Overcoming brain-derived therapeutic resistance in HER2+ breast cancer brain metastasis

19 Brain metastasis of HER2+ breast cancer occurs in about 50% of all women with metastatic HER2+ 20 breast cancer and confers poor prognosis for patients. Despite effective HER2 -targeted treatments of 21 peripheral HER2+ breast cancer with Trastuzumab +/ -HER2 inhibitors, limited brain permeability 22 renders these treatments inefficient for HER2+ breast cancer brain metastasis ( BCBM). The scarcity of 23 suitable patient-derived in-vivo models for HER2+ BCBM has compromised the study of molecular 24 mechanisms that promote growth and therapeutic resistance in brain metastasis. We have generated and 25 characterized new HER2+ BCBM cells (BCBM94) isolated from a patient HER2+ brain metastasis. 26 Repeated hematogenic xenografting of BCBM94 consistently generated BCBM in mice . The clinically 27 used receptor tyrosine kinase inhibitor (RTKi) Lapatinib blocked phosphorylation of all ErbB1-4 28 receptors and induced the intrinsic apoptosis pathway in BCBM94. Neuregulin-1 (NRG1), a ligand for 29 ErbB3 and ErbB4 that is abundantly expressed in the brain, was able to rescue L apatinib-induced 30 apoptosis and clonogenic ability in BCBM94 and in HER2+ BT474. ErbB3 was essential to mediate the 31 NRG1-induced survival pathway that involved PI3K -AKT signalling and the phosphorylation of BAD 32 at serine 136 to prevent apoptosis . High throughput RTK i screening identified the brain penetrable 33 Poziotinib as highly potent compound to reduce cell viability in HER2+ BCBM in the presence of NRG1. 34 Successful in-vivo ablation of BCBM94- and BT474-derived HER2+ brain tumors was achieved upon 35 two weeks of treatment with Poziotinib. MRI revealed BCBM remission upon poziotinib, but not with 36 Lapatinib treatment. In conclusion, we have established a new patient -derived HER2+ BCBM in-vivo 37 model and identified Poziotinib as highly efficacious RTKi with excellent brain penetrability that 38 abrogated HER2+ BCBM brain tumors in our mouse models. 39 40

HER2, breast cancer, brain metastasis, ErbB inhibitors, Poziotinib, Lapatinib, neuregulin-1, 41 brain tumor remission. 42 43 44 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 3

45 Brain metastasis is a fatal complication occurring in 50% of patients with HER2+ breast cancer (BC) and 46 represents one of the most adverse scenarios of HER2+ BC progression (1, 2) as successful treatments 47 are lacking (3). Targeted therapies approved for application in primary and metastatic HER2+ BC involve 48 both monoclonal antibodies (Trastuzumab, Pertuzumab) and small molecule RTK inhibitors (RTKi) such 49 as Lapatinib, Neratinib, and Tucatinib (3, 4). RTKis are generally recommended as a second or third-line 50 regimen for advanced BC patients who were unresponsive or developed resistance to the anti-ErbB mAbs 51 (4). Despite a higher blood-brain barrier penetrability, RTKi monotherapies lack clinical efficacy (4, 5). 52 When used as a monotherapy, the reversible dual EGFR/ErbB2 RTKi Lapatinib showed only a marginal 53 response rate of 2.6-6% in patients with brain metastasis (BM) of HER2+ BC (6, 7). Monotherapy with 54 the covalent dual EGFR/ErbB2 RTKi Neratinib showed a similar (8%) poor efficacy (8), although both 55 Lapatinib and Neratinib proved to be more effective when combined with the deoxycytidine derivative 56 Capecitabine (8, 9) in patients with brain metastases of HER2+ BC. 57 ErbB1-4 engage in ligand- activated homo- and hetero-dimerization with 11 different EGF -like ligands 58 (10). ErbB2 (HER2) overexpression drives malignant transformation in BC as the preferred dimerization 59 partner (11) for other RTKs of the ErbB family to activate downstream signaling via several pathways 60 (e.g., PI3K/Akt, MAPK/ERK, PLCγ) known to mediate cell survival, proliferation, epithelial -61 mesenchymal transition, cell migration and tissue invasion (12). ErbB2-ErbB3 heterodimerization 62 provides the most significant signaling in HER2+ BC (13, 14). 63 Neuregulins (NRG) constitute the largest subgroup of structurally related ErbB ligands (15-17). NRG1 64 specifically bind s to the extracellular domain of ErbB3 and ErbB4 which results in formation of 65 potentially six active ErbB dimers (12, 18). The role of NRGs in cancer progression is tightly linked to 66 the ErbB-driven signaling in breast cancer (19-21). NRGs are commonly expressed in the central nervous 67 system (22). NRG1 protein is expressed mainly by neurons, but also astrocytes, oligodendrocytes and 68 microglia which are main resident cell types of the brain (https://www.proteinatlas.org/) (23) . The 69 expression of ADAM sheddases by brain resident cell populations may contribute to the release and 70 paracrine HER2 activation by NRGs resulting in the progression of brain metastatic HER2+ BC at the 71 brain metastatic niche (24, 25). Studies on breast-to-brain metastasis (BCBM) are hampered by the few 72 patient-derived HER2+ BC cell models capable of generating brain metastases in mice (26), which 73 severely limits mechanistic and therapeutic studies on BCBM. 74 The small molecule RTKi Lapatinib induces apoptosis in several HER2+ BC cell models (27) but several 75 mechanisms of resistance to RTKis have been identified (28), among them the activation of 76 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 4 compensatory pathways involving transcriptional and posttranslational up-regulation of ErbB3 (29), re-77 activation of Akt (30) , and PI3K independent induction of mTOR activity (31). For brain metastatic 78 lesions, the expression of NRG1 in the brain (22, 32) can potentially mediate any of these mechanisms 79 by binding to ErbB3/4 and inducing ErbB activation. ErbB3-PI3K-Akt signaling is known to mediat e 80 the anti-apoptotic response through the regulation of Bcl -2 and IAP protein families (33 -35). Although 81 endogenous NRG1 was shown to mediate resistance of HER2+ BC to both Lapatinib (27) and 82 Trastuzumab (36) through the activation of ErbB3 and concomitant downregulation of apoptosis , a 83 mechanistic explanation of the anti -apoptotic response driven by NRG1 is still lacking. Here we 84 introduce a novel patient -derived hematogenic HER2+ BC brain metastasis model (BCBM94) and 85 identify NRG1-driven mechanisms that rescue L apatinib-induced apoptosis. We show that NRG1 fails 86 to rescue apoptosis by the brain penetrable and irreversible ErbB1/2/4 RTKi Poziotinib and demonstrate 87 the ability of Poziotinib to abrogate with high efficacy HER2+ BC metastatic brain lesions in mice. 88 89

90 Establishment and characterization of a new BCBM94 cell and mouse model 91 A neurosurgical tissue sample was obtained from a metastatic adenocarcinoma to the cerebellum of a 92 female patient diagnosed with invasive ductal carcinoma (grade III, stage T2N1). This research was 93 approved by the Health Research Ethics Board (HREB; protocol #19-038), University of Manitoba. The 94 original patient breast tumor was identified as luminal-B HER2+ BC expressing HER2+ (score 3+), ER+ 95 (Allred score 7), and devoid of PRα/β . Ultrasound-guided xenografting of BCBM94 cells into the left 96 ventricle of Rag2γc-/- mice robustly led to brain metastasis in mice about 3 months after xenografting. 97 The ability of BCBM94 cells to establish hematogenic brain metastases was confirmed in three 98 consecutive rounds of isolation of tumor cells from mouse brain and intracardial re -injection for brain 99 colonization. BC metastatic lesions were of epithelial morphology and formed vascularized and 100 proliferative metastases of different sizes throughout the brain as determined by immunoreactive CD31 101 and Ki67, respectively (Fig. 1A). Strong membrane expression of ErbB2 was detected in the patient 102 primary BC tissue and corresponding brain metastatic tissue, as well as in brain metastases of the 103 experimental animals (Fig. 1B). Membrane expression of ErbB2 was also detected in cultured BCBM94 104 cells (Fig. 1C). When compared to cultured triple-negative MDA-MB-BR or HER2+ cell lines BT474 105 and SKBR3, BCBM94 showed the highest expression of ErbB2 protein, low level of ERα, and complete 106 lack of PRα/β proteins (Fig. 1D). These findings identified BCBM94 cells as novel luminal -B HER2+ 107 BC subtype capable of hematogenous brain metastasis (Fig. 1D). 108 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 5 In-situ hybridization of BCBM94 metastatic mouse brain revealed abundant mouse Nrg1 mRNA 109 expression in different regions, including mouse brain tissue adjacent to BCBM94 brain metastatic 110 lesions, which were highly positive for human ErbB2 transcripts (Fig. 1E). Western blot analysis showed 111 the absence of endogenous NRG1 protein expression in protein lysates of BCBM94 and two other 112 HER2+ BC cell lines (BT474, SKBR3), whereas triple -negative MDA-MB231BR cells had detectable 113 level of NRG1 protein (Fig. 1F). Quantitative RT-PCR detected transcripts for ErbB1 -4 but weak to 114 undetectable levels of NRG1 -4 transcripts in BCBM94 cells (Sup. Fig. 1A) . These results suggested 115 ErbB3/4+ BCBM94 as a target of Nrg1 produced by the mouse brain. 116 NRG1 counteracts Lapatinib cytotoxicity in HER2+ brain metastatic BC cells 117 We evaluated the effects of the small molecule EGFR/ErbB2 RTKi Lapatinib and recombinant human 118 NRG1 (rhNRG1) on the viability of HER2+ brain metastatic BC cell s BCBM94 and BT474. WST-1 119 cell viability assay s showed a dose -dependent decline in BCBM94 cell numbers upon treatment with 120 Lapatinib. The half maximal inhibitory concentration (IC50) of Lapatinib in WST-1 assays for BCBM94 121 cells was reached at 250nM (Fig. 2A). At this IC50 Lapatinib concentration, increasing concentrations of 122 rhNRG1 resulted in a dose-dependent rescue of cell viability. This rescue effect was maximal at 5ng/mL 123 rhNRG1 (Fig. 2B) , which is an NRG1 concentration reported in human serum (37). From hereon, 124 5ng/mL rhNRG1 was used for all in vitro experiments. When combined with Lapatinib at IC50, rhNRG1 125 maintained the viability of BCBM94 cells at 80-90% of control values as determined at 24-72h by WST 126 assays (Fig. 2C). 127 Measuring cellular impedance as a function of cell number and proliferation rate, xCELLigence Real -128 Time Cell Analysis assays confirmed the ability of rhNRG1 to rescue BCBM94 from Lapatinib-induced 129 cytotoxicity (Sup. Fig . 1B, C). These findings were further validated in BT474 cell s where the 130 cytoprotective action of rhNRG1 was even more pronounced (Sup. Fig. 1D). 131 To evaluate the long-term effects of Lapatinib +/-rhNRG1 treatment on the proliferation of BCBM94, 132 we performed colony formation assays. While treatment with 250 nM Lapatinib for 14 days reduced the 133 number of cell colonies by 90% from colony numbers in the control, combined L apatinib/ rhNRG1 134 treatment resulted in colonies numbers similar to the control (Fig. 2D). NRG1 alone did not significantly 135 alter colony numbers (Fig. 2D) . Western blot analysis revealed that the rescue effect was mediated 136 exclusively by exogenous rhNRG1 as Lapatinib did not induce the upregulation of endogenous NRG1 137 transcripts or protein in BCBM94 cells (Sup. Fig. 1E). 138 139 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 6 NRG1 abolishes Lapatinib cytotoxicity by attenuating apoptosis 140 Next, we asked whether the observed viability changes involved regulation of apoptosis. BCBM94 and 141 BT474 HER2+ BC cell models responded to Lapatinib with an increase in cleavage of the active catalytic 142 domain of PARP , but cleaved PARP levels remained undetectable with combined L apatinib/ rhNRG1 143 treatment (Fig. 2E, F). Notably, the sequential addition of rhNRG1 secondary to exposure to L apatinib 144 for 24h was able to mitigate a Lapatinib-induced cleaved PARP induction (Sup. Fig. 2A, B). To explore 145 the events preceding the cleavage of PARP , we evaluated the activity of the effector caspase-3 and 146 caspase-7. Caspase -Glo 3/7 assays showed a prominent activation of caspase -3/7 in BCBM94 cells 147 exposed to Lapatinib. The observed induction of these pro-apoptotic caspases of the intrinsic apoptosis 148 pathway was fully abrogated by rhNRG1 at 5 ng/ml (Fig. 2G). Further investigation of the upstream 149 steps of the apoptotic cascade revealed that Lapatinib mediated activation of the apoptosis- initiating 150 caspase-9 in BCBM94 and BT474 cells which was completely abolished by rhNRG1 (Fig. 2H, I). These 151

identified mitochondrial apoptotic pathways as a target of the anti-apoptotic action of rhNRG1. 152 Anti-apoptotic action of NRG1 is mediated through the mitochondrial pathway 153 Members of the Bcl -2 family mainly regulate the intrinsic apoptotic pathway (38) . In BCBM94 and 154 BT474 cells, rhNRG1 rescued the Lapatinib-induced de- phosphorylation of the BH3- only protein Bad 155 at Ser136, which is an Akt phosphorylation site (Fig. 3A, B) . The ability of the BH3- only and pore -156 forming proteins of the Bcl -2 family to trigger apoptosis is determined not solely by their expression 157 levels but also by cellular localization, phosphorylation, and dimerization status of the se proteins (39). 158 While we did not observe significant changes in the total levels of the Bcl -2 proteins (Suppl. Fig. 2C-159 F), cellular immunofluorescence analysis identified the appearance of Bax aggregates that co-localized 160 with mitochondrial Bak in BCBM94 cells exposed to Lapatinib (Fig. 3D). RhNRG1 preserved the diffuse 161 cellular distribution of B ax and blocked formation of pro -apoptotic Bax/ Bak dimers observed in the 162 presence of Lapatinib only (Fig. 3D). MitoTracker assays showed a reduction of fluorescence signal 163 intensity in Lapatinib-treated BCBM94 indicative of a lower number of active mitochondria (Fig. 3C) 164 and rhNRG1 was able to mitigate this Lapatinib effect on mitochondria (Fig. 3C) . TEM ultrastructural 165 imaging revealed prominent damage to mitochondrial cristae and ruptured outer mitochondrial 166 membranes in BCBM94 cells treated with Lapatinib (Suppl. Fig. 2G), suggestive of mitochondrial 167 damage. Like in control cells, these morphological alterations were only rarely observed in cells 168 receiving dual Lapatinib/ rhNRG1 treatment (Suppl. Fig. 2G) . We concluded that rh NRG1 rescued 169 apoptosis by blocking the pro-apoptotic cascade at the outer mitochondrial membrane. 170 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 7 NRG1 utilizes an ErbB3-Akt pathway to promote Lapatinib resistance 171 In both BCBM94 and BT474 cell models, Lapatinib potently decreased phosphorylation of ErbB1 to 4 172 and reduced total ErbB1 and ErbB4 protein content (Fig. 4A -D). rhNRG1 exclusively rescued ErbB3 173 phosphorylation in both cell lines (Fig. 4A-D). Intriguingly, the patient tissues derived from the primary 174 breast tumor and the brain metastatic tissues used to isolate BCBM94 cells contained immunoreactive 175 phosphorylated ErbB3, as did BCBM94 metastatic lesions in mouse brain (Fig. 4E). The presence of 176 constitutively activated ErbB3 in situ and the ability of rhNRG1 to rescue ErbB3 phosphorylation upon 177 Lapatinib treatment suggested a key role of ErbB3 in NRG1- mediated survival of both HER2+ BC 178 models. While s elective siRNA-mediated ErbB3 knockdown ( KD) alone did not induce apoptosis, 179 treatment of BCBM94ErbB3-KD and BT474ErbB3-KD cells with Lapatinib resulted in a significantly higher 180 level of PAPR cleavage compared to mock transfected cells treated with Lapatinib (Fig. 5A, B). ErbB3 181 KD also attenuated the ability of rhNRG1 to counteract Lapatinib-induced PARP cleavage in BCBM94 182 and BT474 cells (Fig. 5 A, B). ErbB3 KD also markedly weakened the ability of rhNRG1 to rescue 183 BADSer136 phosphorylation under Lapatinib. Significantly reduced cellular levels of phospho-BADSer136 184 were detected upon co-treatment with Lapatinib and rhNRG1 compared to mock transfected BCBM94/ 185 BT474 cells (Fig. 5C, D). Lapatinib further decreased BADSer136 phosphorylation in both ErbB3-KD cell 186 models. While BCBM94ErbB3-KD and BT474ErbB3-KD cells showed a small increase in phospho-BADSer136 187 upon NRG1, overall phospho-BAD levels were negligible compared to mock silenced BCBM94 and 188 BT474 cells co-treated with Lapatinib and rhNRG1 (Fig. 5C, D). Phosphorylation of BAD at Ser136 is 189 mediated by activated AKT (33) and the rhNRG1-mediated rescue of ErbB3 phosphorylation under 190 Lapatinib coincided with increased phosphorylation of the ErbB downstream effector kinase AKT in 191 BCBM94 and BT474 cells (Fig. 5E, F). To demonstrate a causal involvement of A KT in the anti -192 apoptotic regulation by NRG1, we utilized the PI3K/AKT inhibitor PI -103. PI-103 abrogated Akt 193 phosphorylation with a moderate downregulation of total AKT protein in untreated BCBM94 and BT474 194 cells and completely blocked the pAKT rescue by rhNRG1 in Lapatinib exposed cells (Fig. 5G, H). PI-195 103 alone increased PARP cleavage in BCBM94 and BT474 cells and this coincided with a complete 196 loss of BADSer136 phosphorylation and an inability of NRG1 to rescue cells from apoptotic actions upon 197 co-treatment with Lapatinib/ rhNRG1 (Fig. 5G, H). Hence, in our HER2+ BCBM94 and BT474 models 198 NRG1 utilizes an ErbB3-PI3K-AKT-BAD signaling cascade to cause resistance to Lapatinib (Fig.6). 199 NRG1 fails to rescue Po ziotinib cytotoxicity in HER2+ BC in-vitro 200 In search of small molecule ErbB inhibitor s capable of overcoming NRG1 RTKi resistance, we 201 assembled a collection of 50 ErbB inhibitors, covering a range of isoform selectivities, mode of action 202 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 8 (reversible vs covalent) , and predicted CNS penetrance ( Suppl. Table 1 ). We then tested the entire 203 collection of compounds at 22-concentrations, ranging from 0.003 pM to 40 uM in the BCBM94 and 204 HME1 immortalized human normal mammary epithelial cells using a CellTiterGlo viability assay. The 205 dose-response profiles were compared to identify drug candidates that showed high selectivity and 206 efficacy towards BCBM94, but not HME1 cells and were insensitive to the anti -apoptotic rescue by 207 rhNRG1. This screen included the FDA-approved Lapatinib, Neratinib, and Tucatinib currently used as 208 second line treatments in brain metastatic BC patients (40). We identified the irreversible ErbB1/2/4 209 inhibitor Poziotinib which was two orders of magnitude more cytotoxic in malignant BCBM94, with a 210 single digit nanomolar IC50 compared to non-malignant HME1 cells (Fig. 7A). Poziotinib half-maximal 211 activity concentration (AC50) did not significantly differ between BCBM94 and BT474, indicating 212 efficacy against both HER2+ BC models (Fig. 7B) . Furthermore, rhNRG1 failed to rescue HER2+ BC 213 models from the cytotoxic effects of Poziotinib (Fig. 7C). By contrast, the reversible RTKi Lapatinib 214 and Tucatinib or the irreversible RTKi Neratinib only had moderate efficacy and rhNRG1 caused a shift 215 in AC50, indicating that NRG1 successfully rescued BCBM94 from the cytotoxic ErbBi activity of these 216 clinically used drugs (Fig. 7C) . Poziotinib emerged as a promising candidate for overcoming the anti -217 apoptotic action of NRG1. Poziotinib (2.5- 5nM) showed high cytotoxicity in cultured BCBM94 and 218 BT474 cells (Fig. 7D) and rhNRG1 was unable to rescue these HER2+ BC cells as determined by WST 219 (Fig. 7E, F) and cell impedance assays (Sup. Fig. 3A, B). Poziotinib caused apoptosis with significant 220 PARP cleavage (Fig. 7 G, H) and, despite rhNRG1 present, abolished phosphorylation of ErbB3 Y1289 221 (Fig. 7I, J) and Akt S473 (Fig. 7K, L) . We concluded that small molecule ErbB inhibitor Poziotinib 222 blocked NRG1-ErbB3-PI3K-Akt signaling to cause apoptosis in HER2+ BC (Fig.8). 223 Poziotinib eliminates HER2+ BC metastasis in-vivo 224 Computational predictions indicated high brain penetrability of P oziotinib which was confirmed in our 225 pharmacokinetic studies in C57BL mice that demonstrated excellent brain permeability of P oziotinib 226 when applied via two different routes. Our tolerability tests identified subcutaneous administration of 227 Poziotinib to be far superior to oral application for subsequent drug treatments of BC brain metastases in 228 C57BL mice (data not shown). Poziotinib administered at 5 mg/kg PO (per os) and SC (subcutaneous), 229 and 2 mg/kg PO reached therapeutic concentrations in the brain , with concentrations of this RTKi still 230 detectable at IC50 of 2.5 nM at 20h and 8h upon administration, respectively (Suppl. Fig. 3C). 231 To determine the in-vivo efficacy of P oziotinib towards HER2+ BC brain tumors , we orthotopically 232 xenografted BCBM94 and BT474 cells into the right striatum of immunocompromised Rag2γc-/- mice. 233 HER2+ BC b rain tumors were confirmed by magnetic resonance imaging ( MRI) prior to treating the 234 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 9 animals with 100μl of either Lapatinib (80mg/kg, PO), Poziotinib (4mg/kg, SC), or solvent control (PO 235 and SO) for two cycles of 5-days ON and 2 days OFF. MRI volumetry of pre- and post-treatment scans 236 demonstrated a highly significant reduction in BCBM94 tumor volumes with Poziotinib, but not 237 Lapatinib (Fig. 9A). The high efficacy of Poziotinib against HER2+ BC brain tumors was also confirmed 238 in mice xenografted with BT474 cells (Fig. 9B). Ultra-performance liquid chromatography-tandem mass 239 spectrometry (UPLC-MS/MS) analysis of plasma and brain samples collected on the last day of treatment 240 (1h post-dosing) demonstrated that Poziotinib and Lapatinib reached similar concentrations in the brain 241 at the doses administered. Notably, P oziotinib more effectively crossed the blood- brain barrier, with a 242 two-fold higher brain/plasma concentration ratio than Lapatinib (Fig. 9C). Post-treatment FFPE brain 243 sections of mice xenografted with BCBM94 confirmed the observed MRI changes. BCBM94 tumors 244 were exclusively identified in H&E stained tissues of the L apatinib and solvent control groups. These 245 tumors showed phosphorylated ErbB3Y1289 and Ki67+ nuclei and were largely negative in TUNEL tests 246 detecting damaged DNA (Fig. 9D). In sharp contrast, the tumor sites of all mice treated with Poziotinib 247 were devoid of BCBM94 tumor cells, ErbB3 phosphorylated cells , and Ki67+ nuclei, but demonstrated 248 positive TUNEL staining (Fig. 9D) . Similar results were obtained for mice orthotopically xenografted 249 with BT474 cells and treated with Poziotinib or Lapatinib (data not shown). 250 251

252 Despite successful therapies of peripheral BC disease and longer survival, patients with HER2+ BC have 253 an increased risk of brain metastasis from HER2+ tumors (41, 42) . Treatment with humanized 254 monoclonal antibodies against HER2 and/or RTK small molecule inhibitors are not successful for 255 HER2+ brain metastatic disease, in part because the se compounds do not reach therapeutic 256 concentrations in the brain (43 -45). There is an urgent need for experimental in -vivo models to study 257 HER2+ breast cancer brain metastasis because additional environmental factors in the brain metastatic 258 niche can determine treatment responses. Among approximately 30 HER2+ BC cell lines that have been 259 established so far, there are only a few xenogeneic (e.g. BT474) models that can cross the blood- brain-260 barrier (BBB) and repeatedly establish hematogenous brain metastasis (46, 47) . Although BT474 and 261 our new BCBM94 model described here can be classified as a Luminal -B HER2+ molecular subtype, 262 their receptor profile differs substantially. While BCBM94 is ErbB1 -4 -positive, ER -low, and PR -263 negative, BT474 is ErbB1-4-positive, ER-high, and PR-positive. Another unique feature of BCBM94 is 264 that, unlike BT474 which was isolated from primary breast cancer (46), BCBM94 is derived directly 265 from a patient’s breast cancer brain metastasis. The molecular differences in both brain metastatic 266 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 10 HER2+ BC models reflect a degree of molecular heterogeneity in BC brain metastasis. The ability of our 267 BCBM94 model to cross the BBB and establish hematogenous brain metastasis combined with a unique 268 receptor profile make the BCBM94 model a valuable new tool for studying multiple aspects of brain 269 metastasis in HER2+ BC. Since BCBM94 require approximately 3.5 months to establish sizeable brain 270 metastasis for hematogenic and orthotopic xenografting alike, our model reflects the long latency 271 observed for human brain metastasis in HER2+ BC patients. 272 Upregulation of endogenous expression of NRG1 in several HER2+ BC cell lines (BT474, SKBR3) was 273 previously shown to promote autocrine activation of the ErbB3-EGFR signaling axis and mediate 274 resistance to Lapatinib (27). Our current data place exogenous NRG1 at the top of a powerful NRG1-275 ErbB3-PI3K-AKT-BAD signaling axis that mediates anti- apoptotic resistance against Lapatinib in the 276 HER2+ BC metastatic brain niche. NRGs are abundantly expressed by resident cells of the normal brain 277 (48, 49) , as are ADAM metalloproteinases responsible for shedding of the extracellular EGF -like 278 domains of membrane-bound NRG precursors (16, 17, 24, 25) . We localized multiple NRG1+ cells in 279 the TME of BCBM94 BC brain metastases as source for TME-derived NRG1 to induce therapeutic 280 resistance to Lapatinib through paracrine and/or juxtacrine NRG1-ErbB3 signaling in brain metastases 281 of HER2+ BC. Indeed, our in-vitro data revealed that exogenous NRG1 can counteract the cytotoxic and 282 pro-apoptotic actions of Lapatinib in BCBM94 and BT474 cells. 283 The intrinsic apoptosis pathway (50) is primarily regulated by the Bcl-2 family of proteins consisting of 284 the anti-apoptotic, pro-apoptotic BH3-only (apoptosis initiating), and pore-forming (executive) proteins 285 (39) which initiate apoptosis through disruption of the outer mitochondrial membrane and the release of 286 cytochrome C . In both HER2+ BCBM cell models , the apoptotic action of Lapatinib included a 287 significantly reduced phosphorylation of ErbB1 -4 and downstream targets A KT and B AD. 288 Dephosphorylated B AD c an engage in dimer formation with anti- apoptotic B CL-2 family members 289 BCL-XL, B CL-2, and M CL-1, which prevents B AX/ B AK oligomeric pore formation in the outer 290 mitochondrial membrane and the initiation of intrinsic caspase -mediated apoptosis (51). We confirmed 291 BAX/ BAK oligomers by co-immunofluorescence and showed the induction of caspase 9 cleavage and 292 activation of the effector caspases-3/7 in Lapatinib-treated BCBM94 and BT474 cells. 293 In agreement with clinical data, Lapatinib was ineffective in reducing the growth of HER2+ BC tumors 294 in mouse brain, despite reaching concentrations that reduced viability in- vitro. Our results suggest that 295 the presence of brain -derived NRG1, which could compete with Lapatinib for binding to ErbB3/4 296 expressed on both BCBM94 and BT474 BC cells , may account for endocrine therapeutic resistance 297 observed in our mouse studies and in the clinic. To be considered a promising candidate for the treatment 298 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 11 of HER2+ BC brain metastases, any of the 50 small molecule ErbB inhibitor compounds tested in our 299 high-throughput screens had to meet stringent selection criteria. This included good brain permeability, 300 high efficacy at low nanomolar concentrations, selective toxicity to HER2+ BC cells but not to non -301 tumor breast epithelial cell line HME -1, and the ability to inhibit the growth of our HER2+ BC cell 302 models in the presence of exogenous NRG1. Of several potential candidates, the irreversible ErbB 303 inhibitor Poziotinib had the lowest IC50 (low nanomolar) and showed similar efficacy in both BCBM94 304 and BT474 models but had minimal toxicity to HME -1. A powerful inducer of apoptosis, i rreversible 305 RTKi Poziotinib blocked rhNRG1-mediated rescue of the ErbB3 -AKT-Bad signaling cascade in both 306 tested HER2+ brain metastatic BC models. BCBM94 and BT474 xenografts demonstrated a dramatic 307 reduction in tumor volume after only two weeks of treatment with P oziotinib, whereas xenografts 308 continued to grow under Lapatinib. In agreement with a significant tumor volume reduction on MRI 309 scans, histological examination demonstrated the absence of viable tumor cells, the loss of ErbB3 310 phosphorylation, and marked DNA fragmentation at brain injection sites upon Poziotinib treatment. By 311 contrast, BCBM94 lesions treated with Lapatinib remained proliferative and had preserved ErbB3 312 phosphorylation. Poziotinib, alone or in combination with F ulvestrant, was recently shown to attenuate 313 tumor growth, multiorgan metastasis, and mTOR activation in recurrent metastasizing BC cells harboring 314 an HER2 L755S mutation that conferred resistance to the irreversible ErbB inhibitor Neratinib. Recent 315 clinical trial phase I and II studies used orally administered Poziotinib in patients with metasta sising 316 breast cancer (NOV120101-203 trial) (52) and non-small cell lung cancer with HER2 exon 20 insertions 317 (ZENITH20-2 Trial) (53, 54). Although Poziotinib showed meaningful clinical activity in these heavily 318 pretreated HER2+ metastatic BC and non -small cell lung cancer patients, including a small group of 319 patients with brain metastases, the toxic side effects, mainly rash, diarrhea, and stomatitis, frequently led 320 to dose reductions. In our animal studies, we noticed that the same concentration of Poziotinib proven 321 effective at treating BCBM94 and BT474 xenografts showed significantly higher toxicity upon oral than 322 subcutaneous administration in mice. Our pharmacokinetic studies confirmed that either administration 323 route yielded similar brain concentrations, suggesting a possible remedy to the clinical toxicities 324 observed with Poziotinib in patients. Recently, the irreversible ErbB1/2 kinase inhibitor Pyrotinib 325 showed beneficial effects in patients with advanced metastatic HER2+ BC who had progressed under 326 Trastuzumab (55). For patients with radiotherapy- naïve and radiotherapy resistant HER2+ brain 327 metastases, Pyrotinib in combination with capecitabine was reported to have a response rate of 74% and 328 42%, respectively (55) . However, our small molecule compound screens revealed that the AC 50 for 329 Pyrotinib required 10 times higher concentrations than P oziotinib. Importantly and in contrast to 330 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 12 Poziotinib, the efficacy of Pyrotinib was reduced 10 fold in the presence of NRG1 in both of our HER2+ 331 BCBM cell models (Suppl. Table 1). 332 In summary, we have established and characterized a novel patient-derived brain metastatic HER2+ BC 333 cell line capable of hematogenic colonization of mouse brain as pre-clinical model to study HER2+ BC 334 brain metastasis. We unveiled the antiapoptotic mechanism of exogenous NRG1 as a driver of resistance 335 to the reversible RTKi Lapatinib in HER2+ BC brain metastases. Poziotinib prevented the anti-apoptotic 336 actions of NRG1 in-vitro and revealed high efficacy in abrogating HER2+ BC brain metastatic lesions 337 in mice. When administered by SC route, this irreversible RTKi is a promising drug for the management 338 of brain metastases of HER2+ BC. 339 340

341 We confirmed the ability of BCBM94 to establish hematogenous brain metastasis after intracardial 342 xenografting. However, we perform ed orthotopic intracranial, and not intracardial , xenografting of 343 HER2+ BC cell in mice when testing the effects of Lapatinib and Poziotinib in-vivo in order to measure 344 the size of lesion s at identical brain locations during treatments . Future studies will use intracardial 345 xenografting to determine if hematogenous HER2+ BC brain metastases undergo similar regression with 346 Poziotinib. 347 348

349 Cell Culture 350 Human HER2+ BT474 (HTB-30) and SKBR3 (HTB-20) BC cell lines were acquired from ATCC. The 351 brain-seeking subline of the triple-negative MDA-MB-231 cells, MDA-MB-231/BR, was a generous gift 352 of Dr. Patricia Steeg (NCI, Bethesda, Maryland). A patient-derived HER2+ BCBM94 BC cell line was 353 established in the laboratory from BC cells isolated from a surgically resected HER2+ BC brain 354 metastasis. The first mouse brain passage of the BCBM94 model was used in the study. All cell lines 355 were routinely cultured in DMEM/F -12 Ham’s medium supplemented with 10% FBS. Except for data 356 presented in Fig. 6A-C in which 10 % FBS was supplemented during the assay, all treatments were done 357 in DMEM/F-12 Ham’s 1% FBS. hTERT-HME1 (CRL-4010) were acquired from ATCC and cultured in 358 Mammary Epithelial Cell Growth Medium Bullet Kit from Lonza (Bend, OR, Catalog #: CC-3150). 359 siRNA gene silencing 360 ErbB3 silencing was achieved with 20nM of ErbB3 siRNA purchased from Ambion (Cat. AM16708, 361 ID146247). The non- coding siRNA was acquired from Dharmacon (Horizon, St. Louis, MO , Cat. D-362 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 13 001810-10-20). siLentFect lipid reagent (BioRad, Mississauga, ON, Cat. 1703360) was utilized for 363 transfection. 364 Drugs and growth factors 365 Reversible EGFR/ErbB2 RTK inhibitor Lapatinib (Cat. S1028), irreversible pan-ErbB RTK inhibitor 366 Poziotinib (Cat. S7358), and PI3K/Akt/mTOR inhibitor PI -103 (Cat. S1038) were purchased from 367 Selleckchem (Houston, TX). Recombinant human NRG1 was acquired from BioLegend (San Diego, CA, 368 Cat. 551904). High-throughput drug screening assays were used with compounds sourced as indicated 369 in Suppl. Table 1. 370 Blood brain barrier penetrability predictions 371 CNS penetrance was predicted by taking a consensus of the predictions derived from the following 372

1. The Blood–Brain Barrier (BBB) Score (56), using a cutoff >=3.5 to designate a compound 373 as CNS penetrant; 2. B3clf Predictors for Blood- Brain Barrier Permeability with resampling strategies 374 based on B3DB database (57) ; 3. Using the BBB permeability classification model (BBB -Filter) from 375 ADMET Predictor software (58). 376 Cell based assays 377 Cell proliferation reagent WST1 was acquired from Sigma Millipore (Cat. 5015944001) and used as per 378 the manufacturer’s protocol. BCBM94 and BT474 were seeded at 4000 cells/well in 100uL volume in a 379 96-well plate. After 24h the culture medium was replaced with 100uL/well media containing treatment 380 compounds. Caspase-Glo 3/7 Assay System (Promega, Cat. G8091) was used to measure the activity of 381 caspase-3 and caspase-7. Cell seeding and incubation conditions were identical to that described for the 382 WST1 assay. For Colony formation assays (59), BCBM94 were seeded at 20,000 cells/well in 6- well 383 plates. The culture media was replaced every 3 days. After 14 days the cells were fixed and imaged with 384 a D2 inverted microscope (Zeiss, Jena, Germany) . For the high- throughput drug screening assay, 385 compounds were added to 1536- well white-walled plates (Greiner, Monroe, NC, Cat. 789173) using 386 acoustic dispensing. Cells were then seeded at 500 cells/well in 5uL volume. After 72h of incubation 387 (48h for HME1 assay), 2.5uL of CellTiterGlo reagent (Promega, Madison, WI, Cat. G7570) was added 388 to each well and plates were incubated at room temperature for 10 mins prior to luminescence readout. 389 For Mitotracker assays, BCBM94 cells were seeded at 80,000 cells/well on glass coverslips in 6- well 390 plates. At the end of treatments, MitoTracker Red CMXRos (Cell Signaling Technology, Danvers, MA, 391 Cat. 9082) was added to a final concentration of 100 nM/ well and plates were incubated for 30 min at 392 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 14 37°C in the dark. Cells were fixed with ice -cold methanol and t he mean fluorescence intensity was 393 quantified per 100,000 µm² and 200 nuclei per treatment group using Zen software (Zeiss). 394 Quantitative real-time polymerase chain reaction (qRT-RCR) 395 Total RNA was isolated with TRIZol and cDNA was synthesized with qScript cDNA SuperMix 396 (QuantaBio, Beverly, MA, Cat. 95048-025). Primer sequences are listed below and the SYBR Green 397 PCR Master Mix (Applied Biosystems, Cat. 4300155) was used. mRNA expression was analyzed by 398 QuantStudio 3 qRT -PCR System (Applied Biosystems, Fisher Scientific, Winnipeg, MB ), and the 399 Comparative CT (ΔΔCt) Method (60). Primer sequences were: F_huNRG1, 5’-400 ATTGAAAGAGATGAAAAGCCAGG-3’; R _huNRG1, 3’- GCCAGTGATGCTTTGTTAATGC-5’; 401 F_huNRG2, 5’- CTAAGCAAAAAGCCGAGGAGC-3’; R_huNRG2, 3’-402 CTTCCGCTGTTTTTTGGTCTTG-5’; F_huNRG3, 5’- AACACTTATCATTGGAGCCTTC-3’; 403 R_huNRG3, 3’- GGTGTTTCATTTTCTGCCTTTG-5’; F_huNRG4, 5’-404 CTCTGGGTATTGTGTTGGCTG-3’; R_huNRG4, 3’- TGTCCTCCTGCACCAAAAACC-5’. 405 Gel electrophoresis of proteins and Western blotting 406 The whole -cell lysates prepared with Laemmli cell lysis buffer were separated on TGX FastCast 407 acrylamide gels (BioRad, Mississauga, ON). Nitrocellulose membranes and the Trans-Blot Turbo system 408 (Bio-Rad) were used for protein transfer. The non-specific antibody binding sites were blocked with 5% 409 non-fat milk in TBST, pH 7.6 for 1h at RT. A ntibodies and incubation conditions are listed in Table 1. 410 Proteins were visualized using Clarity (Max) Western ECL Substrates (Bio-Rad, Cat. 1705060, Cat. 411 1705062) and ChemiDoc MP Imaging System (Bio -Rad). Band volume quantification (densitometry) 412 was done in the Image Lab (Bio-Rad). 413 In situ hybridization 414 Formalin-fixed, paraffin-embedded (FFPE) mouse brain tissue slides were processed according to the 415 manufacturer’s instructions (61). Hybridization probes against human ErbB2 (Cat. 418741) and mouse 416 Nrg1 (Cat. 441811) were acquired from ACDBio (Newark, CA). Hybridization, amplification, and signal 417 detection were performed according to the manufacturer’s protocol (62). Signal amplification and 418 detection were performed with RNAscope 2.5 HD Duplex Reagent Kit (ACDBio, Cat. 322430). 419 Immunodetection in brain tissues and cells 420 Mouse FFPE brain tissue sections were deparaffinized and re -hydrated. The list of antibodies, antigen 421 retrieval, and staining conditions are listed in Table 1. Non-specific antibody binding sites were blocked 422 with 10% normal goat serum in TBST pH 7.6 for 1h. The chromogenic signal was amplified using HRP-423 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 15 Streptavidin (Jackson ImmunoResearch, West Grove, PA , Cat. 016-030-082) and detected with DAB 424 Substrate Kit (Fisher Scientific, Winnipeg, MB, Cat. PI34002) . BCBM94 and BT474 were seeded at 425 80,000 cells/well on glass coverslips in 6- well plates. Cells were fixed with 4% methanol- free 426 formaldehyde in PBS for 15 min at room temperature (RT) and permeabilized in 0.1% Triton X -100 in 427 PBS for 15 min at RT. Non-specific antibody binding sites were blocked with 5% NGS in PBS for 1h. 428 For antibodies and staining conditions see Table 1. 429 TUNEL assay 430 In Situ Cell Death Detection Kit, POD (Roche, Mississauga, ON , Cat. 11684817910) was used as per 431 manufacturer’s protocol to detect apoptosis in FFPE mouse brain tissue sections. 432 Transmission electron microscopy (TEM) 433 BCBM94 cells were trypsinized, pelleted, and fixed in 3% Glutaraldehyde in 0.1M Sorensen’s buffer 434 and post-fixed in 1% Osmium tetroxide (OsO4) in 0.1M Sorensen’s buffer. The pellets were embedded 435 in EMBed 812 resin (Electron Microscopy Sciences, Hatfield, PA , Cat. 14900). Thin sections were 436 stained with Uranyless (EMS, Cat. 2240920) and Lead citrate (EMS, Cat. 22410) and imaged using 437 Philips TM10 transmission electron microscope. At least 10 cells per sample group were imaged and 438 qualitatively analyzed for signs of mitochondrial damage. 439 In-vivo experiments 440 Ultrasound-guided intracardial injection of BCBM94 cells into RAG2γc -/- mice (50) resulted in brain 441 metastatic tumors after around 3.5 months. The animal research was approved by the Bannatyne Campus 442 Animal Care Committee (protocol #21-017). We opted for orthotopic xenografting into the right striatum 443 using stereotactic surgery to achieve the same lesion size at the identical brain location for BCBM94 and 444 BT474 tumor growth in mouse brain (63). Intracranial tumor growth was monitored using MRI. Upon 445 detection of sizeable brain metastases, the animals were treated with either Lapatinib (PO, 80mg/kg) or 446 Poziotinib (SC, 4mg/kg) for two cycles of 5 days ON and 2 days OFF. Solvent controls were 10% solutol 447 in water (for P oziotinib) and Phosal/PEG300 (for L apatinib). MR images were acquired using a 7T 448 cryogen-free superconducting magnet (MR Solutions , Boston, MA, USA) with a 17cm bore and 449 equipped with a dedicated quadrature mouse head coil. Animals were anesthetized with isoflurane and 450 transferred to a warmed Minerve© bed for head fixation and further anesthetic delivery. Respiration rates 451 were monitored throughout the procedure. Following a preliminary scout scan, acquisition planes were 452 optimized, and whole-brain coronal T2- weighted scans were acquired using a fast spin echo sequence 453 with the following parameters: TR 5000ms, TE 45ms, echo trains 7, FOV 30x30mm 2, matrix size 454 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 16 250x256, total slices 18, slice thickness 0.3mm, and two averages for a total scan time 350s. For MRI 455 lesion volumetry, region of interest (ROI) -based volumetry was performed on pre - and post-treatment 456 sets of MRI scans (n=4/ group). The total MRI lesion volume was quantified as a sum of individual 457 volumes (V = πr2h) for each MRI section containing the lesion. Measurement of drug penetration into 458 mouse brain were performed with C57BL mice. Poziotinib was administered by either oral gavage (PO) 459 or subcutaneous injection ( SC). P lasma and brain tissues were collected immediately after the last 460 treatment and snap-frozen in liquid nitrogen for ultra-performance liquid chromatography-tandem mass 461 spectrometry (UPLC -MS/MS) analysis. The calibration standards and quality control samples were 462 prepared in the blank mouse plasma and brain homogenate. Aliquots of 10 µL samples were mixed with 463 200 µL internal standard in acetonitrile to precipitate proteins in a 96-well plate. 1.0 µL supernatant was 464 injected for the UPLC -MS/MS analysis. MassLynx and TargetLynx were used for data collection and 465 processing (Waters Corp., Milford, MA). For brain histology and IHC, mice were euthanized using 466 isofluorane and cervical dislocation upon two treatment cycles with 5 days ON and 2 days OFF. Brains 467 were removed immediately and hemispheres were fixed in 10% neutral buffered formalin for 24h. FFPE 468 brain sections of 5µm were subjected to H&E staining and IHC. 469 Statistical analysis 470 A paired two-tailed t-test was used to determine if there was a significant difference between the means 471 of the two groups. One-way ANOVA was used to identify any statistically significant difference between 472 the means of three or more groups. The Tukey test was performed to check if there is a statistically 473 significant difference between the pairs of samples that belong to a larger group where the statistical 474 difference proved to be significant in the ANOVA test. 475 476

477 The brain-seeking subline of the MDA -MB-231 cell line, MDA -MB-231/BR, was kindly provided by 478 Dr. Patricia Steeg (NCI, Bethesda, Maryland). We thank the nurses Deb Swan, Coleen Unger and Susan 479 Pearce at the Neurosurgery Clinic, Health Sciences Center, Winnipeg, for their tremendous support with 480 patient consent. The authors acknowledge the support from the staff at the Electron Microscopy Core 481 Platform and the Histology Services, Department of Human Anatomy and Cell Sciences, University of 482 Manitoba. We acknowledge the staff at the Central Animal Care Services and the Small Animal Imaging 483 Core Facility, University of Manitoba, for their help throughout the study. We acknowledge funding 484 support from the Natural Sciences and Engineering Research Council of Canada (TK), CancerCare 485 Manitoba (TK, SHK), the Dr. Paul H.T. Thorlakson Foundation Fund (SHK), the Cancer Research 486 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 17 Society (SHK), Mitacs (DI) , the Max Rady College of Medicine (TT) , the University of Manitoba 487 Graduate Fellowship (JS), and the NCATS/NIH Intramural Research Program (YHL, XX, AW, RC, AK, 488 JJM, MJH). 489 490 Author contributions 491 Conceptualization of study (TK, SHK) ; d ata curation and formal analysis ( DI, YHL, AG, TT , AK); 492 animal surgeries (TK, JS); animal imaging (JS); patient neurosurgery (JB), pathology diagnosis (MDB); 493 drug formulation (RC); drug screening methodology (YHL, MJH), pharmacokinetic data (AW, XX) ; 494 resources and supervision (JJM, MJH, SHK, TK). 495 496 Competing interests: No competing interests declared 497 498 499 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 18

500 1. The-American-Cancer-Society. Treatment of Breast Cancer By Stage. 2022. 501 2. Leone J, Lin N. Systemic Therapy of Central Nervous System Metastases of Breast Cancer. 502 Current oncology reports. 2019;21(6). 503 3. Schlam I, Swain SM. HER2-positive breast cancer and tyrosine kinase inhibitors: the time is 504 now. NPJ Breast Cancer. 2021;7(1):56. 505 4. National-Cancer-Institute. Cancer Treatment. 2020. 506 5. Zhong L, Y L, Xiong L, Wang W, Wu M, Yuan T, et al. Small molecules in targeted cancer 507 therapy: advances, challenges, and future perspectives. Signal transduction and targeted therapy. 508 2021;6(1). 509 6. Lin N, Carey L, Liu M, Younger J, Come S, Ewend M, et al. Phase II trial of lapatinib for brain 510 metastases in patients with human epidermal growth factor receptor 2-positive breast cancer. Journal 511 of clinical oncology : official journal of the American Society of Clinical Oncology. 2008;26(12). 512 7. Lin N, Diéras V, Paul D, Lossignol D, Christodoulou C, Stemmler H, et al. Multicenter phase II 513 study of lapatinib in patients with brain metastases from HER2-positive breast cancer. Clinical cancer 514 research : an official journal of the American Association for Cancer Research. 2009;15(4). 515 8. Freedman R, Gelman R, Anders C, Melisko M, Parsons H, Cropp A, et al. TBCRC 022: A Phase II 516 Trial of Neratinib and Capecitabine for Patients With Human Epidermal Growth Factor Receptor 2-517 Positive Breast Cancer and Brain Metastases. Journal of clinical oncology : official journal of the 518 American Society of Clinical Oncology. 2019;37(13). 519 9. Bachelot T, Romieu G, Campone M, Diéras V, Cropet C, Dalenc F, et al. Lapatinib plus 520 capecitabine in patients with previously untreated brain metastases from HER2-positive metastatic 521 breast cancer (LANDSCAPE): a single-group phase 2 study. The Lancet Oncology. 2013;14(1). 522 10. Ullrich A, Schlessinger J. Signal transduction by receptors with tyrosine kinase activity. Cell. 523 1990;61(2). 524 11. Citri A, Skaria K, Yarden Y. The deaf and the dumb: the biology of ErbB-2 and ErbB-3. 525 Experimental cell research. 2003;284(1). 526 12. Roskoski R. The ErbB/HER family of protein-tyrosine kinases and cancer. Pharmacological 527 research. 2014;79. 528 13. Weitsman G, Barber PR, Nguyen LK, Lawler K, Patel G, Woodman N, et al. HER2-HER3 dimer 529 quantification by FLIM-FRET predicts breast cancer metastatic relapse independently of HER2 IHC 530 status. Oncotarget. 2016;7(32):51012-26. 531 14. Berghoff AS, Bartsch R, Preusser M, Ricken G, Steger GG, Bago-Horvath Z, et al. Co-532 overexpression of HER2/HER3 is a predictor of impaired survival in breast cancer patients. Breast. 533 2014;23(5):637-43. 534 15. Jia R, Zhao H, Wang S. Neuregulin Signaling in the Tumor Microenvironment. Advances in 535 experimental medicine and biology. 2021;1270. 536 16. Luo X, Prior M, He W, Hu X, Tang X, Shen W, et al. Cleavage of neuregulin-1 by BACE1 or 537 ADAM10 protein produces differential effects on myelination. The Journal of biological chemistry. 538 2011;286(27). 539 17. Fleck D, van Bebber F, Colombo A, Galante C, Schwenk B, Rabe L, et al. Dual cleavage of 540 neuregulin 1 type III by BACE1 and ADAM17 liberates its EGF-like domain and allows paracrine 541 signaling. The Journal of neuroscience : the official journal of the Society for Neuroscience. 542 2013;33(18). 543 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 19 18. Wieduwilt M, Moasser M. The epidermal growth factor receptor family: biology driving 544 targeted therapeutics. Cellular and molecular life sciences : CMLS. 2008;65(10). 545 19. Li Q, Ahmed S, Loeb J. Development of an autocrine neuregulin signaling loop with malignant 546 transformation of human breast epithelial cells. Cancer research. 2004;64(19). 547 20. Cabrera R, Mao S, Surve C, Condeelis J, Segall J. A novel neuregulin - jagged1 paracrine loop in 548 breast cancer transendothelial migration. Breast cancer research : BCR. 2018;20(1). 549 21. Jeong H, Kim J, Lee Y, Seo J, Hong S, Kim A. Neuregulin-1 induces cancer stem cell 550 characteristics in breast cancer cell lines. Oncology reports. 2014;32(3). 551 22. The Human Protein Atlas. 2022(21.1). 552 23. Law AJ, Shannon Weickert C, Hyde TM, Kleinman JE, Harrison PJ. Neuregulin-1 (NRG-1) mRNA 553 and protein in the adult human brain. Neuroscience. 2004;127(1):125-36. 554 24. Guo Z, Su Y, Wang Y, Wang W, Guo D. The expression pattern of Adam10 in the central 555 nervous system of adult mice: Detection by in situ hybridization combined with 556 immunohistochemistry staining. Molecular medicine reports. 2016;14(3). 557 25. Dominguez-Garcia S, Castro C, Geribaldi-Doldán N. ADAM17/TACE: a key molecule in brain 558 injury regeneration. Neural regeneration research. 2019;14(8). 559 26. Valiente M, Van Swearingen AED, Anders CK, Bairoch A, Boire A, Bos PD, et al. Brain Metastasis 560 Cell Lines Panel: A Public Resource of Organotropic Cell Lines. Cancer Res. 2020;80(20):4314-23. 561 27. Xia W, Petricoin E, Zhao S, Liu L, Osada T, Cheng Q, et al. An heregulin-EGFR-HER3 autocrine 562 signaling axis can mediate acquired lapatinib resistance in HER2+ breast cancer models. Breast cancer 563 research : BCR. 2013;15(5). 564 28. D'Amato V, Raimondo L, Formisano L, Giuliano M, De Placido S, Rosa R, et al. Mechanisms of 565 lapatinib resistance in HER2-driven breast cancer. Cancer treatment reviews. 2015;41(10). 566 29. Garrett J, Olivares M, Rinehart C, Granja-Ingram N, Sánchez V, Chakrabarty A, et al. 567 Transcriptional and posttranslational up-regulation of HER3 (ErbB3) compensates for inhibition of the 568 HER2 tyrosine kinase. Proceedings of the National Academy of Sciences of the United States of 569 America. 2011;108(12). 570 30. Hegde P, Rusnak D, Bertiaux M, Alligood K, Strum J, Gagnon R, et al. Delineation of molecular 571 mechanisms of sensitivity to lapatinib in breast cancer cell lines using global gene expression profiles. 572 Molecular cancer therapeutics. 2007;6(5). 573 31. SW B, J Z, MH T, D Y. PI3K-independent mTOR activation promotes lapatinib resistance and IAP 574 expression that can be effectively reversed by mTOR and Hsp90 inhibition. Cancer biology & therapy. 575 2015;16(3). 576 32. Breuleux M. Role of heregulin in human cancer. Cellular and molecular life sciences : CMLS. 577 2007;64(18). 578 33. Datta S, Dudek H, Tao X, Masters S, Fu H, Gotoh Y, et al. Akt phosphorylation of BAD couples 579 survival signals to the cell-intrinsic death machinery. Cell. 1997;91(2). 580 34. Cheng J, Jiang X, Fraser M, Li M, Dan H, Sun M, et al. Role of X-linked inhibitor of apoptosis 581 protein in chemoresistance in ovarian cancer: possible involvement of the phosphoinositide-3 582 kinase/Akt pathway. Drug resistance updates : reviews and commentaries in antimicrobial and 583 anticancer chemotherapy. 2002;5(3-4). 584 35. Anandharaj A, Cinghu S, Park W. Rapamycin-mediated mTOR inhibition attenuates survivin 585 and sensitizes glioblastoma cells to radiation therapy. Acta biochimica et biophysica Sinica. 586 2011;43(4). 587 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 20 36. Yang L, Y L, Shen E, Cao F, Li L, Li X, et al. NRG1-dependent activation of HER3 induces primary 588 resistance to trastuzumab in HER2-overexpressing breast cancer cells. International journal of 589 oncology. 2017;51(5). 590 37. Shibuya M, Komi E, Wang R, Kato T, Watanabe Y, Sakai M, et al. Measurement and comparison 591 of serum neuregulin 1 immunoreactivity in control subjects and patients with schizophrenia: an 592 influence of its genetic polymorphism. Journal of neural transmission (Vienna, Austria : 1996). 593 2010;117(7). 594 38. Hardwick JM, Soane L. Multiple functions of BCL-2 family proteins. Cold Spring Harb Perspect 595 Biol. 2013;5(2). 596 39. Singh R, Letai A, Sarosiek K. Regulation of apoptosis in health and disease: the balancing act of 597 BCL-2 family proteins. Nature reviews Molecular cell biology. 2019;20(3). 598 40. Zimmer AS, Van Swearingen AED, Anders CK. HER2-positive breast cancer brain metastasis: A 599 new and exciting landscape. Cancer Rep (Hoboken). 2022;5(4):e1274. 600 41. Kuksis M, Gao Y, Tran W, Hoey C, Kiss A, Komorowski AS, et al. The incidence of brain 601 metastases among patients with metastatic breast cancer: a systematic review and meta-analysis. 602 Neuro Oncol. 2021;23(6):894-904. 603 42. Lin N, Winer E. Brain metastases: the HER2 paradigm. Clinical cancer research : an official 604 journal of the American Association for Cancer Research. 2007;13(6). 605 43. Hurvitz SA, O'Shaughnessy J, Mason G, Yardley DA, Jahanzeb M, Brufsky A, et al. Central 606 Nervous System Metastasis in Patients with HER2-Positive Metastatic Breast Cancer: Patient 607 Characteristics, Treatment, and Survival from SystHERs. Clin Cancer Res. 2019;25(8):2433-41. 608 44. Kuksis M, Gao Y, Tran W, Hoey C, Kiss A, Komorowski A, et al. The incidence of brain 609 metastases among patients with metastatic breast cancer: a systematic review and meta-analysis. 610 Neuro-oncology. 2021;23(6). 611 45. O'Sullivan CC, Davarpanah NN, Abraham J, Bates SE. Current challenges in the management of 612 breast cancer brain metastases. Semin Oncol. 2017;44(2):85-100. 613 46. Dai X, Cheng H, Bai Z, Li J. Breast Cancer Cell Line Classification and Its Relevance with Breast 614 Tumor Subtyping. Journal of Cancer. 2017;8(16). 615 47. L M, M V. Animal models of brain metastasis. Neuro-oncology advances. 2021;3(Suppl 5). 616 48. Buonanno A, Fischbach G. Neuregulin and ErbB receptor signaling pathways in the nervous 617 system. Current opinion in neurobiology. 2001;11(3). 618 49. Mei L, Nave K. Neuregulin-ERBB signaling in the nervous system and neuropsychiatric diseases. 619 Neuron. 2014;83(1). 620 50. Fernald K, Kurokawa M. Evading apoptosis in cancer. Trends in cell biology. 2013;23(12). 621 51. Mann J, Githaka JM, Buckland TW, Yang N, Montpetit R, Patel N, et al. Non-canonical BAD 622 activity regulates breast cancer cell and tumor growth via 14-3-3 binding and mitochondrial 623 metabolism. Oncogene. 2019;38(18):3325-39. 624 52. Kim TM, Lee KW, Oh DY, Lee JS, Im SA, Kim DW, et al. Phase 1 Studies of Poziotinib, an 625 Irreversible Pan-HER Tyrosine Kinase Inhibitor in Patients with Advanced Solid Tumors. Cancer Res 626 Treat. 2018;50(3):835-42. 627 53. Elamin YY, Robichaux JP, Carter BW, Altan M, Gibbons DL, Fossella FV, et al. Poziotinib for 628 Patients With HER2 Exon 20 Mutant Non-Small-Cell Lung Cancer: Results From a Phase II Trial. J Clin 629 Oncol. 2022;40(7):702-9. 630 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 21 54. Le X, Cornelissen R, Garassino M, Clarke JM, Tchekmedyian N, Goldman JW, et al. Poziotinib in 631 Non-Small-Cell Lung Cancer Harboring HER2 Exon 20 Insertion Mutations After Prior Therapies: 632 ZENITH20-2 Trial. J Clin Oncol. 2022;40(7):710-8. 633 55. Qi X, Shi Q, Xuhong J, Zhang Y, Jiang J. Pyrotinib-based therapeutic approaches for HER2-634 positive breast cancer: the time is now. Breast Cancer Res. 2023;25(1):113. 635 56. Gupta M, HJ L, Barden C, Weaver D. The Blood-Brain Barrier (BBB) Score. Journal of medicinal 636 chemistry. 2019;62(21). 637 57. F M, J C, PW A. Predictors for Blood-Brain Barrier Permeability with resampling strategies 638 based on B3DB database. 2019. 639 58. SumulationsPlus. ADMET Predictor®. 640 59. Franken N, Rodermond H, Stap J, Haveman J, van Bree C. Clonogenic assay of cells in vitro. 641 Nature protocols. 2006;1(5). 642 60. Livak K, Schmittgen T. Analysis of relative gene expression data using real-time quantitative 643 PCR and the 2(-Delta Delta C(T)) Method. Methods (San Diego, Calif). 2001;25(4). 644 61. ACDBio. RNAscope Sample Preparation and Pretreatment Guide for FFPE Tissue 2013 [ 645 62. ACDBio. RNAscope 2.5 HD Duplex Detection Kit (Chromogenic) 2019 [ 646 63. Baumann BC, Dorsey JF, Benci JL, Joh DY, Kao GD. Stereotactic intracranial implantation and in 647 vivo bioluminescent imaging of tumor xenografts in a mouse model system of glioblastoma 648 multiforme. J Vis Exp. 2012(67). 649 650 651 652 653 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 22 MAIN FIGURES 654 655 656 Figure 1. HER2+ BCBM94 cells establish hematogenic brain metastasis in mice. 657 (A) BCBM94 cells produce hematogenous brain metastasis upon intracardiac xenografting in RAG2yc-658 /- mice. H&E and IHC staining for Ki67+ nuclei and CD31+ endothelial cells in FFPE mouse brain 659 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 23 tissues containing BCBM94 lesions. Magnification 200x (B,C) BCBM94 cells retain HER2 expression 660 in-vivo and in-vitro. HER2 IHC staining of FFPE tissue sections of patient’s breast, patient’s brain, and 661 mouse brain containing BCBM94 lesions. Magnification 200x. ICC staining of cultured formaldehyde -662 fixed BCBM94 cells show strong membrane expression of HER2 (ErbB2). (D) BCBM94 is a Luminal-663 B HER2+ BC model. Western blot (WB) comparing total protein expression of HER2 (ErbB2), ER α, 664 and PRα/β in the triple-negative MDA-MB-BR and HER2+ BCBM94, BT474, and SKBR3 cell lines in-665 vitro. (E) NRG1 is expressed in the TME of BCBM94 brain metastases. In-situ expression of NRG1 and 666 ErbB2 mRNA in the FFPE mouse brain tissue containing BCBM94 metastases was assessed with the 667 RNAscope 2.5 HD Duplex assay. Black arrows indicate NRG1 mRNA (red dots) within non- tumoral 668 cells in the mouse brain. NRG1+ cells are abundantly present in various regions of the mouse brain, 669 including the TME of HER2+ BCBM94 (blue) metastasis. Magnification 200x. ( F) HER2+ BC cell 670 models are devoid of NRG1. WB compared NRG1 protein expression in the triple-negative MDA-MB-671 BR and HER2+ BCBM94, BT474, and SKBR3 cell lines in-vitro. 672 673 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 24 674 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 25 Figure 2. NRG1 rescues BCBM94 cells from Lapatinib-induced cytotoxicity. 675 (A-D) Cell viability of BCBM94 under Lapatinib (Lap) +/- rhNRG1 treatment was assessed in the WST-676 1 assay. The endpoint absorbance readouts were used for quantification of the relative cell viability (mean 677 +/- SD, n=3) (A-C). Cell proliferation and colony formation potential of BCBM94 cells under Lap +/ -678 rhNRG1 treatment were assessed in colony formation assays (D). The bar chart presents the average 679 number of colonies formed by BCBM94 cells under Lap +/- rhNRG1 treatment (n=2) (D). Representative 680 images of colonies are shown below the plot (D). (E-I) NRG1 counteracts Lapatinib-induced apoptosis. 681 The chart presents cleaved / full length PARP protein ratio in BCBM94 cells under Lap +/ -rhNRG1 682 conditions measured by Western blot (WB) and quantified with densitometry (n=3) (E). A representative 683 WB is shown below the plot (E). WB detection of cleaved and full-length PARP proteins in BT474 cells 684 is shown under Lap +/- rhNRG1 conditions (n=1) (F). Relative luminescence values represent the activity 685 of caspase-3 and caspase -7 under Lap +/ -rhNRG1 conditions measur ed using a CaspaseGlo 3/7 assay 686 (mean +/- SD, n=3) (G). Detection of the cleaved/ pro- caspase-9 protein ratio in BCBM94 cells under 687 Lap +/ -rhNRG1 conditions was measured by WB and quantified with densitometry (n=3) (H) . A 688 representative WB is shown below the chart (H). WB detection of cleaved / pro - caspase-9 proteins in 689 BT474 cells under L ap +/ - rhNRG1 conditions (n=1) (I). Bar charts present mean +/ - SD,*p<0.05, 690 **p<0.01. 691 692 693 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 26 694 695 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 27 Figure 3. Anti-apoptotic actions of NRG1 involve BCL2 proteins 696 (A,B) NRG1 rescues Bad phosphorylation under Lapatinib (Lap). The phospho-/ total Bad protein ratio 697 in BCBM94 cells under Lap +/ -rhNRG1 conditions was determined by WB and quantified with 698 densitometry (n=3) (A). A representative WB is shown below the chart (A). WB images show expression 699 of phospho- and total Bad proteins in BT474 cells under L ap +/- rhNRG1 conditions (n=1) (B). (C) 700 NRG1 protects mitochondria from L apatinib-induced damage. Relative fluorescence intensity values 701 represent the number of active mitochondria in BCBM94 cells under Lap +/-rhNRG1 conditions detected 702 with MitoTracker® (n=2). Representative IF images are shown below the graph, magnification 200x (C). 703 (D) NRG1 prevents aggregation of the mitochondria outer membrane pore -formers Bax and Bak under 704 Lapatinib. BCBM94 cells treated with Lap +/-rhNRG1 were PFA-fixed for detection of Bak and Bax by 705 ICC/IF. The white arrow indicates punctate Bax aggregates co -localizing with Bak under L apatinib 706 treatment (n=2). Magnification 630x. Bar charts present mean +/- SD, *p<0.05, **p<0.01. 707 708 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 28 709 710 Figure 4. NRG1 rescues ErbB3 phosphorylation under HER2 inhibition in BCBM cells 711 (A-D) NRG1 rescues ErbB3 phosphorylation under L apatinib (Lap). Protein ratios for phospho- / total 712 ErbB2 and phospho-/ total ErbB3 were determined in BCBM94 cells under Lap +/-rhNRG1 by WB and 713 quantified with densitometry (n=3) (A). Total ErbB1 (EGFR) and total ErbB3 protein levels in BCBM94 714 cells under Lap +/ -rhNRG1 conditions were measured by WB and quantified with densitometry (n=3) 715 (B). Representative WBs are shown below the charts (A, B). WB images show the detection of phospho-716 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 29 ErbB1-4 and total ErbB1-4 proteins in BT474 cells under Lap +/ -rhNRG1 conditions (n=1) (C,D). (E) 717 Phosphorylated ErbB3 is expressed by BCBM94 tumors in -vivo. IHC analysis of patient’s breast, 718 patient’s brain, and mouse brain FFPE tissue sections containing BCBM94 lesions, magnification 200x. 719 Bar charts present mean +/- SD, *p<0.05, **p<0.01. 720 721 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 30 722 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 31 Figure 5. ErbB3 signaling mediates anti -apoptotic actions of NRG1 under Lapatinib (A, B) 723 Knockdown of ErbB3 increases PARP cleavage under Lapatinib (Lap) and mitigates NRG1 rescue. The 724 cleaved/ full -length PARP protein ratio in BCBM94 cells under Lap +/ -rhNRG1 +/ -ErbB3siRNA 725 conditions was measured by WB and quantified with densitometry (n=3) (A). A representative WB is 726 shown below the chart (A). Protein levels of cleaved and full-length PARP proteins in BT474 cells under 727 Lap +/-rhNRG1 +/-ErbB3siRNA conditions are shown by WB (n=1) (B). (C, D) Knockdown of ErbB3 728 attenuates rhNRG1 -mediated rescue of phospho- Bad under combined Lap/ rhNRG1 treatment. The 729 graph presents phospho -/ total Bad protein ratio in BCBM94 cells under Lap +/ -rhNRG1 +/ -730 ErbB3siRNA conditions measured by WB and quantified with densitometry (n=3) (C). A representative 731 WB is shown below the chart (C). Protein levels of phospho- and total Bad proteins in BT474 cells under 732 Lap +/ -rhNRG1 +/ -ErbB3siRNA conditions were detected by WB (n=1) (D). ( E, F) NRG1 rescues 733 expression and phosphorylation of Akt under Lapatinib. WB images show expression of phospho- and 734 total Akt proteins in BCBM94 (representative examples, n=3) (E) and BT474 (n=1) (F) cells under Lap 735 +/-rhNRG1 +/-ErbB3siRNA conditions. (G, H) The anti-apoptotic action of NRG1 is mediated through 736 Akt. The graph presents phospho- / total Bad protein ratio in BCBM94 cells under Lap +/ -rhNRG1 737 treatment measured by WB and quantified with densitometry; the PI3K inhibitor PI -103 was used at 10 738 µM (G). Representative WB images are shown below the chart (n=3) (G) and present the levels of 739 phospho-/ total-Akt and cleaved/ full -length PARP proteins in BCBM94. Protein levels of phospho- / 740 total-Akt, cleaved/ full -length PARP and phospho- / total Bad proteins in BT474 cells under Lap +/ -741 rhNRG1 exposure and treatment with PI -103 were determined by WB (n=1). Bar charts present mean 742 +/- SD, **p<0.01 743 744 745 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 32 746 747 Figure 6. Schematic illustration of the NRG1 actions that rescue Lapatinib-induced apoptosis in 748 HER2+ BCBM cells. 749 750 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 33 751 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 34 Figure 7. Poziotinib-induced cytotoxicity is not diminished by NRG1 752 (A) Fifty ErbB inhibitors were tested for differential sensitivity in BCBM94 versus HME1. 53 samples 753 (3 compounds were represented by two separate batches) were tested at 22 concentrations ranging from 754 0.003pM to 40uM (BCBM94 n=4, HME1 n=1). The area under the curve (AUC, y-axis) of the non-linear 755 fitting of the dose -response data was calculated as a metric to compare differential sensitivity to the 756 compounds. The potency of each compound in the BCBM94 model is indicated on the x -axis. 757 Compounds in the lower left corner of the plot are the most potent and BCBM94-selective. (B) Poziotinib 758 is highly efficacious in reducing cell viability of both BCBM94 and BT474 HER2+ BC models. ErbB 759 inhibitors were tested at concentrations ranging from 0.003 pM to 40uM and viability was measured 760 using a CellTiterGlo assay after 72 h (BCBM94 n=4, BT474 n=1). ( C) rhNRG1 reduces the cytotoxic 761 activity of many ErbB inhibitors. The 50 ErbB inhibitors were tested +/ - 5ng/mL rhNRG1 and viability 762 was measured after 72h using CellTiterGlo. The differential sensitivity was assessed by comparing IC50 763 values (ratio) under both conditions (BCBM94 n=3, BT474 n=1). Data points that fall near the 764 intersection of the dotted lines, including Poziotinib (Poz), represent compounds with equipotent activity 765 +/- NRG1. Compounds to the upper right of the plot are those where NRG1 reduced the cytotoxic effect 766 of the compound. ( D-F) rhNRG1 failed to rescue tested HER2+ BC models from P oziotinib-mediated 767 cytotoxicity. Cell viability of BCBM94 (n=3) (D, E) and BT474 (n=3) (F) under Poz +/ -rhNRG1 768 treatment was assessed in the WST -1 assay. The endpoint absorbance readouts were used for 769 quantification of the relative cell viability. ( G, H) rhNRG1 fail ed to counteract P oziotinib-induced 770 apoptosis. WB images show the levels of cleaved/ full -length PARP in BCBM94 (representative 771 examples, n=3) ( G) and BT474 (n=1) ( H) cells under Poz +/ -rhNRG1 treatment. ( I, J) rhNRG1 wa s 772 unable to rescue ErbB3 phosphorylation under Poziotinib. Representative WB images show the levels of 773 phospho-/ total ErbB3 in BCBM94 (n=3) (I) and BT474 (n=1) (J) cells under Poz +/-rhNRG1 treatment. 774 (K, L) rhNRG1 failed to rescue Akt phosphorylation under Poziotinib. Representative WB images show 775 the levels of phospho- / total ErbB3 in BCBM94 (n=3) (K) and BT474 (n=) (L) cells under Poz +/ -776 rhNRG1 treatment. Bar charts present mean +/- SD, *p<0.05; **p<0.01. 777 778 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 35 779 780 Figure 8. Schematic illustration showing the inability of NRG1 to rescue phosphorylation of ErbB3 781 and AKT resulting in PARP cleavage and Poziotinib-induced apoptosis in BCBM cells. 782 783 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 36 784 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 37 Figure 9. Poziotinib effectively reduced brain metastases in mice 785 BCBM94 and BT474 cells were orthotopically xenografted into SCID and RAG2yc -/- mice. Upon 786 detection of sizable brain metastases with MRI, the animals were treated with either Lapatinib (80mg/kg), 787 Poziotinib (4mg/kg), or solvent control for two 5- day cycles with 2 days off treatment in between. (A) 788 Treatment with P oziotinib resulted in a significant reduction of BCBM94 and BT474 brain metastatic 789 tumors. The ROI -based volumetry was performed on pre - and post -treatment sets of MRI scans 790 (n=4/group). Bar charts present mean +/ - SD, **p<0.01. Representative MR images are shown below 791 the charts (A, B). (C) Poziotinib is more brain penetrable than Lapatinib. Drug concentrations were 792 measured by UPLC -MS/MS analysis of plasma and brain tissues taken 1 hour after the last drug 793 administration (n=4/group). (D) Poziotinib abrogate d ErbB3 phosphorylation, inhibit ed proliferation, 794 and induce d apoptosis within BCBM94 brain metastasis . H&E, IHC , and TUNEL analysis of post -795 treatment mouse brain FFPE tissues is shown for solvent control, L apatinib and P oziotinib treatment 796 groups. The white dashed lines mark the margins of the metastatic lesion. ErbB3 phosphorylation and 797 presence of Ki67+ nuclei we re observed only in lesions of the solvent control and L apatinib groups. 798 TUNEL assay shows fragmented DNA in green color. Magnification: H&E 100x, Ki67 200x, TUNEL 799 and pErbB3 400x. 800 801 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 38 SUPPLEMENTARY FIGURES 802 803 804 Suppl. Figure 1: NRG1 counteracts Lapatinib-induced loss of viability 805 A) BCBM94 cells have very low expression of NRG1-4 mRNA in-vitro. qRT-PCR compared mRNA 806 levels of NRG1-4 in the triple-negative MDA-MB-BR and HER2+ SKBR3 and BCBM94 cells. A cycle 807 count higher than 30 (Ct > 30) is considered as low expression independent of relative values. Values for 808 MDA-MB-BR are set to 1. To calculate the relative expression of NRG3 in BCBM94, Ct values for 809 MDA-MB-BR and SKBR3 were set to 40 (maximum cycle number). ND : not detected. B,C) rhNRG1 810 rescued BCBM94 cells from Lapatinib -induced cytotoxicity. Real -time cell analysis based on 811 impedance measurements was performed with the xCELLigence assay. The graph depicts measurements 812 of BCBM94 cells treated with Lap +/-rhNRG1 over a period of 25 hours (B). The terminal 24h readouts 813 of the xCELLigence assay were plotted on the bar graph (C). D) rhNRG1 rescued BT474 cells from 814 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 39 Lapatinib-induced cytotoxicity. Cell viability of BT474 cells under Lap +/ - rhNRG1 treatment was 815 assessed in the WST -1 assay. The endpoint absorbance readouts were used for quantification of the 816 relative cell viability. E) Lapatinib did not upregulate endogenous expression of NRG1 in BCBM94 817 cells. WB images present the level of NRG1 protein in BCBM94 cells under two different concentrations 818 of Lapatinib. MD-MB-BR cells serve as positive control for NRG1. 819 820 821 822 823 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 40 Suppl. Figure 2: NRG1 counteracts Lapatinib-induced apoptosis 824 A, B) rhNRG1 rescued BCBM94 and BT474 cells from Lapatinib -induced apoptosis even after 825 pre-incubation with Lapatinib. BCBM94 and BT474 cells were treated with Lapatinib (Lap) for the 826 first 24h and either Lapatinib or Lap/ rhNRG1 for another 24h. WB images show the levels of cleaved 827 and full -length PARP proteins in BCBM94 (A) and BT474 (B) cells under the above-mentioned 828 conditions. C-F) Total levels of the antiapoptotic and proapoptotic (effectors) proteins of the Bcl-2 829 family were unaffected by Lap +/ -rhNRG1 treatment. WB images present the levels of Bcl -2, Bcl-830 XL, Mcl-1, Bax and Bak protein in BCBM94 (C,E) and BT474 (D,F) cell lines under Lap +/ -rhNRG1 831 treatment for the indicated times. G) Transmission electron microscopy (TEM) images. NRG1 832 maintained mitochondrial cristae structure under Lapatinib. TEM images of BCBM94 cells incubated 833 with Lap +/ -rhNRG1. Black arrows indicate damaged mitochondria under L apatinib treatment. 834 Magnifications are indicated on the image. 835 836 837 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 41 Suppl. Figure 3: NRG1 fails to rescue poziotinib-induced viability loss. 838 A, B) rhNRG1 failed to rescue BCBM94 cells from Poziotinib (Poz)-induced cytotoxicity. Real-time 839 cell analysis using the xCELLigence assay show cell impedance measurements in BCBM94 cells treated 840 with Poz +/-rhNRG1 over a period of 25 hours (A). The terminal 24h readouts of the xCELLigence assay 841 were plotted on the bar graph (B). Bar charts present mean +/ - SD, **p<0.01; ***p<0.001; 842 ****p<0.0001. C) Poziotinib brain concentrations. Poziotinib was administered in C57BL/6 mice by 843 oral gavage (PO) at 2 and 5 mg/kg and by subcutaneous injections (SC) at 1 and 5 mg/kg. Concentrations 844 in brain tissues were determined by UPLC -MS/MS bioanalytical method at different time points after 845 drug administration. The dotted line represents the IC90 concentration, as determined by in-vitro viability 846 experiments. 847 848 849 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint 42 850 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 22, 2024. ; https://doi.org/10.1101/2024.02.19.581073doi: bioRxiv preprint