mutated and \u003cem\u003eAPC\u003c/em\u003e mutated PDOs were less sensitive to the standard chemotherapies 5-FU (p\u0026thinsp;=\u0026thinsp;0.03) and SN-38 (active metabolite of irinotecan; p\u0026thinsp;=\u0026thinsp;0.01) than the corresponding wild-type PDOs, respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). A systematic evaluation of co-occurring mutations in pairs of genes (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea) showed that \u003cem\u003eTP53\u003c/em\u003e and \u003cem\u003eAPC\u003c/em\u003e co-mutated PDOs had lower drug sensitivities in general (Supplementary Table\u0026nbsp;5) and a lower mean drug sensitivity score (DSS) across the drug libraries (p\u0026thinsp;=\u0026thinsp;0.02). This was also the case for the poor-prognostic subgroup defined by co-occurring mutations of \u003cem\u003eRAS\u003c/em\u003e/\u003cem\u003eBRAF\u003c/em\u003e\u003csup\u003eV600E\u003c/sup\u003e and \u003cem\u003eTP53\u003c/em\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb). The strongest positive association to drug sensitivity was observed with co-occurring \u003cem\u003eRAS\u003c/em\u003e and \u003cem\u003ePIK3CA\u003c/em\u003e mutations, and these PDOs had strong sensitivity to the combination chemotherapy FLIRI (5-FU, leucovorin and SN-38; Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec), although not to any of the corresponding single drugs (Supplementary Table\u0026nbsp;5).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eComparisons of DSS values according to the transcriptomic iCMS classification showed higher activity of EGFR inhibitors in iCMS2 PDOs, attributed to enrichment with \u003cem\u003eRAS/BRAF\u003c/em\u003e\u003csup\u003eV600E\u003c/sup\u003e mutations in the iCMS3 group (odds ratio 32, p\u0026thinsp;=\u0026thinsp;3.5\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;13\u003c/sup\u003e by Fisher\u0026rsquo;s exact test). However, higher activity of gemcitabine (chemotherapy) and alisertib (Aurora kinase A inhibitor) in iCMS3 PDOs was independent of \u003cem\u003eRAS/BRAF\u003c/em\u003e\u003csup\u003e\u003cem\u003eV600E\u003c/em\u003e\u003c/sup\u003e mutation status (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ed; Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ec).\u003c/p\u003e \u003cp\u003eMultiplex fluorescent immunohistochemistry of selected diagnostic markers, proteins involved in drug resistance, and other proteins of interest (n\u0026thinsp;=\u0026thinsp;12) showed that the activity of 78% of drugs (n\u0026thinsp;=\u0026thinsp;28 of 33 single drugs) was significantly correlated to the expression of at least one protein among PDOs (n\u0026thinsp;=\u0026thinsp;77\u0026ndash;134 PDOs from 37\u0026ndash;67 patients; Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ee). The strongest drug-protein interaction was a negative correlation between expression of the ABC transporter ABCB1 and activity of the PI3K/AKT pathway inhibitor gedatolisib (Spearman\u0026acute;s rho=-0.5, p\u0026thinsp;=\u0026thinsp;9.3x10\u003csup\u003e\u0026minus;\u0026thinsp;10\u003c/sup\u003e). In general, proteins known to be involved in drug resistance and stress response, including ABCB1, ABCG2, UGT1A, and HSF1, were negatively correlated with the activity of several drugs. The strongest positive correlations were found between CDX2 expression and the standard chemotherapies 5-FU and TAS-102 (trifluridine/tipiracil). 5-FU sensitivity was also correlated with RCC2 expression, a protein suggested to be associated with benefit from adjuvant chemotherapy in locoregional CRC\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e. High expression of either CDX2 or RCC2 (\u0026gt;\u0026thinsp;70th percentiles) accounted for the majority of 5-FU-sensitive PDOs (DSS\u0026thinsp;\u0026gt;\u0026thinsp;14.3; odds ratio 2.7, p\u0026thinsp;=\u0026thinsp;0.01; Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ed). Furthermore, TP53 expression was negatively correlated with sensitivity to the PARP inhibitor olaparib. High TP53 expression was found exclusively in PDOs with missense \u003cem\u003eTP53\u003c/em\u003e mutations (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ee), consistent with our previous study suggesting that wild-type TP53 activity is needed for PARP inhibitor sensitivity in CRC cells\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. Additionally, RIPK1 expression was positively correlated to the activity of the SMAC mimetic LCL161, which targets inhibitor of apoptosis proteins\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003ch3\u003eClinicopathological and pharmacological associations\u003c/h3\u003e\n\u003cp\u003ePatients with right-sided primary tumors have limited benefit from anti-EGFR treatment\u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e, and the living biobank confirmed sensitivity to EGFR inhibitors only in \u003cem\u003eRAS\u003c/em\u003e/\u003cem\u003eBRAF\u003c/em\u003e wild-type PDOs derived from patients with left-sided or rectal primary tumors (p\u0026thinsp;=\u0026thinsp;0.02 by Wilcoxon test; Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eb). In contrast, right-sided primary tumor location was associated with higher sensitivity to the standard chemotherapies SN-38 and TAS-102, independent of \u003cem\u003eRAS\u003c/em\u003e/\u003cem\u003eBRAF\u003c/em\u003e mutation status (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eb and Supplementary Table\u0026nbsp;6). Patient sex was not associated with sensitivity to any drug, including 5-FU (p\u0026thinsp;=\u0026thinsp;0.7 by Wilcoxon test), while high patient age was weakly correlated to 5-FU sensitivity (Spearman rho\u0026thinsp;=\u0026thinsp;0.28, p\u0026thinsp;=\u0026thinsp;0.02; Supplementary Table\u0026nbsp;7). Diagnosis with multiple CRLMs on radiological imaging before liver surgery was associated with low drug sensitivity in general, although a statistically significant correlation was found for SN-38 and idasanutlin only (Supplementary Table\u0026nbsp;8). Cox proportional hazards analysis indicated that high \u003cem\u003eex vivo\u003c/em\u003e drug sensitivities were associated with better overall survival among the patients (mean DSS across 24 drugs as predictor; hazard ratio 0.91, 95% confidence interval 0.82-1.0, p\u0026thinsp;=\u0026thinsp;0.07), and the prognostic association was significant for four drugs (atorvastatin, bemcentinib, and two EGFR inhibitors; Supplementary Table\u0026nbsp;9). Sensitivity to the cholesterol-lowering agent atorvastatin and the AXL inhibitor bemcentinib maintained significant prognostic associations in multivariable models with clinicopathological variables and \u003cem\u003eRAS\u003c/em\u003e/\u003cem\u003eBRAF\u003c/em\u003e mutation status (multivariable hazard ratio 0.81 and 0.79, 95% confidence interval 0.71\u0026ndash;0.93 and 0.64\u0026ndash;0.96, respectively; Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003ePDOs of CRLMs represent two main morphological phenotypes\u003c/h3\u003e\n\u003cp\u003ePDOs were classified into two main morphological phenotypes based on HE stains of paraffin embedded samples (n\u0026thinsp;=\u0026thinsp;213; Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea, Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). A cystic phenotype was identified in 28% of PDOs and was characterized by organized cuboidal and columnar cells forming cystic/glandular structures with a well-defined central lumen. A solid phenotype was found in 38% of PDOs and these lacked a central lumen and had predominantly solid growth patterns, including cytoplasmic vacuoles or individually disaggregated cells. The remaining PDOs (34%) had traits of both the cystic and solid phenotypes and were classified as mixed (less than 50% of the individual structures showed clear characteristics of either phenotype).\u003c/p\u003e \u003cp\u003eMultiplex fluorescent immunohistochemistry showed differential expression of the CRC diagnostic markers KRT20 and CDX2 according to the morphological phenotypes (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eb and Extended Data Fig.\u0026nbsp;8a-b). CDX2 had highest expression in cystic PDOs, KRT20 in solid PDOs, and the mixed phenotype showed intermediate expression levels of both proteins. The solid phenotype was also enriched with \u003cem\u003eRAS\u003c/em\u003e and \u003cem\u003eBRAF\u003c/em\u003e\u003csup\u003e\u003cem\u003eV600E\u003c/em\u003e\u003c/sup\u003e mutations (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ec), and 92% of \u003cem\u003eBRAF\u003c/em\u003e\u003csup\u003e\u003cem\u003eV600E\u003c/em\u003e\u003c/sup\u003e PDOs had a solid phenotype (Extended Data Fig.\u0026nbsp;8c). Consistently, enrichment with a \u003cem\u003eBRAF\u003c/em\u003e mutant-like expression signature in solid PDOs was also the strongest distinction between the subtypes on the transcriptomic level (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ee, Extended Data Fig.\u0026nbsp;8e). Solid PDOs were also more MSI-like and enriched with signatures of apoptosis and inflammation, including TNF-α signaling and IFN-α response, as well as the iCMS3 class (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ed). In contrast, cystic PDOs were MSS-like and enriched with cancer stem cell signatures and iCMS2. Cystic PDOs were also generally more sensitive to drugs, estimated as a higher mean sensitivity across the 24 drugs evaluated in both libraries (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ef and Supplementary Table\u0026nbsp;10). This included higher sensitivity to standard treatments for CRC such as 5-FU and EGFR inhibitors (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ef), reflecting higher expression of CDX2 and less frequent mutations of \u003cem\u003eRAS\u003c/em\u003e/\u003cem\u003eBRAF\u003c/em\u003e\u003csup\u003e\u003cem\u003eV600E\u003c/em\u003e\u003c/sup\u003e (Extended Data Fig.\u0026nbsp;8d).\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eModeling metastatic heterogeneity with multi-lesion PDOs\u003c/h2\u003e \u003cp\u003eMorphological and pharmacological tumor heterogeneity was analyzed among PDOs of distinct metastatic lesions from each of 66 patients (n\u0026thinsp;=\u0026thinsp;175 PDOs; Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea). Most patients had PDOs of the same morphological phenotype (n\u0026thinsp;=\u0026thinsp;40, 61%). However, the mixed group represents an inherently heterogeneous phenotype, and beyond the 18% of patients (n\u0026thinsp;=\u0026thinsp;12) with exclusively mixed PDOs, another 23% (n\u0026thinsp;=\u0026thinsp;15) had inter-metastatic heterogeneity with a combination of mixed and either cystic or solid PDOs. Furthermore, heterogeneity of the cystic and solid phenotypes was observed in 12% of patients (n\u0026thinsp;=\u0026thinsp;8), and a morphological phenotype switch after recurrence was observed in one of six evaluable patients (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea).\u003c/p\u003e \u003cp\u003eHeterogeneity of drug sensitivities was less prominent in intra-patient than inter-patient comparisons, evaluated as lower pharmacological heterogeneity scores between patient-matched sample pairs and estimated based on Euclidean distances of DSS values (n\u0026thinsp;=\u0026thinsp;24 overlapping drugs; Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eb; p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001 by one-way ANOVA). This was supported by hierarchical clustering analysis, showing clustering of PDOs from the same patient (Supplementary Fig.\u0026nbsp;4). There was no difference in the level of spatial heterogeneity among lesions from the same resection and longitudinal heterogeneity among recurrent lesions sampled at re-resections (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eb). The intra-patient pharmacological heterogeneity score was not associated with any clinicopathological variable or patient survival (Supplementary Table\u0026nbsp;11).\u003c/p\u003e \u003cp\u003eThe solid PDOs had lower mean DSS values than the cystic in all patients with morphological heterogeneity except one (p\u0026thinsp;=\u0026thinsp;0.027, Wilcoxon paired test; Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ec), indicating that morphological heterogeneity promoted pharmacological heterogeneity. This was supported by higher pharmacological heterogeneity scores among PDOs in patients with morphological heterogeneity (p\u0026thinsp;=\u0026thinsp;0.017 by Welch\u0026rsquo;s t-test, Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ed). However, the highest pharmacological heterogeneity score was found in a patient with three cystic PDOs and heterogeneous \u003cem\u003eAPC\u003c/em\u003e\u003csup\u003e\u003cem\u003eR213X\u003c/em\u003e\u003c/sup\u003e mutation status (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ee; Extended Data Fig.\u0026nbsp;9a). The mutated PDO was less sensitive to several chemotherapies (5-FU, methotrexate, gemcitabine) and more sensitive to the SMAC mimetic LCL161 (Extended Data Fig.\u0026nbsp;9b). Notably, all three PDOs were highly sensitive to the EGFR inhibitor afatinib (within the top 3rd -11th percentiles among all PDOs) and moderately sensitive to combination chemotherapy with FLIRI (25th -55th percentiles), and this corresponded with similar clinical responses to neoadjuvant FOLFIRI plus an anti-EGFR antibody in all three lesions (partial response with 35\u0026ndash;41% reduction of tumor sizes; Extended Data Fig.\u0026nbsp;9a).\u003c/p\u003e \u003cp\u003eThere were no clear differences in the level of intra-patient pharmacological heterogeneity among drugs or classes of drugs, estimated as the maximum DSS difference for each drug between any pair of PDOs per patient (Extended Data Fig.\u0026nbsp;9c; the low heterogeneity of encorafenib and idasanutlin was likely due to low drug activity in general). The highest heterogeneity estimates (\u0026gt;\u0026thinsp;95th percentile) were evenly distributed across drugs (median of 3 high scores per drug, 95% confidence interval 2\u0026ndash;5), supporting that pharmacological heterogeneity is more dependent on PDO biology than on the specific drug.\u003c/p\u003e \u003cp\u003e \u003cb\u003eProspective treatment selection based on\u003c/b\u003e \u003cb\u003eex vivo\u003c/b\u003e \u003cb\u003epharmacogenomics\u003c/b\u003e\u003c/p\u003e \u003cp\u003eTwo patients with promising \u003cem\u003eex vivo\u003c/em\u003e data for standard or well-tested treatments for metastatic CRC were treated according to their pharmacogenomics profile. Drug nominations were made relative to a reference dataset of the PDOs screened with the largest drug library (n\u0026thinsp;=\u0026thinsp;114 PDOs from 55 patients screened with lib2 of 47 drugs). The first patient (Pt137) was a man with known history of ulcerative colitis who was diagnosed with rectal cancer and synchronous CRLMs. He had stable disease after first-line treatment with FOLFIRI plus the antiangiogenic agent bevacizumab, and mixed response among metastatic lesions to subsequent neoadjuvant treatment with FOLFOXIRI plus bevacizumab (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ea). PDO culturing was attempted for five resected CRLMs and the primary tumor, but the two liver lesions with response to neoadjuvant treatment were necrotic and did not propagate \u003cem\u003eex vivo\u003c/em\u003e (Extended Data Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea). The dual EGFR-ERBB2 inhibitor lapatinib was the most active drug in the four established PDOs (highest mean DSS), although with considerable inter-tumor heterogeneity (ranked in the top 2nd -26th percentiles relative to the reference; Extended Data Fig.\u0026nbsp;10a). Molecular profiling confirmed \u003cem\u003eERBB2\u003c/em\u003e amplification and \u003cem\u003eRAS\u003c/em\u003e wild-type status in all tissue samples and PDOs analyzed, but \u003cem\u003ein situ\u003c/em\u003e expression of ERBB2 was heterogeneous and corresponded with the heterogeneous \u003cem\u003eex vivo\u003c/em\u003e sensitivity to lapatinib (lowest in the primary tumor and highest in the T5 CRLM). The patient had recurrence in the liver after resection, and recurrent CRLMs progressed on 3rd line treatment with FOLFIRI plus the anti-EGFR antibody cetuximab. Based on the pharmacogenomics data and previous clinical studies\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e, the patient was treated with a combination of the two anti-ERBB2 antibodies trastuzumab and pertuzumab in the 4th line. However, the CRLMs progressed on the first evaluation after four treatment cycles, likely attributed to the heterogeneity of ERBB2 expression and/or an \u003cem\u003eERBB2\u003c/em\u003e\u003csup\u003e\u003cem\u003eS310F\u003c/em\u003e\u003c/sup\u003e mutation later detected in all the PDOs and corresponding tumor tissue samples and potentially associated with pertuzumab resistance\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e,\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe second patient (Pt160) was a man diagnosed with sigmoidal colon cancer and synchronous CRLMs (\u003cem\u003eKRAS\u003c/em\u003e\u003csup\u003eG\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003eA\u003c/sup\u003e mutated). He underwent simultaneous surgery of the primary tumor and metastases after four cycles of neoadjuvant treatment with FOLFOX (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eb). The patient continued FOLFOX after surgery, until recurrence with non-resectable liver and lymph node metastases. The treatment was switched to FOLFIRI plus bevacizumab (totally 16 cycles over 273 days), which resulted in initial disease stabilization in the liver, except one small new metastasis (\u0026lt;\u0026thinsp;5 mm) detected after nine cycles (remained stable for three additional cycles). Following a brief treatment break, four additional cycles of FOLFIRI plus bevacizumab were given due to progression in the lungs and adrenal gland. Magnetic resonance imaging confirmed stable disease in the liver and progression in the lungs after completing the 16 treatment cycles. The patient subsequently progressed also on 3rd line chemotherapy and experimental treatment with immune checkpoint inhibitors. PDOs were successfully established from biopsies of two CRLMs taken after 3rd line chemotherapy but before immunotherapy. Both PDOs had strong \u003cem\u003eex vivo\u003c/em\u003e sensitivity to SN-38 (top 1st -3rd percentile relative to reference; Extended Data Fig.\u0026nbsp;10b) and combinations of SN-38 with 5-FU and leucovorin (preclinical FLIRI; top 1st -6th percentile) and the PLK1 inhibitor volasertib (top 1st -2nd percentile). FOLFIRI was re-administered in the 5th line based on the \u003cem\u003eex vivo\u003c/em\u003e drug sensitivity data, 300 days after completing FOLFIRI and bevacizumab in the 2nd line (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eb). The evaluation after 47 days and 3 cycles of FOLFIRI showed partial response of the two largest liver lesions and a reduction of most lung metastases (50% reduction of the total disease burden). Treatment with FOLFIRI was continued for an additional 60 days until progression of the lung metastases.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003ePatients with CRC have limited benefit from precision medicine guided by cancer genome sequencing\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e,\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. This study reports the preclinical development of a functional oncology platform that is the basis for an ongoing intervention trial of metastatic CRC (NCT05725200). Clinical translation was illustrated by successful rechallenge with combination chemotherapy guided by \u003cem\u003eex vivo\u003c/em\u003e drug sensitivity testing. Rechallenge with first-line therapies can provide a survival benefit over other late-line treatment options\u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e,\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e. However, studies of anti-EGFR therapies highlight the need for treatment guidance based on monitoring of resistance clones in liquid biopsies\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e, and there are currently no biomarkers to guide chemotherapy rechallenge in a similar manner. This study proposed drug sensitivity testing of PDOs as a potential approach. The turnaround time from sampling to a completed drug screen is a concern with diagnostic use of PDOs. In this respect, it is worth noting that rechallenge in the 5th line was successfully guided by biopsies taken before start of 4th line treatment. However, 4th line treatment had no effect in this patient, and dedicated longitudinal studies are needed to address whether response and/or exposure to intervening treatment has a potentially modifying effect on the predictive power of PDOs for subsequent treatment lines.\u003c/p\u003e \u003cp\u003eThe other example of \u003cem\u003eex vivo\u003c/em\u003e pharmacogenomics-guided treatment highlighted the impact of tumor heterogeneity. This patient showed heterogeneous \u003cem\u003eex vivo\u003c/em\u003e sensitivity to ERBB2 inhibition in a genetic background of \u003cem\u003eERBB2\u003c/em\u003e amplified tumors and derived no clinical benefit from anti-ERBB2 treatment. The \u003cem\u003eex vivo\u003c/em\u003e drug sensitivities reflected closely the heterogeneous expression levels of ERBB2, and the clinical data were consistent with a previous study showing that \u003cem\u003eERBB2\u003c/em\u003e amplification does not confer treatment response without protein overexpression\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e. Notably, the patient received targeted treatment for recurrent metastases without analyses of the recurrent lesions, and the relative impact of protein expression levels and co-occurring genetic aberrations (\u003cem\u003eERBB2\u003c/em\u003e amplification and S310 mutation) on the treatment outcome was not clear. Across the living biobank, patients showed large variation in the level of inter-metastatic heterogeneity of \u003cem\u003eex vivo\u003c/em\u003e drug sensitivities, and we found no strong clinicopathological or molecular correlates to potentially predict this heterogeneity. Furthermore, heterogeneity was observed for all drugs and with no apparent proclivity for heterogeneity of specific drug classes. However, no patient had heterogeneous sensitivity to all drugs, indicating potential to bypass vulnerability to heterogeneity by selecting treatments with similar activity across lesions. This was illustrated with the most heterogeneous patient, who showed large variation in \u003cem\u003eex vivo\u003c/em\u003e sensitivities to several drugs but similar sensitivity to standard of care agents, consistent with clinical response to the corresponding treatment in all lesions. These data highlight the importance of performing functional precision oncology in the context of tumor heterogeneity, although profiling of multiple samples collected at a single timepoint cannot account for tumor dynamics and potential heterogeneity of adaptive treatment responses and acquired resistance.\u003c/p\u003e \u003cp\u003eMulti-level data analyses suggested associations of PDO morphologies, molecular profiles and drug sensitivities, supporting the presence of multimodal phenotypes of CRCs. Specifically, PDOs with a cystic morphology had higher drug sensitivities in general and expressed the intestinal lineage marker CDX2\u003csup\u003e48\u003c/sup\u003e, as well as stem cell signatures of the bottom of the colonic crypt\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e. Positive correlation of CDX2 expression with sensitivity to several drugs, including standard chemotherapies, is consistent with benefit from 1st line chemotherapy in CDX2-positive metastatic CRCs\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e,\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e. However, studies of adjuvant treatment in stage II-III CRCs have suggested that loss of CDX2 is associated with chemotherapy benefit\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e,\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e. This inconsistency is not likely attributable to marker heterogeneity, since CDX2 expression is consistent between primary and metastatic tumors\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e, but context dependency can be rationalized by an indirect effect of CDX2 on chemosensitivity by regulating genes involved in drug efflux and metabolism\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e. Compared to cystic PDOs, the solid morphology showed several features of cancer aggressiveness and expressed differentiation markers associated with the top of colonic crypts, including KRT20\u003csup\u003e55,56\u003c/sup\u003e. KRT20 might mark a plastic cellular phenotype with capacity for de-differentiation into stem-like cellular states, which is a hallmark of cancer development and therapy resistance\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e,\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e. Notably, the PDO culturing conditions induce stem cell behavior and might have impacted this distinction between the two PDO phenotypes\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e, although approximately two thirds of the PDOs recapitulated the histological morphologies of the original tumors. We did not evaluate the cystic and solid PDO phenotypes in relation to the consensus histopathological growth patterns of CRLMs, since the desmoplastic and replacement growth patterns are primarily defined at the interface between the tumors and liver parenchyma\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e. Spatial transcriptomics has suggested that the growth patterns have distinct expression markers in the cancer cell compartment\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e, but we found no consistent distinctions of these markers between the cystic and solid PDOs (data not shown). Additional studies are needed to potentially consolidate the PDO-derived and tumor-derived histopathological phenotypes, including a potential association between the favorable prognosis of the desmoplastic growth pattern\u003csup\u003e\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e and the stronger drug sensitivities of cystic PDOs. The living biobank supported a favorable prognosis of patients with strong \u003cem\u003eex vivo\u003c/em\u003e drug sensitivities.\u003c/p\u003e \u003cp\u003eComprehensive analyses of sample pairs indicated strong molecular correspondence of PDOs and their original CRLMs, but also highlighted known limitations of cancer stem cell-derived cultures. Beyond the failure to model the tumor microenvironment and its influence on cancer drug activity\u003csup\u003e\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e, the PDOs showed signs of adaptation to culturing conditions, with higher metabolic and proliferative activity. It has been shown that the cellular states of PDOs can to some extent be controlled by manipulation of growth factors in the culture media\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. Careful adaptation of culturing conditions for features such as physiological hypoxia and niche factors can increase the culture success rate and even improve the fidelity of organoids as cancer models\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e. In this study, standardization of protocols was favored over sample-wise optimization, as standardization is an important consideration in clinical translation of functional precision oncology\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e. Furthermore, a large proportion of patients received chemotherapy prior to resection and sampling, and successful \u003cem\u003eex vivo\u003c/em\u003e culturing appeared to be dependent on poor chemosensitivity. This resulted in selection of cells and patients in the living biobank, including selection for poor-prognosis patients, but also reflected a clinically relevant situation where functional precision oncology is used to identify new treatment options after development of chemoresistance. A limitation of the study was the lack of co-clinical evaluation of subsequent treatments given to the patients. Radiological images were analyzed for the sampled specimens in a lesion-wise manner, but longitudinal radiological data of recurrent lesions were not available to evaluate the prospective predictive value of PDOs in an observational setting. Ultimately, the benefit and representativeness of PDOs as pharmacogenomic models need validation in prospective intervention trials\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn conclusion, this study reports the establishment of a living PDO biobank and analyses of multi-modal pharmaco-omics data of resected CRLMs in the context of tumor heterogeneity. The pharmaco-omics data represent a preclinical resource for functional precision oncology and have been made available to the scientific community. The resource also serves as reference for an ongoing intervention trial, to support the interpretation of \u003cem\u003eex vivo\u003c/em\u003e drug sensitivities into clinical \u0026ldquo;actionability\u0026rdquo;. As illustrated with examples of prospective treatment selection, top-scoring drugs are nominated only if the \u003cem\u003eex vivo\u003c/em\u003e activity is high in the prospective patient relative to patients in the reference. Furthermore, the study suggests that \u003cem\u003eex vivo\u003c/em\u003e pharmacogenomics of multiple metastatic lesions from each patient can aid the selection of treatments with lower vulnerability to tumor heterogeneity.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003ePatients and specimens\u003c/h2\u003e \u003cp\u003ePatients were included from an observational study of liver resection for metastatic CRC at Oslo University Hospital, involving prospective biobanking of metastatic lesions larger than 5 mm\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e,\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e. Patients in this study (n\u0026thinsp;=\u0026thinsp;132; Supplementary Table\u0026nbsp;1) were treated between December 2017 and 2021 and selected to increase the number with multiple CRLMs (85% of the included patients). Metastatic lesions were considered distinct when clearly separated from the neighboring lesions in the same or distant liver segments by radiological evaluation and visual examination after surgery. Longitudinal samples of recurrent CRLMs were collected from eight patients treated by one (n\u0026thinsp;=\u0026thinsp;6) or two (n\u0026thinsp;=\u0026thinsp;2) hepatic re-resections. Tumor tissue samples were collected immediately after surgery, and parallel samples were either frozen in liquid nitrogen and stored at \u0026minus;\u0026thinsp;80\u0026deg;C until processing for molecular profiling, or stored and transported in ice-cold basal media for organoid culture within 24 h.\u003c/p\u003e \u003cp\u003eClinicopathological data were prospectively collected from hospital medical records. Patients received systemic treatment according to standard national protocols, and evaluation of radiological response to neoadjuvant treatment was performed from magnetic resonance imaging (MRI) and/or computed tomography images in a lesion-wise manner. The last examination before start of chemotherapy was used as baseline and compared to the first examination after finishing each line of chemotherapy. The same imaging modality was used for pre- and post-chemotherapy comparisons if available, and MRI was preferred if both modalities were available. Data registration was finished at the time of resection for collection of PDO. The CRLMs sampled to establish PDOs were identified on images based on the segment localization according to the surgical or radiological description at the time of tissue specimen collection. If there was more than one CRLM in one segment, the largest and/or most central lesion in the segment was considered. The CRLM identification/selection and all measurements of lesion size was performed by one radiologist (KKL), and difficult cases reviewed by another radiologist (TS) and consensus reached. Each tumor was measured in its largest diameter on axial images in mm and response expressed lesion-wise as % size change (in mm) relative to baseline.\u003c/p\u003e \u003cp\u003e The study was conducted in accordance with the Declaration of Helsinki and all patients provided written informed consent. The study has been approved by the Norwegian Data Protection Authority and Regional Committee for Medical and Health Research Ethics, South-Eastern Norway (REC numbers 1.2005.1629, 2010/1805, 2017/780).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003ePatient-derived tumor organoids\u003c/h2\u003e \u003cp\u003ePDOs were established as previously described\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. In short, tumor specimens were minced with a scalpel, filtered with a 70 \u0026micro;m pore mesh, washed with ice-cold culture media, and collected by centrifugation at 400g, 4\u0026deg;C for 5 min. The tissue pellet was resuspended in Growth Factor Reduced Matrigel (Corning, NY, USA) and dispensed as 25 \u0026micro;l drops in 6-well tissue culture plates, fed with 3 ml organoid growth media supplemented with 10 \u0026micro;M Y-27632 (Merk, Darmstadt, Germany), and incubated at 37\u0026deg;C in a humidified atmosphere with 5% CO\u003csub\u003e2\u003c/sub\u003e. The composition of the organoid growth media was based on a previous study\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e and consisted of advanced DMEM/F-12 (Gibco, MA, USA) supplemented with 10 mM HEPES (Gibco), 2 mM GlutaMAX (Gibco), 50 U/ml penicillin/streptomycin (Gibco), 1x B27 Supplement (Gibco), 10 nM [Leu15]-Gastrin I (Merk), 1 mM N-acetyl-l-cysteine (Merk), 50 ng/mL EGF (Gibco), 100 ng/mL Noggin (Preprotech, NJ, UDA), 500 nM A83-01 (Tocris, Bristol, UK) and 10 \u0026micro;M SB202190 (Merk). Organoids were passaged by enzymatic digestion with TrypLE Express (Gibco) for 5 min \u0026minus;\u0026thinsp;10 min at 37\u0026deg;C. Organoid growth media was refreshed every two to four days without Y-27632. Each PDO was propagated with more than 1x10\u003csup\u003e6\u003c/sup\u003e organoids for cryopreservation, drug sensitivity assays, immunohistochemistry and sampling for DNA/RNA extraction.\u003c/p\u003e \u003cp\u003eMycoAlert Mycoplasma Detection Assay was used to ensure that PDOs were not contaminated (Lonza, Bazel, Switzerland). PDOs were authenticated by comparisons with the original tumor tissue using the AmpFLSTR Identifiler PCR Amplification Kit (Thermo Fisher Scientific, MA, USA).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eDrug sensitivity screening\u003c/h2\u003e \u003cp\u003eA medium-throughput drug screen of two drug libraries comprising totally 55 small molecule inhibitors at seven (lib2) or nine (lib1) different concentrations, as well as cetuximab at two (lib1) and five (lib2) concentrations and six drug combinations in seven concentrations was performed in PDOs (Supplementary Table\u0026nbsp;4, Supplementary Fig.\u0026nbsp;5). Lib2 (n\u0026thinsp;=\u0026thinsp;47) is an updated version of lib1 (n\u0026thinsp;=\u0026thinsp;41), in which inactive drugs and noninformative concentrations have been replaced with new drugs or combinations. The libraries were customized for CRC, and include all agents and combinations used in standard of care, as well as drugs approved for other cancer types and drugs in pre-clinical and clinical testing for CRC or other cancer types. Lib1 was screened in 117 PDOs from 59 patients, and lib2 in 116 PDOs from 55 patients, including 22 PDOs from 12 patients screened with both libraries. Drug dilutions and printing for sensitivity screens were carried out at the FIMM High Throughput Biomedicine Unit, which is hosted by the University of Helsinki and supported by HiLIFE and Biocenter Finland. Drugs were preprinted in 384-well tissue culture plates with the liquid acoustic dispensing technology Echo 550 (Labcyte Inc). The monoclonal antibody cetuximab was added manually right before screening. Each plate included 100 \u0026micro;M benzethonium chloride and 0.1% DMSO as positive and negative controls, respectively. Each PDO was screened in two parallel technical replicates. The drug screens were performed by adding 10 \u0026micro;l of 50% Matrigel in growth media and 30 \u0026micro;l of a 3% Matrigel suspension of 450\u0026ndash;600 organoids filtered with 70 \u0026micro;m mesh size to each well of a 384-well plate. PDOs were incubated for 96 h at 37\u0026deg;C in a humidified 5% CO\u003csub\u003e2\u003c/sub\u003e atmosphere. Viability was measured using the CellTiter-Glo 3D Cell Viability Assay (Promega, WI, USA) according to the manufacturer's instructions and luminescence readouts on a Victor 3 microplate reader (Perkin Elmer, Waltham, MA, USA). Drug sensitivity estimates were obtained as previously described\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. Briefly, CellTiter-Glo luminescence readouts were converted to relative viability estimates by normalization based on the median of the negative and positive control wells. Drug-response curves were modeled using logistic regression function logLogisticRegression in the R package PharmacoGx (v3.0.2) on the viability data points, and drug sensitivity scores (DSS) were calculated from the curves using the R package \u0026ldquo;DSS\u0026rdquo;\u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e. PDO growth speed was determined based on the CTG 3D assay readouts (n\u0026thinsp;=\u0026thinsp;121) or mean diameter (measured from micrographs, n\u0026thinsp;=\u0026thinsp;69) of PDOs after four days of growth in 0.1% DMSO, divided by the baseline viability or mean PDO diameter at the time of seeding, respectively.\u003c/p\u003e \u003cp\u003eDrugs with technical variation (variability between drug lots, technical replicates and/or visual inspection of the dose response curves) and drugs with no activity (maxDSS\u0026thinsp;\u0026lt;\u0026thinsp;10 and Emax\u0026thinsp;\u0026gt;\u0026thinsp;0.50) were excluded from downstream analysis (Supplementary Table\u0026nbsp;4). Correlation analysis of DSS values indicated similar activity of drugs with the same molecular target (three EGFR inhibitors, two MEK inhibitors, three PI3K/MTOR inhibitors), mode of action (three inhibitors of ATR, WEE2 or CHK1), as well as drug combinations and their respective single agents (combinations with 5-FU, combinations with SN-38; Supplementary Fig.\u0026nbsp;6). There was no correlation between the number of lesions analyzed and DSS heterogeneity based on median pair-wise Euclidean distances (Supplementary Fig.\u0026nbsp;8e-f).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eDNA/RNA extraction\u003c/h2\u003e \u003cp\u003ePDOs were collected from Matrigel and dissociated into single cells with TrypLE. RNA and DNA were isolated from both the PDOs and their corresponding fresh frozen tumor tissue samples using the Allprep DNA/RNA/miRNA Universal Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The quantity of nucleic acids was determined using a NanoDrop ND-1000 Spectrophotometer, and the quality of the RNA was assessed using a 2100 Bioanalyzer (RNA 6000 Nano kit; Agilent Technologies, Santa Clara, CA, USA). All tumor samples were found to be microsatellite stable (MSS) by the PCR-based MSI Analysis System, Version 1.2 (Promega #MD1641).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eMutation analyses\u003c/h2\u003e \u003cp\u003eSequencing was performed with a custom gene panel (Twist Bioscience, San Francisco, CA, USA) targeting all coding regions of \u003cem\u003eAPC, TP53, KRAS, NRAS, BRAF, ARID1A, PIK3CA, PTEN, EGFR, ERBB2, FBXW7, JAK1, MYC, NF1, CTNNB1, SMAD4\u003c/em\u003e, \u003cem\u003eSMAD2, CCND2\u003c/em\u003e, and \u003cem\u003eMDM2\u003c/em\u003e, as well as selected pathogenic exonuclease domain mutations of \u003cem\u003ePOLE\u003c/em\u003e (n\u0026thinsp;=\u0026thinsp;11; Supplementary Table\u0026nbsp;12). Multiple PDOs from patients with homogeneous and heterogenous mutations are shown in Supplementary Fig.\u0026nbsp;8a. Sequencing libraries were prepared from 50ng genomic DNA using the Library Preparation Enzymatic Fragmentation (EF) Kit and target enrichment workflow with DNA purification beads according to the manufacturer's instructions (Twist Bioscience). Sequencing was done on the Illumina MiniSeq system in a 2 x 73 base-pair paired-end mode using the MiniSeq High Output Kit (150-cycle; Illumina, San Diego, CA, USA). Raw sequencing reads were assessed using FastQC (v.0.11.8), before further preprocessing. Alignment to the GRCh38 human reference genome was performed using BWA (v.0.7.17), and file format conversion and refinement of sequencing reads was performed using Picard (2.19.0) and GATK (v4.1.2). Somatic variants were called with MuTect2 and annotated by ANNOVAR (version 2016Feb01). Variant calling was done in either tumor-normal or tumor-only mode depending on the availability of a matched normal tissue sample. In tumor-normal mode, only candidate somatic non-synonymous single nucleotide variants and insertion-deletions annotated as \u0026ldquo;PASS\u0026rdquo; or \u0026ldquo;clustered events\u0026rdquo; were kept and further filtered to include only loci with a minimum variant allele frequency of 5% and more than 5 reads in the tumor sample. The coverage threshold at each locus was set to a minimum of 15 reads, and a variant allele frequency less than 1.5% was accepted in the normal tissue. In tumor-only mode (98% of the PDOs and 89% of the CRLMs), additional filtering was performed to discard germline variants\u003csup\u003e\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e\u003c/sup\u003e. The median depth of coverage across the 20 genes was 570X (range 229 to 1033).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eGene expression analyses\u003c/h2\u003e \u003cp\u003ePDOs and matching tumor tissue samples were analyzed for gene expression on Affymetrix Human Transcriptome 2.0 arrays (n\u0026thinsp;=\u0026thinsp;119 PDOs and 50 matching tissue) or by RNA sequencing (n\u0026thinsp;=\u0026thinsp;92 PDOs, 5 overlapping with arrays). Seven PDOs were not analyzed due to low RNA yields.\u003c/p\u003e \u003cp\u003eMicroarray experiments were performed with 100 ng of total RNA as input and following the manufacturer\u0026rsquo;s protocol (Thermo Fisher Scientific, Waltham, MA, USA). RNA sequencing was performed in 2\u0026times;101 base-pair paired-end mode on the Illumina NovaSeq 6000 platform (Illumina) at the Oslo University Hospital Genomics Core Facility to a median depth of 71.6\u0026nbsp;million uniquely mapped read pairs per sample (10-90th percentile 60.6\u0026ndash;83.1\u0026nbsp;million reads). Sample preparation was performed by ribosomal RNA depletion using the Ribo-Zero Gold rRNA Depletion kit and sequence library generation with the TruSeq Stranded Total RNA Library Prep Gold kit (Illumina). Raw intensity CEL-files from microarray experiments were processed in two runs (PDO samples only or PDOs and matching tumor tissue together) according to the robust multi-array average approach\u003csup\u003e\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u003c/sup\u003e, using the function justRMA in the R package affy (v1.80.0)\u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e\u003c/sup\u003e and custom Entrez CDF file from Brainarray (hta20hsentrezgcdf_25.0.0)\u003csup\u003e\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e\u003c/sup\u003e. Gene annotations according to the GRCh38 genome assembly were retrieved using the function getBM in the R package biomaRt (v2.58.0)\u003csup\u003e\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e\u003c/sup\u003e.Protein-coding genes annotated with HGNC symbols were retained. Raw RNA sequencing reads were processed by adapter trimming with Trimmomatic (v.0.38), read alignment to the human reference genome GRCh38.p13 (v.41) using STAR (v.2.7.6a) with 2-pass mapping and the feature annotation file gencode.v41.annotation.gtf, as well as quantification of reads mapping to Ensembl gene ids using the HTSeq-count tool (v.2.0.2)\u003csup\u003e\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e\u003c/sup\u003e. Annotations were converted to HGNC symbols with biomaRt. Count data were normalized as transcripts per kilobase million (TPM). For differential gene expression analyses, count data were normalized with trimmed mean of M-values (TMM) and summarized as counts per million (CPM) using the R package edgeR. Normalized gene expression estimates were log2-transformed (after adding a constant of 0.1 to TPM values).\u003c/p\u003e \u003cp\u003eSimilarity of PDOs to CRC tissue on the transcriptomic level was evaluated with the R package CancerCellNet (v.0.2.0) using the function broadClass_predict and a classifier trained with CRC tissue samples from TCGA, obtained with the function broadClass_train and default settings\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. Classification according to iCMS was performed using the approach and gene template described in the original publication\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e\u003c/sup\u003e. Most PDOs (94%) derived from distinct lesions of the same patient were classified in the same iCMS group (Supplementary Fig.\u0026nbsp;8b). Principal components analysis was performed using the R package FactoMineR (v2.9) on genes (n\u0026thinsp;=\u0026thinsp;3,000) with the highest cross-sample 10-90th percentile range. Gene set enrichment analysis of a custom gene set collection (n\u0026thinsp;=\u0026thinsp;106) was performed with the R package GSA (v1.03.0)\u003csup\u003e\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u003c/sup\u003e for sample group comparisons, including false discovery rate adjustment of the p-value, and the R package GSVA (v1.50.0)\u003csup\u003e\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e\u003c/sup\u003e for single-sample scoring by the gene set variation analysis approach.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eHistopathology and multiplex immunohistochemistry\u003c/h2\u003e \u003cp\u003ePDOs were fixed with 4% paraformaldehyde in PBS for 20 minutes and embedded in the protective gel from Shandon Cytoblock Cell Block Preparation System (Thermo Scientific). Fixed cells were dehydrated, infused with paraffin wax, and embedded in paraffin to create histology blocks. The blocks were cut into 4 \u0026micro;m sections for staining with hematoxylin and eosin (HE) and antibodies.\u003c/p\u003e \u003cp\u003eMorphological resemblance between tumor tissues and their corresponding PDOs was evaluated by a pathologist on HE stained sections from 46 paired samples from 26 patients. The comparisons were based on structural patterns of the epithelial cell compartment, including small/acinar and dilated/cystic gland-like structures, cribriform/complex structures, solid growth, cytoplasmatic vacuoles, and presence of single cells. Separately, HE stained sections of all 213-paraffin embedded PDOs were used to determine the morphological cystic, solid and mixed phenotypes as described in the results.\u003c/p\u003e \u003cp\u003eFluorescence-based multiplex immunohistochemistry and digital image analyses were used to analyze \u003cem\u003ein situ\u003c/em\u003e expression of fourteen proteins in 136 PDOs and two corresponding tumor tissue samples from 67 patients (Supplementary Fig.\u0026nbsp;7a-d). Fluorescence staining was based on Opal kits (NEL810001KT (includes fluorophores Opal 520, 570 and 690, DAPI, antibody diluent and anti-mouse/rabbit HRP secondary antibodies) and FP1495001KT (Opal 620) from Akoya Biosciences (Marlborough, MA, USA) and reagents were used according to the manufacturer's recommendations (Akoya Biosciences), unless otherwise noted below. Five multiplex stains were developed using primary antibodies against the following targets (see Supplementary Table\u0026nbsp;13 for staining sequence and pairing/dilution of fluorophores): CDX2 (1:400, clone EPR2764Y, Cell Marque, CA, USA), KRT20 (1:400 and 1:1000, clone Ks20.8, Agilent Dako, Glostrup, Denmark), CDH1 (1:10.000, clone 36, BD Biosciences, NJ, USA), KRT7 (1:400, clone OV-TL 12/30, Agilent Dako), KI67 (undiluted, clone MIB-1, DAKO/Agilent), HSF1 (1:100 clone D3L8I, cell signaling technology, MA, USA), UGT1A (1:300 and 1:500, clone B-4, Santa Cruz, CA, USA), RIPK1 (1:50 clone E8S7U, cell signaling technology), TP53 (1:12, clone DO-7, Agilent Dako), RCC2 (1:100 clone D14F3, cell signaling technology), ABCG2 (1:50, clone D5V2K, cell signaling technology), CFTR (1:2000, clone 24\u0026thinsp;\u0026minus;\u0026thinsp;1, R\u0026amp;D systems, MN, USA), ERBB2 (1:500, polyclonal, catalogue number AO485, DAKO/Agilent), and ABCB1 (1:100, clone E1Y7S, cell signaling technology). Deparaffinization and the initial antigen retrieval were performed in a PT-link module (DAKO/Agilent) at 97\u003csup\u003eo\u003c/sup\u003eC for 20 min, using 3-in-1 high-pH buffer (catalogue number K8004, DAKO/Agilent). The following rounds of heat treatment for antibody stripping were also performed in the PT-link module, however instead using high/low pH buffers from Akoya (catalogue numbers AR9001KT \u0026amp; AR6001KT, respectively), specified in Supplementary Table\u0026nbsp;13. All primary antibodies were incubated for 30 min at room temperature. Cell nuclei were stained with DAPI prior to mounting with Prolong Diamond Antifade Mountant (Life Technologies/Thermo Fisher Scientific). The sections were scanned at 10x magnification, and multispectral images were acquired at 20x magnification using the Vectra 3 Automated Quantitative Pathology Imaging System (Akoya Biosciences). Image analysis was performed using inForm Image Analysis Software (Akoya Biosciences), which used a supervised machine learning algorithm to accurately segment the cells and nuclei. All images were manually checked after segmentation and poor-quality regions of the samples were excluded (e.g. due to sample folds). Mean relative expression of each protein was quantified and normalized to the total cellular content in each PDO image. KRT7 and ERBB2 were excluded from analysis due to no or very low expression levels, respectively. Protein expression heterogeneity was less pronounced in intra-patient than inter-patient comparisons, as indicated by lower heterogeneity scores between patient-matched sample pairs, calculated using Euclidean distances of protein expression profiles (Supplementary Fig.\u0026nbsp;8c). There was no correlation between the number of lesions analyzed and protein expression heterogeneity based on median pair-wise Euclidean distances (Supplementary Fig.\u0026nbsp;8d).\u003c/p\u003e \u003cp\u003eERBB2 protein expression (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ea) was also visualized using the DAB chromogen on the Autostainer Link 48 system. Deparaffinization and antigen retrieval were performed in the PT-link module, as stated above, except that low-pH buffer was used (catalogue number K8005, DAKO/Agilent). Primary antibody against ERBB2 (1:50, clone CB11, Leica Biosystems, Nussloch, Germany) was incubated for 30 min at room temperature. Anti-mouse EnVision\u0026thinsp;+\u0026thinsp;System HRP labelled polymer (catalogue number K4001, DAKO/Agilent) was incubated for 30 min prior to incubation with DAB chromogen (catalogue number K3468, DAKO/Agilent) for 10 min. Hematoxylin (catalogue number 01800, Histolab) was used as counterstain and incubated for 30 sec. Slides were rinsed in water before dehydrating and mounting.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eStatistics\u003c/h2\u003e \u003cp\u003eStatistical analyses and data visualization were performed with R software version 4.2.1, Graph Pad Prism version 10.2.2 or SPSS version 29. Clinicopathological variables were summarized and compared using the \"gtsummary\" v1.7.0 R package. Chi-squared tests and visualization were performed with the \"ggstatsplot\" v0.11.0 package in R. Spearman and Pearson correlation analyses were performed using the \"cor\" function. Wilcoxon tests were performed using the \"wilcox_test\" function from the \"rstatix\" v0.7.2 R package. Kruskal\u0026ndash;Wallis tests of three groups of continuous variables were performed using the \u0026ldquo;ggpubr\u0026rdquo; v0.6.0 R package. Euclidean distances were calculated using the dist function in R. Heatmaps were generated using the \"pheatmap\" v1.0.12 R package with the \"complete\" clustering method. Survival analyses were conducted using the \"survival\" v3.3-5 and \"survminer\" v0.4.9 R packages. Overall survival was estimated from the date of liver surgery and using death from any cause as events. Patients still alive at the end of the study were censored at the last follow-up or four years after liver surgery. Cox proportional hazards models were used to evaluate prognostic associations of clinicopathological variables. P-values lower than 0.05 were considered statistically significant. All statistical tests were two-sided.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003eData availability\u003c/p\u003e\n\u003cp\u003eAll data generated and analyzed in this study are available in public repositories or as Data Sets in the manuscript. Mutation data (n=20 genes in 150 PDOs and 74 tumor tissue samples) are available as Data Set 1. The microarray gene expression data (n=119 PDOs and 50 tumor tissue samples) have been deposited to the NCBI\u0026rsquo;s Gene Expression Omnibus under accession code GSE294511 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE294511]. The raw RNA sequencing data (n=92 PDOs) are considered patient identifiable and subject to secure storage regulations in accordance with Norwegian legislation and the ethical approval of the study by the Regional Committee for Medical and Health Research Ethics, South Eastern Norway, but count data of protein-coding genes are available as Data Set 2. Protein expression data from multiplex immunohistochemistry (n=14 proteins in 136 PDOs from 67 patients) are available as Data Set 3. Drug sensitivity scores (library 1: n=41 drugs in 117 PDOs; and library 2: n=47 drugs in 116 PDOs) are available as Data Set 4.\u003c/p\u003e\n\u003cp\u003eCode availability\u003c/p\u003e\n\u003cp\u003eAll data processing and analyses were performed with published software packages and computer code and have been described and cited in the Results and/or Methods sections. No custom code was developed in the study.\u003c/p\u003e\n\u003cp\u003eAcknowledgments\u003c/p\u003e\n\u003cp\u003eWe would like to thank the study nurses Magdalena Maria Kowalewska-Harbiyeli and Vlora Krasniqi Hulaj for their help in collecting patient samples and clinical data. We also appreciate the technical support provided by the principal engineer Mette Ekn\u0026aelig;s and head engineer Merete Hektoen. Additionally, we acknowledge the contributions of former lab members Barbara Niederdorfer, Jonas Langerud, Jarle Bruun, Peter W. Eide, Christer A. Andreassen and Kaja C. G. Berg in the initial stages of this project.\u003c/p\u003e\n\u003cp\u003eThe study was funded by grants from the Norwegian Cancer Society (project numbers 182759 and 223319 to RAL, project numbers 208336 and 246954 to AS, project number 297971 to KK), the South-Eastern Norway Regional Health Authority (project number 2023101 to AS and 2024108 to RAL), the Research Council of Norway (project number 287899 to AS), and the Oslo University Hospital (\u0026ldquo;Strategic research area, 2019-2024\u0026rdquo; to RAL and AS).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003evan Renterghem, A.W.J., van de Haar, J. \u0026amp; Voest, E.E. Functional precision oncology using patient-derived assays: bridging genotype and phenotype. \u003cem\u003eNat Rev Clin Oncol\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 305-317 (2023).\u003c/li\u003e\n\u003cli\u003eKornauth, C.\u003cem\u003e et al.\u003c/em\u003e Functional Precision Medicine Provides Clinical Benefit in Advanced Aggressive Hematologic Cancers and Identifies Exceptional Responders. \u003cem\u003eCancer Discov\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 372-387 (2022).\u003c/li\u003e\n\u003cli\u003eMalani, D.\u003cem\u003e et al.\u003c/em\u003e Implementing a Functional Precision Medicine Tumor Board for Acute Myeloid Leukemia. \u003cem\u003eCancer Discov\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 388-401 (2022).\u003c/li\u003e\n\u003cli\u003eOoft, S.N.\u003cem\u003e et al.\u003c/em\u003e Prospective experimental treatment of colorectal cancer patients based on organoid drug responses. \u003cem\u003eESMO Open\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 100103 (2021).\u003c/li\u003e\n\u003cli\u003eSato, T.\u003cem\u003e et al.\u003c/em\u003e Long-term expansion of epithelial organoids from human colon, adenoma, adenocarcinoma, and Barrett\u0026apos;s epithelium. \u003cem\u003eGastroenterology\u003c/em\u003e \u003cstrong\u003e141\u003c/strong\u003e, 1762-72 (2011).\u003c/li\u003e\n\u003cli\u003evan de Wetering, M.\u003cem\u003e et al.\u003c/em\u003e Prospective derivation of a living organoid biobank of colorectal cancer patients. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e161\u003c/strong\u003e, 933-45 (2015).\u003c/li\u003e\n\u003cli\u003eFujii, M.\u003cem\u003e et al.\u003c/em\u003e A Colorectal Tumor Organoid Library Demonstrates Progressive Loss of Niche Factor Requirements during Tumorigenesis. \u003cem\u003eCell Stem Cell\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 827-38 (2016).\u003c/li\u003e\n\u003cli\u003eOoft, S.N.\u003cem\u003e et al.\u003c/em\u003e Patient-derived organoids can predict response to chemotherapy in metastatic colorectal cancer patients. \u003cem\u003eSci Transl Med\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e(2019).\u003c/li\u003e\n\u003cli\u003eVlachogiannis, G.\u003cem\u003e et al.\u003c/em\u003e Patient-derived organoids model treatment response of metastatic gastrointestinal cancers. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e359\u003c/strong\u003e, 920-926 (2018).\u003c/li\u003e\n\u003cli\u003eNarasimhan, V.\u003cem\u003e et al.\u003c/em\u003e Medium-throughput Drug Screening of Patient-derived Organoids from Colorectal Peritoneal Metastases to Direct Personalized Therapy. \u003cem\u003eClin Cancer Res\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 3662-3670 (2020).\u003c/li\u003e\n\u003cli\u003eLetai, A. Functional Precision Medicine: Putting Drugs on Patient Cancer Cells and Seeing What Happens. \u003cem\u003eCancer Discov\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 290-292 (2022).\u003c/li\u003e\n\u003cli\u003eJensen, L.H.\u003cem\u003e et al.\u003c/em\u003e Precision medicine applied to metastatic colorectal cancer using tumor-derived organoids and in-vitro sensitivity testing: a phase 2, single-center, open-label, and non-comparative study. \u003cem\u003eJ Exp Clin Cancer Res\u003c/em\u003e \u003cstrong\u003e42\u003c/strong\u003e, 115 (2023).\u003c/li\u003e\n\u003cli\u003eMcGranahan, N. \u0026amp; Swanton, C. Clonal Heterogeneity and Tumor Evolution: Past, Present, and the Future. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e168\u003c/strong\u003e, 613-628 (2017).\u003c/li\u003e\n\u003cli\u003eDagogo-Jack, I. \u0026amp; Shaw, A.T. Tumour heterogeneity and resistance to cancer therapies. \u003cem\u003eNat Rev Clin Oncol\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 81-94 (2018).\u003c/li\u003e\n\u003cli\u003eHu, Z.\u003cem\u003e et al.\u003c/em\u003e Quantitative evidence for early metastatic seeding in colorectal cancer. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, 1113-1122 (2019).\u003c/li\u003e\n\u003cli\u003eDang, H.X.\u003cem\u003e et al.\u003c/em\u003e The clonal evolution of metastatic colorectal cancer. \u003cem\u003eSci Adv\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, eaay9691 (2020).\u003c/li\u003e\n\u003cli\u003eCai, J.\u003cem\u003e et al.\u003c/em\u003e Single-cell exome sequencing reveals polyclonal seeding and TRPS1 mutations in colon cancer metastasis. \u003cem\u003eSignal Transduct Target Ther\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 247 (2024).\u003c/li\u003e\n\u003cli\u003eLangerud, J.\u003cem\u003e et al.\u003c/em\u003e Multiregional transcriptomics identifies congruent consensus subtypes with prognostic value beyond tumor heterogeneity of colorectal cancer. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 4342 (2024).\u003c/li\u003e\n\u003cli\u003eMoorman, A.\u003cem\u003e et al.\u003c/em\u003e Progressive plasticity during colorectal cancer metastasis. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e637\u003c/strong\u003e, 947-954 (2025).\u003c/li\u003e\n\u003cli\u003eBrunsell, T.H.\u003cem\u003e et al.\u003c/em\u003e Heterogeneous radiological response to neoadjuvant therapy is associated with poor prognosis after resection of colorectal liver metastases. \u003cem\u003eEur J Surg Oncol\u003c/em\u003e \u003cstrong\u003e45\u003c/strong\u003e, 2340-2346 (2019).\u003c/li\u003e\n\u003cli\u003eZhou, J.\u003cem\u003e et al.\u003c/em\u003e Mapping lesion-specific response and progression dynamics and inter-organ variability in metastatic colorectal cancer. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 417 (2023).\u003c/li\u003e\n\u003cli\u003eOu, F.S.\u003cem\u003e et al.\u003c/em\u003e Evaluation of Intratumoral Response Heterogeneity in Metastatic Colorectal Cancer and Its Impact on Patient Overall Survival: Findings from 10,551 Patients in the ARCAD Database. \u003cem\u003eCancers (Basel)\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e(2023).\u003c/li\u003e\n\u003cli\u003eGeevimaan, K.\u003cem\u003e et al.\u003c/em\u003e Patient-Derived Organoid Serves as a Platform for Personalized Chemotherapy in Advanced Colorectal Cancer Patients. \u003cem\u003eFront Oncol\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 883437 (2022).\u003c/li\u003e\n\u003cli\u003eSchumacher, D.\u003cem\u003e et al.\u003c/em\u003e Heterogeneous pathway activation and drug response modelled in colorectal-tumor-derived 3D cultures. \u003cem\u003ePLoS Genet\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, e1008076 (2019).\u003c/li\u003e\n\u003cli\u003eRoerink, S.F.\u003cem\u003e et al.\u003c/em\u003e Intra-tumour diversification in colorectal cancer at the single-cell level. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e556\u003c/strong\u003e, 457-462 (2018).\u003c/li\u003e\n\u003cli\u003eKim, S.C.\u003cem\u003e et al.\u003c/em\u003e Multifocal Organoid Capturing of Colon Cancer Reveals Pervasive Intratumoral Heterogenous Drug Responses. \u003cem\u003eAdv Sci (Weinh)\u003c/em\u003e, e2103360 (2021).\u003c/li\u003e\n\u003cli\u003eMo, S.\u003cem\u003e et al.\u003c/em\u003e Patient-Derived Organoids from Colorectal Cancer with Paired Liver Metastasis Reveal Tumor Heterogeneity and Predict Response to Chemotherapy. \u003cem\u003eAdv Sci (Weinh)\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, e2204097 (2022).\u003c/li\u003e\n\u003cli\u003eThng, D.K.H.\u003cem\u003e et al.\u003c/em\u003e A functional personalised oncology approach against metastatic colorectal cancer in matched patient derived organoids. \u003cem\u003eNPJ Precis Oncol\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 52 (2024).\u003c/li\u003e\n\u003cli\u003eBruun, J.\u003cem\u003e et al.\u003c/em\u003e Patient-Derived Organoids from Multiple Colorectal Cancer Liver Metastases Reveal Moderate Intra-patient Pharmacotranscriptomic Heterogeneity. \u003cem\u003eClin Cancer Res\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 4107-4119 (2020).\u003c/li\u003e\n\u003cli\u003eJoanito, I.\u003cem\u003e et al.\u003c/em\u003e Single-cell and bulk transcriptome sequencing identifies two epithelial tumor cell states and refines the consensus molecular classification of colorectal cancer. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e54\u003c/strong\u003e, 963-975 (2022).\u003c/li\u003e\n\u003cli\u003ePeng, D.\u003cem\u003e et al.\u003c/em\u003e Evaluating the transcriptional fidelity of cancer models. \u003cem\u003eGenome Med\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 73 (2021).\u003c/li\u003e\n\u003cli\u003eVecchione, L.\u003cem\u003e et al.\u003c/em\u003e A Vulnerability of a Subset of Colon Cancers with Potential Clinical Utility. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e165\u003c/strong\u003e, 317-30 (2016).\u003c/li\u003e\n\u003cli\u003eYaeger, R.\u003cem\u003e et al.\u003c/em\u003e Clinical Sequencing Defines the Genomic Landscape of Metastatic Colorectal Cancer. \u003cem\u003eCancer Cell\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 125-136 e3 (2018).\u003c/li\u003e\n\u003cli\u003eMendelaar, P.A.J.\u003cem\u003e et al.\u003c/em\u003e Whole genome sequencing of metastatic colorectal cancer reveals prior treatment effects and specific metastasis features. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 574 (2021).\u003c/li\u003e\n\u003cli\u003eBergsland, C.H.\u003cem\u003e et al.\u003c/em\u003e Prediction of relapse-free survival according to adjuvant chemotherapy and regulator of chromosome condensation 2 (RCC2) expression in colorectal cancer. \u003cem\u003eESMO Open\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, e001040 (2020).\u003c/li\u003e\n\u003cli\u003eSmeby, J.\u003cem\u003e et al.\u003c/em\u003e Molecular correlates of sensitivity to PARP inhibition beyond homologous recombination deficiency in pre-clinical models of colorectal cancer point to wild-type TP53 activity. \u003cem\u003eEBioMedicine\u003c/em\u003e \u003cstrong\u003e59\u003c/strong\u003e, 102923 (2020).\u003c/li\u003e\n\u003cli\u003eKryeziu, K.\u003cem\u003e et al.\u003c/em\u003e Increased sensitivity to SMAC mimetic LCL161 identified by longitudinal ex vivo pharmacogenomics of recurrent, KRAS mutated rectal cancer liver metastases. \u003cem\u003eJ Transl Med\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 384 (2021).\u003c/li\u003e\n\u003cli\u003eArnold, D.\u003cem\u003e et al.\u003c/em\u003e Prognostic and predictive value of primary tumour side in patients with RAS wild-type metastatic colorectal cancer treated with chemotherapy and EGFR directed antibodies in six randomized trials. \u003cem\u003eAnn Oncol\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 1713-1729 (2017).\u003c/li\u003e\n\u003cli\u003eMeric-Bernstam, F.\u003cem\u003e et al.\u003c/em\u003e Pertuzumab plus trastuzumab for HER2-amplified metastatic colorectal cancer (MyPathway): an updated report from a multicentre, open-label, phase 2a, multiple basket study. \u003cem\u003eLancet Oncol\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 518-530 (2019).\u003c/li\u003e\n\u003cli\u003eGupta, R.\u003cem\u003e et al.\u003c/em\u003e Pertuzumab Plus Trastuzumab in Patients With Colorectal Cancer With ERBB2 Amplification or ERBB2/3 Mutations: Results From the TAPUR Study. \u003cem\u003eJCO Precis Oncol\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, e2200306 (2022).\u003c/li\u003e\n\u003cli\u003eZhang, Y.\u003cem\u003e et al.\u003c/em\u003e Identification of an Activating Mutation in the Extracellular Domain of HER2 Conferring Resistance to Pertuzumab. \u003cem\u003eOnco Targets Ther\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 11597-11608 (2019).\u003c/li\u003e\n\u003cli\u003eDiwanji, D.\u003cem\u003e et al.\u003c/em\u003e Structures of the HER2-HER3-NRG1beta complex reveal a dynamic dimer interface. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e600\u003c/strong\u003e, 339-343 (2021).\u003c/li\u003e\n\u003cli\u003eDi Nicolantonio, F.\u003cem\u003e et al.\u003c/em\u003e Precision oncology in metastatic colorectal cancer - from biology to medicine. \u003cem\u003eNat Rev Clin Oncol\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 506-525 (2021).\u003c/li\u003e\n\u003cli\u003eSveen, A., Kopetz, S. \u0026amp; Lothe, R.A. Biomarker-guided therapy for colorectal cancer: strength in complexity. \u003cem\u003eNat Rev Clin Oncol\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 11-32 (2020).\u003c/li\u003e\n\u003cli\u003eBazarbashi, S.\u003cem\u003e et al.\u003c/em\u003e Efficacy of Chemotherapy Rechallenge Versus Regorafenib or Trifluridine/Tipiracil in Third-Line Setting of Metastatic Colorectal Cancer: A Multicenter Retrospective Comparative Study. \u003cem\u003eJCO Glob Oncol\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, e2300461 (2024).\u003c/li\u003e\n\u003cli\u003eSartore-Bianchi, A.\u003cem\u003e et al.\u003c/em\u003e Circulating tumor DNA to guide rechallenge with panitumumab in metastatic colorectal cancer: the phase 2 CHRONOS trial. \u003cem\u003eNat Med\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 1612-1618 (2022).\u003c/li\u003e\n\u003cli\u003eCiardiello, D.\u003cem\u003e et al.\u003c/em\u003e The role of anti-EGFR rechallenge in metastatic colorectal cancer, from available data to future developments: A systematic review. \u003cem\u003eCancer Treat Rev\u003c/em\u003e \u003cstrong\u003e124\u003c/strong\u003e, 102683 (2024).\u003c/li\u003e\n\u003cli\u003eBadia-Ramentol, J.\u003cem\u003e et al.\u003c/em\u003e The prognostic potential of CDX2 in colorectal cancer: Harmonizing biology and clinical practice. \u003cem\u003eCancer Treat Rev\u003c/em\u003e \u003cstrong\u003e121\u003c/strong\u003e, 102643 (2023).\u003c/li\u003e\n\u003cli\u003eZhang, B.Y.\u003cem\u003e et al.\u003c/em\u003e Lack of Caudal-Type Homeobox Transcription Factor 2 Expression as a Prognostic Biomarker in Metastatic Colorectal Cancer. \u003cem\u003eClin Colorectal Cancer\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 124-128 (2017).\u003c/li\u003e\n\u003cli\u003eAasebo, K.\u003cem\u003e et al.\u003c/em\u003e CDX2: A Prognostic Marker in Metastatic Colorectal Cancer Defining a Better BRAF Mutated and a Worse KRAS Mutated Subgroup. \u003cem\u003eFront Oncol\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 8 (2020).\u003c/li\u003e\n\u003cli\u003eDalerba, P.\u003cem\u003e et al.\u003c/em\u003e CDX2 as a Prognostic Biomarker in Stage II and Stage III Colon Cancer. \u003cem\u003eN Engl J Med\u003c/em\u003e \u003cstrong\u003e374\u003c/strong\u003e, 211-22 (2016).\u003c/li\u003e\n\u003cli\u003eBruun, J.\u003cem\u003e et al.\u003c/em\u003e Prognostic, predictive, and pharmacogenomic assessments of CDX2 refine stratification of colorectal cancer. \u003cem\u003eMol Oncol\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 1639-1655 (2018).\u003c/li\u003e\n\u003cli\u003eShigematsu, Y.\u003cem\u003e et al.\u003c/em\u003e CDX2 expression is concordant between primary colorectal cancer lesions and corresponding liver metastases independent of chemotherapy: a single-center retrospective study in Japan. \u003cem\u003eOncotarget\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 17056-17065 (2018).\u003c/li\u003e\n\u003cli\u003eDelhorme, J.B.\u003cem\u003e et al.\u003c/em\u003e CDX2 controls genes involved in the metabolism of 5-fluorouracil and is associated with reduced efficacy of chemotherapy in colorectal cancer. \u003cem\u003eBiomed Pharmacother\u003c/em\u003e \u003cstrong\u003e147\u003c/strong\u003e, 112630 (2022).\u003c/li\u003e\n\u003cli\u003eChan, C.W.\u003cem\u003e et al.\u003c/em\u003e Gastrointestinal differentiation marker Cytokeratin 20 is regulated by homeobox gene CDX1. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e \u003cstrong\u003e106\u003c/strong\u003e, 1936-41 (2009).\u003c/li\u003e\n\u003cli\u003eLugli, A., Tzankov, A., Zlobec, I. \u0026amp; Terracciano, L.M. Differential diagnostic and functional role of the multi-marker phenotype CDX2/CK20/CK7 in colorectal cancer stratified by mismatch repair status. \u003cem\u003eMod Pathol\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 1403-12 (2008).\u003c/li\u003e\n\u003cli\u003eShimokawa, M.\u003cem\u003e et al.\u003c/em\u003e Visualization and targeting of LGR5(+) human colon cancer stem cells. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e545\u003c/strong\u003e, 187-192 (2017).\u003c/li\u003e\n\u003cli\u003eMehta, A. \u0026amp; Stanger, B.Z. Lineage Plasticity: The New Cancer Hallmark on the Block. \u003cem\u003eCancer Res\u003c/em\u003e \u003cstrong\u003e84\u003c/strong\u003e, 184-191 (2024).\u003c/li\u003e\n\u003cli\u003eLatacz, E.\u003cem\u003e et al.\u003c/em\u003e Histopathological growth patterns of liver metastasis: updated consensus guidelines for pattern scoring, perspectives and recent mechanistic insights. \u003cem\u003eBr J Cancer\u003c/em\u003e \u003cstrong\u003e127\u003c/strong\u003e, 988-1013 (2022).\u003c/li\u003e\n\u003cli\u003eFleischer, J.R.\u003cem\u003e et al.\u003c/em\u003e Molecular differences of angiogenic versus vessel co-opting colorectal cancer liver metastases at single-cell resolution. \u003cem\u003eMol Cancer\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 17 (2023).\u003c/li\u003e\n\u003cli\u003eStremitzer, S.\u003cem\u003e et al.\u003c/em\u003e Immune phenotype and histopathological growth pattern in patients with colorectal liver metastases. \u003cem\u003eBr J Cancer\u003c/em\u003e \u003cstrong\u003e122\u003c/strong\u003e, 1518-1524 (2020).\u003c/li\u003e\n\u003cli\u003eJin, M.Z. \u0026amp; Jin, W.L. The updated landscape of tumor microenvironment and drug repurposing. \u003cem\u003eSignal Transduct Target Ther\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 166 (2020).\u003c/li\u003e\n\u003cli\u003eWalaas, G.A.\u003cem\u003e et al.\u003c/em\u003e Physiological hypoxia improves growth and functional differentiation of human intestinal epithelial organoids. \u003cem\u003eFront Immunol\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 1095812 (2023).\u003c/li\u003e\n\u003cli\u003eBose, S.\u003cem\u003e et al.\u003c/em\u003e A path to translation: How 3D patient tumor avatars enable next generation precision oncology. \u003cem\u003eCancer Cell\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, 1448-1453 (2022).\u003c/li\u003e\n\u003cli\u003eBrunsell, T.H.\u003cem\u003e et al.\u003c/em\u003e High Concordance and Negative Prognostic Impact of RAS/BRAF/PIK3CA Mutations in Multiple Resected Colorectal Liver Metastases. \u003cem\u003eClin Colorectal Cancer\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, e26-e47 (2020).\u003c/li\u003e\n\u003cli\u003eYadav, B.\u003cem\u003e et al.\u003c/em\u003e Quantitative scoring of differential drug sensitivity for individually optimized anticancer therapies. \u003cem\u003eSci Rep\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 5193 (2014).\u003c/li\u003e\n\u003cli\u003eMoosavi, S.H.\u003cem\u003e et al.\u003c/em\u003e Molecular prognostic factors for liver transplantation of unresectable metastatic colorectal cancer. \u003cem\u003eBr J Cancer\u003c/em\u003e (2025).\u003c/li\u003e\n\u003cli\u003eIrizarry, R.A.\u003cem\u003e et al.\u003c/em\u003e Exploration, normalization, and summaries of high density oligonucleotide array probe level data. \u003cem\u003eBiostatistics\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 249-264 (2003).\u003c/li\u003e\n\u003cli\u003eGautier, L., Cope, L., Bolstad, B.M. \u0026amp; Irizarry, R.A. affy--analysis of Affymetrix GeneChip data at the probe level. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 307-15 (2004).\u003c/li\u003e\n\u003cli\u003eDai, M.\u003cem\u003e et al.\u003c/em\u003e Evolving gene/transcript definitions significantly alter the interpretation of GeneChip data. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, e175 (2005).\u003c/li\u003e\n\u003cli\u003eDurinck, S., Spellman, P.T., Birney, E. \u0026amp; Huber, W. Mapping identifiers for the integration of genomic datasets with the R/Bioconductor package biomaRt. \u003cem\u003eNat. Protoc.\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 1184-1191 (2009).\u003c/li\u003e\n\u003cli\u003eEilertsen, I.\u003cem\u003e et al.\u003c/em\u003e Technical differences between sequencing and microarray platforms impact transcriptomic subtyping of colorectal cancer. \u003cem\u003eCancer Lett.\u003c/em\u003e \u003cstrong\u003e469\u003c/strong\u003e, 246-255 (2020).\u003c/li\u003e\n\u003cli\u003eJoanito, I.\u003cem\u003e et al.\u003c/em\u003e Single-cell and bulk transcriptome sequencing identifies two epithelial tumor cell states and refines the consensus molecular classification of colorectal cancer. \u003cem\u003eNat. Genet.\u003c/em\u003e \u003cstrong\u003e54\u003c/strong\u003e, 963-975 (2022).\u003c/li\u003e\n\u003cli\u003eEfron, B. \u0026amp; Tibshirani, R. On testing the significance of sets of genes. \u003cem\u003eAnn. Appl. Stat.\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 107-129 (2007).\u003c/li\u003e\n\u003cli\u003eHanzelmann, S., Castelo, R. \u0026amp; Guinney, J. GSVA: gene set variation analysis for microarray and RNA-seq data. \u003cem\u003eBMC Bioinformatics\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 7 (2013).\u003cbr\u003e \u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6507406/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6507406/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study reports the establishment and pharmacogenomics analyses of tumor heterogeneity in a living biobank of tumor organoids of 213 liver metastases from 102 patients with metastatic colorectal cancer. Successful organoid culturing reflected poorer chemosensitivity and patient survival. Molecular fidelity was demonstrated in tumor-organoid sample pairs, and multi-modal phenotypes were proposed based on organoid morphologies. Cystic morphology was associated with intestinal stem cell markers and higher drug sensitivities, and solid morphology with markers of cancer cell plasticity and aggressiveness. Potential to identify treatments with less vulnerability to tumor heterogeneity was supported by multi-lesion analyses in 65 patients. Complexity of clinical translation was illustrated by two prospective cases of pharmacogenomics-guided treatment, including successful chemotherapy rechallenge and targeted therapy resistance in cancers with low and high tumor heterogeneity, respectively. All pharmacogenomics data are available as a functional oncology resource and serve as reference for an ongoing intervention trial, supporting the interpretation of ex vivo drug sensitivities into prospective clinical \u0026ldquo;actionability\u0026rdquo;.\u003c/p\u003e","manuscriptTitle":"Ex vivo modeling of morphological, molecular and pharmacological tumor heterogeneity of metastatic colorectal cancer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-09 10:41:21","doi":"10.21203/rs.3.rs-6507406/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"d334ae09-279e-437a-b6f9-57820a7b2fb9","owner":[],"postedDate":"May 9th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":47987058,"name":"Biological sciences/Cancer/Cancer models"},{"id":47987059,"name":"Biological sciences/Cancer/Tumour heterogeneity"},{"id":47987060,"name":"Biological sciences/Cancer/Gastrointestinal cancer/Colorectal cancer"},{"id":47987061,"name":"Biological sciences/Cancer/Cancer therapy"},{"id":47987062,"name":"Biological sciences/Cancer/Metastasis"}],"tags":[],"updatedAt":"2025-06-04T16:07:17+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-09 10:41:21","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6507406","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6507406","identity":"rs-6507406","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}