Stratifying macrophages based on their infectious burden identifies novel host targets for intervention during Crohn’s disease associated adherent-invasiveEscherichia coliinfection

38 Bacterial infection is a dynamic process resulting in a heterogenous population of infected 39 and uninfected cells. These cells respond differently based on their bacterial load and 40 duration of infection. In the case of infection of macrophages with Crohn’s disease (CD) 41 associated adherent-invasive Escherichia coli (AIEC), understanding the drivers of pathogen 42 success may allow targeting of cells where AIEC replicate to high levels. Here we show that 43 stratifying immune cells based on their bacterial load identifies novel pathways and 44 therapeutic targets not previously associated with AIEC when using a traditional 45 homogeneous infected population approach. Using flow cytometry-based cell sorting we 46 stratified cells into those with low or high intracellular pathogen loads, or those which were 47 bystanders to infection. Immune cells transcriptomics revealed a diverse response to the 48 varying levels of infection while pathway analysis identified novel intervention targets that 49 were directly related to increasing intracellular AIEC numbers. Chemical inhibition of 50 identified targets reduced AIEC intracellular replication or inhibited secretion of tumour 51 necrosis factor alpha (TNF), a key cytokine associated with AIEC infection. Our results 52 have identified new avenues of intervention in AIEC infection that may also be applicable to 53 CD through the repurposing of already available inhibitors. Additionally, they highlight the 54 applicability of immune cell stratification post-infection as an effective approach for the study 55 of microbial pathogens. 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint

75 Infection is a dynamic process with a highly heterogenous population of host cells infected to 76 varying degrees by infiltrating microorganisms. These differing microbial loads can lead to a 77 variety of outcomes for the infected cells and a heterogeneity in host responses. Replicating 78 the dynamics of infection using either in vivo or in vitro models of disease is challenging, but 79 these models have proved highly useful tools in understanding specific aspects of infection. 80 While heterogeneity fundamentally underlies in vivo models of disease, in vitro models by 81 design are often based on the interaction between a single pathogen-and a particular host 82 cell type in a more controlled environment. This reduction in complexity has clarified aspects 83 of the host or microbial response to infection, confirming or raising hypotheses for later 84 testing in more complex models. 85 In vitro models of bacterial infection often require high multiplicities of infection (MOIs) to 86 ensure a bacterial intracellular burden high enough to enable host-pathogen dynamics to 87 proceed in a measurable way over time. While MOIs into the hundreds are common, these 88 rarely result in homogenous infection by, or phagocytosis of, all bacteria present within the 89 system. What results is a mixture of sub-populations with varying degrees of infectious load, 90 with either no bacterial infection having occurred, low levels of intracellular bacteria or a high 91 intracellular bacterial load. Yet these diverse sub-populations have traditionally been studied 92 as a single homogenous population, leading to the potential loss of information critical to 93 understanding the infection process. For example, there may be contrasting outcomes in 94 immune cells where bacteria are overcome in some cells, while actively replicating 95 intracellularly in others, yet the basis of these outcomes are generally not investigated in in 96 vitro models. 97 Adherent-invasive Escherichia coli (AIEC) is a pathobiont isolated in increased frequency 98 from the intestine of CD patients relative to healthy controls (Darfeuille-Michaud et al., 1998; 99 Martinez-Medina et al., 2009; Nadalian et al., 2021). CD is a multifactorial disease with 100 genetic susceptibility, dietary factors and microorganisms all playing a role in disease 101 pathogenesis. Rising incidence, the increasingly young age of onset, and incurability of the 102 disease mean that as well as reducing quality of life, CD is a significant burden on health 103 care systems across the world (Bassi et al., 2004; Rao et al., 2017). While genetic 104 susceptibilities linked to CD are well defined, specific defects in autophagy and protection 105 against intracellular bacteria have not explained why bacteria such as AIEC are found with 106 increasing prevalence. AIEC lacks many of the classic virulence factors associated with E. 107 coli pathotypes and its persistence in the CD gut is likely mediated via metabolic success 108 and adaption to the conditions in the inflamed CD gut (Ormsby et al., 2019, 2020; Cho et al., 109 2022; Sugihara et al., 2022). A hall mark of AIEC infection is replication to high levels within 110 infected macrophages, where it can stall cell death pathways, a likely contributory factor in 111 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint granuloma formation (Meconi et al., 2007; Dunne et al., 2013). With a paucity of information 112 regarding the key drivers for the success of infection in the host-pathogen relationship, the 113 treatment of AIEC infection in the context of CD has proved challenging, although recent 114 progress has been made (Boucher and Barnich, 2022; Douadi et al., 2022; Gerner et al., 115 2022; Titécat et al., 2022). However, while AIEC replicates and persists to high levels in 116 some infected macrophages this does not occur in all infected cells. Here we show that the 117 population of AIEC infected macrophages is highly heterogenous, and this is reflected in the 118 vastly different responses of cells to infection. While many cells remain uninfected, or have 119 overcome AIEC infection, these cells remain within the studied in vitro population 120 contributing to outputs and thus disguising the response to infection in cells where AIEC are 121 actively infecting. By stratifying macrophages based on their infectious load, we identified 122 host pathways significantly differentially expressed in direct response to infectious burden, 123 information lost when treating cells as a single homogenous population. By inhibiting the 124 identified differentially expressed pathways, which had not previously been linked to AIEC 125 infection, we could block bacterial intracellular replication and release of the cytokine tumour 126 necrosis factor alpha (TNF), known to be a critical driver of inflammation in both AIEC 127 infection and CD. 128 Our approach here shows that stratifying immune cells based on their bacterial load 129 identifies novel pathways and therapeutic targets not detected using a traditional 130 homogenous population approach. By focusing on host responses directly linked to bacterial 131 success in cells where they are overwhelming the immune response, a more relevant and 132 useful understanding of the complexities of infection can be gained. 133 134

135 Cell culture and infection 136 RAW 264.7 cells were seeded at a density of 2 x 105 cells/ml into a T75 flask with 15 ml of 137 Roswell Park Memorial Institute (RPMI) media (supplemented with 3% foetal bovine serum 138 (FBS), penicillin/streptomycin and L-glutamate). Six hours post-cell seeding, RAW 264.7 139 cells were treated with 100 ng/ml of lipopolysaccharide (LPS) and incubated overnight. RAW 140 264.7 cells in RPMI-1640 with 3% FBS without antibiotics, were then infected with LF82 141 carrying the prpsMGFP plasmid (LF82rpsMGFP) at a multiplicity of infection (MOI) of 100 for 142 1 hour (Li et al., 2022). Post-infection, extracellular bacteria were removed by washing with 143 fresh RPMI media (3% FBS) containing 50 μg/ml of gentamicin and the media was replaced 144 with fresh RPMI media (3% FBS, 50 μg/ml gentamicin). After 24 hours, cells were harvested 145 using cell scrapers. Suspended cells were washed and maintained in fluorescence 146 associated cell sorting (FACS) solution (2% FBS in phosphate buffered saline (PBS)). The 147 viability of cell cultures was assessed using 7-aminoactinomycin D (7-AAD) viability staining 148 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint solution at a final concentration of 0.25 μg/million cells. For each experiment four 149 independent biological replicates were carried out with 4 technical replicates within each. 150 151 Sorting of infected cells 152 Flow cytometry was performed on a BD FACSAria IIU with BD FACSDiva software version 153 9.0.1 (BD Biosciences, Franklin Lakes, NJ) paired with FlowJo Version 6.3.2 analysis 154 software (Tree Star Inc., Ashland, OR). The instrument has not been altered and has a fixed-155 alignment cuvette flow cell and four-laser base configuration. Each sample was first 156 examined using forward scatter (FSC) versus side scatter (SSC). Green fluorescent protein 157 (GFP) was excited by a 488 nm, 20 mW Coherent laser and the emissions detected with a 158 530/30 bandpass filter set while 7-AAD was excited by a 561 nm, 50 mW Coherent laser, 159 and the emissions picked up in the 660/20 Bandpass filter set. Based on measurements 160 obtained from the analysis of 10,000 events for each samples, gating strategies were 161 established for the selection of cells of interest using FSC, SSC, and fluorescence emission 162 properties (Figure S1). Actual cells were easily distinguished from debris by gating on FSC 163 and SSC. 7AAD was used to identify live/dead cells. In LF82 prpsMGFP infected RAW 164 264.7 cells, living cells were gated based on their lack of 7AAD staining. A gating strategy 165 was then established for the three populations of infected cells by determining their GFP 166 fluorescence intensity. The identification of different intracellular bacterial burdens as No, 167 Low and High, were used to sort the cells into three separate populations, representing cells 168 with no bacteria (No), cells with less than 5 bacteria (Low) and cells with more than 5 169 bacteria (High). The control group cells, where bacteria had not been added, were sorted in 170 the same number as for the other three groups. Data was acquired for each population for 171 80,000 cells. To simplify the description of the four groups of cells in the following text, the 172 terms "Control", "No", "Low" and "High" will be used to indicate their infection status. Sorted 173 cells were collected into 1.5 ml microfuge tubes containing 800 μl of RNAlater solution 174 (Invitrogen AM7020) stopping cellular transcriptional changes. RNA from four independent 175 biological repeats was collected and kept at -80°C until RNA was extracted. 176 177 RNA isolation 178 RNA was extracted using an RNeasy PowerMicrobiome Kit (QIAGEN, 26000-50) using the 179 manufacturer’s protocol. RNA extracts were kept at -80°C. Both quantity and quality of RNA 180 were assessed by using an Agilent 2100 Bioanalyzer (Agilent Technologies). RNA yields 181 ranged from 3.47 to 18.6 ng/μl. RNA integrity numbers (RIN) of a sample are generated by 182 the 2100 Bioanalyzer to indicate the level of degradation and have been shown to predict 183 gene expression suitability reliably. RIN scores ranged from 8.7 to 10, indicating high-quality 184 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint RNA suitable for gene expression analysis by RNA sequencing (RNA-seq) (Fleige and 185 Pfaffl, 2006). 186 187 Library construction, RNA-seq, and bioinformatics 188 At least 10 ng of RNA was isolated per sample and provided to Glasgow Polyomics 189 (University of Glasgow) for RNA-seq, the generation of cDNA, sequencing, and 190 bioinformatics. The cDNA libraries were created using the Quantseq (FWD) kit from 191 Lexogen. The kit creates a library from the polyA end of transcripts, creating fragments 192 terminating in the polyA sequence and sequencing towards this. The libraries were 193 sequenced at 75bp, paired end, to a mean depth of 10 million reads per sample, using an 194 illumine (NextSeq 2000). The data was quality controlled and aligned using Galaxy (server: 195 http://antioch.tcrc.gla.ac.uk/). Firstly, read quality was explored using FastQC, then trimmed 196 using Trimmomatic (Bolger, Lohse and Usadel, 2014), under default settings. Reads were 197 mapped to the reference genome (GRCm39) and transcriptome (v110) using Hisat2 (Kim et 198 al., 2019), under default settings. Read counts were produced using HTseq-count (Anders, 199 Pyl and Huber, 2015), which were then normalised, and pairwise differential expression 200 calculated using DESeq2 (Team RC., 2014). Searchlight (Cole et al., 2021) was used to 201 explore and visualise the data. Each pertinent pairwise comparison was entered as a DE 202 workflow, with (adjusted p 1). A single MDE workflow was 203 used combining each of No, Low and High vs Control comparisons. For the pathway 204 analysis the KEGG and GO pathway databases were used with (adjusted p < 0.05). 205 206

207 RAW 264.7 cells that had been incubated with LF82rpsmGFP, or control uninfected cells, 208 were subjected to flow cytometry-based cell sorting to isolate cells based on the intensity of 209 green fluorescence and the number of intracellular bacteria enumerated by colony formation 210 unit (CFU) counts. The experimental procedure is outlined in Figure 1a. Based on 211 fluorescence intensity, RAW264.7 cells co-incubated with LF82rpsmGFP led to 3 distinct 212 populations of cells (each population was at least 80,000 cells) (Figure 1b); those that 213 remained uninfected despite being in proximity to LF82rpsmGFP (No), those with a bacterial 214 load with an average of 1-2 bacteria per cell (Low) and those with approximately 7 bacteria 215 per cell (High) (Figure 1c). The control uninfected cells had no contact with LF82rpsmGFP 216 (Control). 217 RNA extraction was carried out from sorted cells and differential expression analysis was 218 undertaken. Each population of Control, No, Low and High RAW 264.7 cells were compared 219 to each other to identify differentially expressed genes (DEGs) between each group (Figure 220 2). Principle component analysis (PCA) clearly showed the cells from the infected population 221 clustering together and away from uninfected cells as expected (Figure 2a). While it was 222 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint clear from the resulting heatmap and a count of significant DEGs that the response in control 223 cells was significantly changed in comparison to cells in proximity to, or with intracellular 224 LF82rpsmGFP, it was also noted that there was a significant change in response between 225 each of the No, Low and High sub-populations of cells (Figure 2b-c). Comparing DEGs 226 between Control cells and those where LF82rpsmGFP were present (No, Low and High), 227 28.8% of significant DEGs were common to all cells from this population (Figure 3a). 228 However, there were also significant changes in responses between the cells in the infected 229 population with 32.2% of DEGs between the infected and uninfected populations unique to 230 the High group, 11.5% of DEGs unique to the Low group, and 8% of DEGs unique to No 231 group (Fig. 3a). This pointed towards a clearly heterogenous population with cells that were 232 uninfected but bystanders to infection of other cells (No group) having their own unique 233 response, acting in a more similar fashion to infected rather than uninfected cells. Pathway 234 analysis was conducted on the 32.2% of DEGs unique to the High and Control sub-235 populations. The outcome clearly identified several pathways associated with the immune 236 response that were activated in the High group, including the nuclear factor NF-kappa B 237 (NF- κB) pathway, while pathways related to the cell cycle were inhibited in the High group 238 (Fig 3b). 239 Analysis of cytokine gene expression again clearly indicated differences between sub-240 populations within the total infected population. While expression of many cytokine-related 241 genes was increased within the infected population, the bystander cells without bacteria (No 242 group) were noted to have lower expression of several related genes (e.g. TNF receptor: 243 Tnfrsf1b), while other genes were expressed at similar levels to those cells with High 244 bacterial load (e.g. Tnf; Figure 4). Therefore, these bystander cells, were clearly contributing 245 to inflammation through cytokine production but were not being influenced to the same 246 extent by circulating cytokines such as TNF. 247 Having determined both intracellular LF82rpsmGFP numbers and gene expression in 248 response to intracellular bacterial load, we used this data to determine signatures of host 249 gene expression in response to LF82rpsmGFP and identify host pathways that were 250 expressed or repressed in response to infectious burden (Figure 5). As pathway analysis 251 was not possible in the context of three pairwise comparisons (No vs Low, No vs High and 252 Low vs High) due to their low number of significant DEGs, this approach of determining 253 signatures of infection allowed us to extract valuable information related to infection status 254 and drivers of increased infectious burden. We could therefore move past the simple 255 comparison of DEGs in the context of infected versus uninfected cells and examine 256 significant DEGs in the context of the heterogenous AIEC infected population. Two 257 signatures of infection were tested, Signature 1 selected for significantly increased DEGs 258 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint that increased stepwise in direct response to increasing LF82rpsmGFP burden. 516 genes 259 fitting these criteria (Fig. 5a). Signature 2 selected for significantly increased DEGs that had 260 an inverse relationship with intracellular LF82rpsmGFP burden, their expression decreasing 261 as bacterial burden increased, 222 genes fitted these criteria (Fig. 5b). Signature 1 clearly 262 showed that, as bacterial numbers increased, there was a corresponding increase in 263 pathways related to inflammation, chemotaxis and response to bacterial stimuli (Fig. 5a, 264 Table 1). Signature 2 showed that increasing intracellular LF82rpsmGFP load was inversely 265 related to pathways for RNA metabolism, ribosome assembly and cell differentiation, all of 266 which were significantly lower in cells with higher intracellular bacterial loads (Fig. 5b, Table 267 2). 268 To further investigate the importance of these pathways to LF82 infection several significant 269 DEGs were selected from the highlighted Signature 1 pathways with each DEG showing the 270 Signature 1 stepwise increase in expression correlating with increased intracellular 271 LF82rpsmGFP burden (Fig. 6, Table 3). Chemical inhibitors were identified for a number of 272 these Signature 1 gene products that could be used to test their role in LF82rpsmGFP 273 infection; ST034307 for Adcy1, clomipramine for Itch, trametinib for Map2k1, 274 necrosulforamide for Mlkl, and GSK2636771 for Pik3cb (Table 4). Importantly none of the 275 selected inhibitors had previously been tested in the context of either AIEC infection or CD. 276 RAW 264.7 cells were again exposed to LF82rpsmGFP and phagocytosis was allowed to 277 occur prior to treatment to prevent any inhibition of pathogen uptake influencing the results. 278 Each of ST034307, clomipramine and GSK2636771 were seen to influence intracellular 279 bacterial burden at 24 hours post infection (hpi) (Fig. 7ab and 7e). While the inhibitors were 280 not toxic to bacteria during growth, it was clear that at high concentrations certain inhibitors 281 were cytotoxic to cells (Fig. S2). So, while ST034307 inhibition of Adcy1 function 282 significantly decreased intracellular LF82rpsmGFP at 24 hpi, it was seen to induce increases 283 in cytotoxicity when used at the effective 10 M concentration, with this increase in 284 cytotoxicity becoming significant upon infection. Clomipramine was determined to exhibit the 285 most significant effects, reducing intracellular bacterial burden 3 log-fold (Fig. 7b). While 286 clomipramine exhibited some cytotoxicity this was at a higher concentration than those that 287 reduced intracellular bacterial load (Fig. S2). However, to rule out any cytotoxic effects on 288 bacterial load, a reduced 1 M concentration of clomipramine was tested over a longer time 289 course (72 hpi) and the intracellular bacterial burden of live cells determined. Clomipramine 290 was observed to significantly reduce both the number of infected cells and the intracellular 291 bacterial burden in the remaining infected cells (Fig. 8). This effect of clomipramine was 292 observed at 24 hpi (Fig. 8a) and continued over 48 (Fig. 8b) and 72 hpi (Fig. 8c) with the 293 number of infected cells reducing by half and the number of cells with a High bacterial 294 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint burden reducing by over two-thirds. No changes in bacterial burden were observed with the 295 other inhibitors. 296 297 Given pathway analysis using Signature 1 had highlighted a significant role for inflammation 298 and migration of immune cells in response to increasing bacterial burden we next 299 determined any effects of the identified inhibitors on TNF release by the infected cells. 300 TNF levels were determined post-infection and treatment with the inhibitors (Fig. 9). 301 Trametinib significantly inhibited TNF release by both infected and uninfected cells at both 302 100 nM and 1 M with the reduction apparent at early (6 hpi) and later points of infection (24 303 hpi). While ST034307 inhibited TNF, the reduction was only apparent at a higher cytotoxic 304 concentrations of the inhibitor (10 M) in infected cells (Fig. 9b and 9d). Most interestingly 305 however it is noticeable that trametinib, while it significantly reduced TNF release by 306 infected cells, did not reduce intracellular LF82 burden. Contrastingly clomipramine, which 307 reduced intracellular LF82 burden, had no impact on TNF release. This disconnect 308 between AIEC infection and cytokine release by immune cells has not previously been 309 described and may offer future opportunities for intervention to prevent inflammation despite 310 bacterial burdens. 311 312

313 Bacterial infection, both in vitro and in vivo, results in a heterogenous population of cells, 314 comprising those infected to differing levels by the pathogen, and those that remain 315 uninfected, termed bystander cells. The diversity of outcomes at the cellular level presents a 316 conundrum as regards studying infection, as the mixed population can have an array of 317 bacterial burdens resulting in diverse host and microbial gene expression. This 318 heterogeneity makes interpretation of the host response particularly difficult as it can mask 319 crucial host mediators of infection. 320 Here we demonstrate this heterogeneity within an LF82-treated well of RAW 264.7 cells in 321 vitro. Close to 60% of cells carry intracellular LF82 at 24 hpi, but this results in any 322 subsequent analysis of host gene expression in response to infection including the 323 remaining 40% of cells that are uninfected. Even within the LF82-bearing cells our data 324 demonstrates that two thirds of these cells contain less than 5 bacteria, within only 20% of 325 the total population of cells bearing more than 5 bacteria. Given that intracellular replication 326 in immune cells has been described as a critical phenotypic marker of this pathobiont, the 327 fact that only one fifth of the infected population of cells meet this criterium makes it 328 challenging to study (Bringer et al., 2012). Any host transcriptional changes in response to 329 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint intracellular replication of AIEC will be difficult to pick up in downstream analysis due to 330 being masked by the transcriptional changes in the remaining 80% of cells. 331 To overcome the challenges of a heterogeneously infected population, here we took an 332 approach of cell sorting based on intracellular bacterial load followed by RNA sequencing. 333 This enabled us to stratify the heterogenous population into distinct population subsets, each 334 with its own characteristics of being uninfected or infected and, if infected, stratified into 335 further sub-populations based on their intracellular LF82 burden. Clearly there were 336 significant differences between cells exposed to LF82 and unexposed and uninfected cells. 337 Surprisingly bystander cells from wells where LF82 was present, but with no intracellular 338 LF82, displayed a phenotypic shift that mirrored that of infected cells, and which was distinct 339 from cells from uninfected wells. Over 400 genes were significantly differentially regulated 340 between both these uninfected populations, with 77 of the DEGs from these bystander cells 341 unique to them and not identified in infected cells from the same well. This indicated that 342 while also responding to LF82 in a manner similar to infected cells, these bystander cells 343 were a unique population in themselves. 344 Interestingly our analyses also indicated that uninfected bystander cells were directly 345 contributing to inflammation despite not being actively infected with AIEC. It is likely that 346 immune activation of these bystander cells is driven by either contact with bacteria, bacteria-347 derived molecules and vesicles being shed into the media, or immune cell derived TNF 348 (Bringer et al., 2012; Jung et al., 2017; Qu, Zhu and Zhang, 2022). However, the relative 349 contribution of each to bystander cell activation cannot be ascertained from the data 350 generated here, but understanding this could be informative in the context of CD given the 351 importance of TNF in driving inflammation in CD. If bystander cell activation, and their 352 subsequent contribution to inflammation, was dependent on TNF, anti-TNF therapy as 353 used in CD would block activation of these cells. Also as immune cells with a high 354 intracellular bacterial burden are likely to be the primary source of TNF sparking a 355 subsequent inflammatory cascade, targeting this small population of highly infected cells to 356 remove AIEC would offer most therapeutic benefit. 357 Differences in cells either exposed to, or infected with, LF82 were further underlined through 358 direct comparison of gene expression amongst these groups. Again, unique DEGs were 359 found for each group with some DEGs common to more than one sub-population within an 360 infected well. A major advantage of our approach was the ability to directly match host gene 361 expression to bacterial load across the stratified sub-populations. Given the depth of data 362 available, and the significant number of DEGs identified, an approach of enrichment analysis 363 whereby signatures of gene expression were correlated to bacterial load was undertaken. 364 This approach identified DEGs and pathways directly responding to increasing or decreasing 365 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint intracellular bacterial load. While gene expression may fluctuate due to bacterial load, using 366 signatures of infection across populations allowed us to concentrate on DEGs whose 367 expression was directly related to infectious burden. This approach of identifying signatures 368 of gene expression in response to intracellular infectious load revealed several pathways 369 related to increasing or decreasing bacterial load. Given their likely importance to success of 370 infection we targeted these pathways using chemical inhibitors, selecting target proteins 371 from the significant DEGs within these pathways of interest. This enabled testing their role in 372 mediating both intracellular replication of AIEC and its induction of inflammation. 373 The targets chosen; Adcy1, Pik3cb, Mlkl, Map2k1 and Itch, each represented a unique 374 pathway in which they displayed the Signature 1 phenotype of increasing in direct response 375 to bacterial burden within the cell. None of these genes had to date been associated with 376 AIEC infection or used as a target to inhibit bacterial infection, although PIK3cb and MLKL 377 had previously been suggested as targets for therapeutic intervention in IBD, while MAP2K1 378 has a currently approved kinase inhibitor targeted towards it for IBD treatment 379 (Pierdomenico et al., 2014; Bruckner et al., 2020; Winkelmann et al., 2021). ITCH has been 380 directly implicated in pathogenesis of nucleotide-binding oligomerization domain-containing 381 protein 2 (NOD2) mediated inflammatory disease and it is directly involved in ubiquitination 382 and tagging of host proteins for proteasomal degradation, a system we have previously 383 shown to be exploited during AIEC infection (Dunne et al., 2013). Concentrations of each 384 inhibitor used were those previously published in the literature although it was noted that 385 some caused increased cytotoxicity during testing on RAW 264.7 cells, and this was 386 exacerbated by infection in the case of the Adcy1 inhibitor ST034307 (Břehová et al., 2021). 387 Inhibition of Adcy1, Mlkl or Pik3cb function had no significant effect on LF82 infection over 388 the time tested, with no reduction in either intracellular bacterial burden or release of 389 inflammatory cytokines. However, the inhibitor of Itch, trametinib, alongside the inhibitor of 390 Map2K1, clomipramine, generated intriguing results. While clomipramine significantly 391 reduced both intracellular burden of LF82 and the number of cells infected with LF82, 392 trametinib significantly inhibited TNF release. Intriguingly in the case of both inhibitors, they 393 decoupled intracellular proliferation and cytokine release which have been shown to be 394 interdependent during AIEC infection (Bringer et al., 2012; Douadi et al., 2022). Kinase 395 inhibitors such as trametinib can block cell proliferation, arrest the cell cycle and induce cell 396 death as well as blocking extracellular signal-regulated kinase (ERK) signalling, which plays 397 a role in cytokine secretion during AIEC infection (Hedl and Abraham, 2012; Hoffner, MSN, 398 ANP-BC, AOCNP and Benchich, MSN, NP-C, AOCNP, 2018). Given proliferation of infected 399 cells is unlikely as cell cycle arrest is already occurring during LF82 infection based on the 400 Signature 2 pathways identified, the reduction in TNF secretion observed is likely due to 401 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint trametinib interruption of signalling pathways, such as that controlled by ERK, upstream of 402 TNF release. 403 The mechanism of action of clomipramine in the context of LF82 infection was more 404 challenging to interpret. Used to treat obsessive compulsive disorder, clomipramine effects 405 are likely mediated through reducing re-uptake of norepinephrine and serotonin. However, it 406 has recently been used to treat both viral and parasitic infections with a suggested 407 mechanism of action related to its effects on lysosomal pH undermining viral protease 408 efficacy (Vater et al., 2017; Nobile et al., 2020; Strauss et al., 2021; Khan et al., 2022). With 409 lysosomal defence integral to combatting AIEC infection this may explain the phenotype 410 observed here (Spalinger et al., 2022). Clomipramine effect on intracellular LF82 replication 411 was clear cut, significantly reducing both the intracellular bacterial load within cells and the 412 number of cells carrying bacteria. Most strikingly, given previous work describing how LF82 413 intramacrophage replication and TNF release were intertwined, this reduction in LF82 414 numbers showed no effect on TNF release. This disconnect between TNF mediated 415 inflammation and AIEC intracellular replication, which to now have described as mutually 416 dependent, may help in unravelling the complex host-AIEC relationship. 417 The data presented here therefore clearly demonstrates that stratifying infected populations 418 of immune cells into distinct sub-populations based on their bacterial load can reveal new 419 therapeutic targets in infection. Here this approach has shed light on tackling a crucial 420 population of inflammatory immune cells in CD, those heavily infected with AIEC. This 421 targeted approach is relatively simplistic but clearly showed promise, with chemical inhibition 422 of target genes either blocking intracellular replication or reducing secretion of TNF. This is 423 the first time an approach has specifically targeted and been effective against heavily AIEC 424 infected immune cells. 425 426

427 The authors gratefully acknowledge the Sii-Flow Cytometry Core Facility at the University of 428 Glasgow for their support & assistance in this work. 429 430

431 Anders, S., Pyl, P. T. and Huber, W. (2015) ‘HTSeq--a Python framework to work with high-432 throughput sequencing data’, Bioinformatics (Oxford, England). Bioinformatics, 31(2), pp. 433 166–169. doi: 10.1093/BIOINFORMATICS/BTU638. 434 Bassi, A. et al. (2004) ‘Cost of illness of inflammatory bowel disease in the UK: a single 435 centre retrospective study’, Gut. BMJ Publishing Group, 53(10), pp. 1471–1478. doi: 436 10.1136/GUT.2004.041616. 437 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Bolger, A. M., Lohse, M. and Usadel, B. (2014) ‘Trimmomatic: a flexible trimmer for Illumina 438 sequence data’, Bioinformatics (Oxford, England). Bioinformatics, 30(15), pp. 2114–2120. 439 doi: 10.1093/BIOINFORMATICS/BTU170. 440 Boucher, D. and Barnich, N. (2022) ‘Phage Therapy Against Adherent-invasive E. coli: 441 Towards a Promising Treatment of Crohn’s Disease Patients?’, Journal of Crohn’s & colitis. 442 NLM (Medline), 16(10), pp. 1509–1510. doi: 10.1093/ecco-jcc/jjac070. 443 Břehová, P. et al. (2021) ‘Acyclic nucleoside phosphonates with 2-aminothiazole base as 444 inhibitors of bacterial and mammalian adenylate cyclases’, European Journal of Medicinal 445 Chemistry, 222. doi: 10.1016/j.ejmech.2021.113581. 446 Bringer, M. A. et al. (2012) ‘Replication of Crohn’s disease-associated AIEC within 447 macrophages is dependent on TNF-α secretion’, Laboratory Investigation. Lab Invest, 92(3), 448 pp. 411–419. doi: 10.1038/labinvest.2011.156. 449 Bruckner, R. S. et al. (2020) ‘DOP86 An IFN-STAT1-MLKL axis drives programmed necrosis 450 of Paneth cells in Crohn’s ileitis’, Journal of Crohn’s and Colitis. Oxford Academic, 451 14(Supplement_1), pp. S125–S126. doi: 10.1093/ECCO-JCC/JJZ203.125. 452 Cho, Y. H. et al. (2022) ‘Inflammatory bowel disease-associated adherent-invasive 453 Escherichia coli have elevated host-defense peptide resistance’, FEMS microbiology letters. 454 FEMS Microbiol Lett, 369(1). doi: 10.1093/FEMSLE/FNAC098. 455 Cole, J. J. et al. (2021) ‘Searchlight: automated bulk RNA-seq exploration and visualisation 456 using dynamically generated R scripts’, BMC Bioinformatics. BioMed Central Ltd, 22(1), pp. 457 1–21. doi: 10.1186/S12859-021-04321-2/TABLES/1. 458 Darfeuille-Michaud, A. et al. (1998) ‘Presence of adherent Escherichia coli strains in ileal 459 mucosa of patients with Crohn’s disease.’, Gastroenterology, 115(6), pp. 1405–1413. 460 Douadi, C. et al. (2022) ‘Anti-TNF Agents Restrict Adherent-invasive Escherichia coli 461 Replication Within Macrophages Through Modulation of Chitinase 3-like 1 in Patients with 462 Crohn’s Disease’, Journal of Crohn’s & colitis. J Crohns Colitis, 16(7), pp. 1140–1150. doi: 463 10.1093/ECCO-JCC/JJAB236. 464 Dunne, K. A. et al. (2013) ‘Increased S-nitrosylation and proteasomal degradation of 465 caspase-3 during infection contribute to the persistence of adherent invasive Escherichia coli 466 (AIEC) in immune cells.’, PloS one, 8(7), p. e68386. doi: 10.1371/journal.pone.0068386. 467 Fleige, S. and Pfaffl, M. W. (2006) ‘RNA integrity and the effect on the real-time qRT-PCR 468 performance’, Molecular aspects of medicine. Mol Aspects Med, 27(2–3), pp. 126–139. doi: 469 10.1016/J.MAM.2005.12.003. 470 Gerner, R. R. et al. (2022) ‘Siderophore Immunization Restricted Colonization of Adherent-471 Invasive Escherichia coli and Ameliorated Experimental Colitis’, mBio. mBio, 13(5). doi: 472 10.1128/MBIO.02184-22. 473 Hedl, M. and Abraham, C. (2012) ‘Nod2-Induced Autocrine Interleukin-1 Alters Signaling by 474 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint ERK and p38 to Differentially Regulate Secretion of Inflammatory Cytokines’, 475 Gastroenterology. NIH Public Access, 143(6), p. 1530. doi: 476 10.1053/J.GASTRO.2012.08.048. 477 Hoffner, MSN, ANP-BC, AOCNP, B. and Benchich, MSN, NP-C, AOCNP, K. (2018) 478 ‘Trametinib: A Targeted Therapy in Metastatic Melanoma’, Journal of the Advanced 479 Practitioner in Oncology. Harborside Press, 9(7), p. 741. doi: 10.6004/jadpro.2018.9.7.5. 480 Jung, A. L. et al. (2017) ‘Legionella pneumophila infection activates bystander cells 481 differentially by bacterial and host cell vesicles’, Scientific Reports. Springer US, 7(1), pp. 1–482 11. doi: 10.1038/s41598-017-06443-1. 483 Khan, E. et al. (2022) ‘Anti-depressants and COVID-19: A New Ray of Hope’, Psychiatria 484 Danubina. Psychiatr Danub, 34(1), pp. 171–173. doi: 10.24869/PSYD.2022.171. 485 Kim, D. et al. (2019) ‘Graph-based genome alignment and genotyping with HISAT2 and 486 HISAT-genotype’, Nature biotechnology. Nat Biotechnol, 37(8), pp. 907–915. doi: 487 10.1038/S41587-019-0201-4. 488 Li, X. et al. (2022) ‘Inhibition of proline tyrosine kinase 2 (Pyk2) phosphorylation during 489 adherent-invasive Escherichia coli infection inhibits intra-macrophage replication and 490 inflammatory cytokine release’, bioRxiv. Cold Spring Harbor Laboratory, p. 491 2022.11.14.516411. doi: 10.1101/2022.11.14.516411. 492 Martinez-Medina, M. et al. (2009) ‘Molecular diversity of Escherichia coli in the human gut: 493 New ecological evidence supporting the role of adherent-invasive E. coli (AIEC) in Crohn’s 494 disease’, Inflammatory Bowel Diseases, 15(6), pp. 872–882. doi: 10.1002/ibd.20860. 495 Meconi, S. et al. (2007) ‘Adherent-invasive Escherichia coli isolated from Crohn’s disease 496 patients induce granulomas in vitro’, Cellular Microbiology, 9(5), pp. 1252–1261. doi: 497 10.1111/j.1462-5822.2006.00868.x. 498 Nadalian, B. et al. (2021) ‘Prevalence of the pathobiont adherent-invasive Escherichia coli 499 and inflammatory bowel disease: a systematic review and meta-analysis’, Journal of 500 gastroenterology and hepatology. J Gastroenterol Hepatol, 36(4), pp. 852–863. doi: 501 10.1111/JGH.15260. 502 Nobile, B. et al. (2020) ‘Clomipramine Could Be Useful in Preventing Neurological 503 Complications of SARS-CoV-2 Infection’, Journal of neuroimmune pharmacology : the official 504 journal of the Society on NeuroImmune Pharmacology. J Neuroimmune Pharmacol, 15(3), 505 pp. 347–348. doi: 10.1007/S11481-020-09939-2. 506 Ormsby, M. J. et al. (2019) ‘Inflammation associated ethanolamine facilitates infection by 507 Crohn’s disease-linked adherent-invasive Escherichia coli’, EBioMedicine. Elsevier B.V., 43, 508 pp. 325–332. doi: 10.1016/j.ebiom.2019.03.071. 509 Ormsby, M. J. et al. (2020) ‘Propionic Acid Promotes the Virulent Phenotype of Crohn’s 510 Disease-Associated Adherent-Invasive Escherichia coli’, Cell Reports, 30(7). doi: 511 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint 10.1016/j.celrep.2020.01.078. 512 Pierdomenico, M. et al. (2014) ‘Necroptosis is active in children with inflammatory bowel 513 disease and contributes to heighten intestinal inflammation’, The American journal of 514 gastroenterology. Am J Gastroenterol, 109(2), pp. 279–287. doi: 10.1038/AJG.2013.403. 515 Qu, M., Zhu, H. and Zhang, X. (2022) ‘Extracellular vesicle-mediated regulation of 516 macrophage polarization in bacterial infections’, Frontiers in Microbiology, 13. doi: 517 10.3389/fmicb.2022.1039040. 518 Rao, B. B. et al. (2017) ‘The Cost of Crohn’s disease: varied healthcare expenditure patterns 519 across distinct disease trajectories’, Inflammatory bowel diseases. NIH Public Access, 23(1), 520 p. 107. doi: 10.1097/MIB.0000000000000977. 521 Spalinger, M. R. et al. (2022) ‘Original research: Autoimmune susceptibility gene PTPN2 is 522 required for clearance of adherent-invasive Escherichia coli by integrating bacterial uptake 523 and lysosomal defence’, Gut. BMJ Publishing Group, 71(1), p. 89. doi: 10.1136/GUTJNL-524 2020-323636. 525 Strauss, M. et al. (2021) ‘Differential tissue distribution of discrete typing units after drug 526 combination therapy in experimental Trypanosoma cruzi mixed infection’, Parasitology. 527 Parasitology, 148(13), pp. 1595–1601. doi: 10.1017/S0031182021001281. 528 Sugihara, K. et al. (2022) ‘Mucolytic bacteria license pathobionts to acquire host-derived 529 nutrients during dietary nutrient restriction’, Cell reports. Cell Rep, 40(3). doi: 530 10.1016/J.CELREP.2022.111093. 531 Titécat, M. et al. (2022) ‘Safety and Efficacy of an AIEC-targeted Bacteriophage Cocktail in a 532 Mice Colitis Model’, Journal of Crohn’s & colitis. J Crohns Colitis, 16(10), pp. 1617–1627. 533 doi: 10.1093/ECCO-JCC/JJAC064. 534 Vater, M. et al. (2017) ‘New insights into the intracellular distribution pattern of cationic 535 amphiphilic drugs’, Scientific Reports. Nature Publishing Group, 7. doi: 536 10.1038/SREP44277. 537 Winkelmann, P. et al. (2021) ‘The PI3K pathway as a therapeutic intervention point in 538 inflammatory bowel disease’, Immunity, inflammation and disease. Immun Inflamm Dis, 9(3), 539 pp. 804–818. doi: 10.1002/IID3.435. 540 541 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Figures 542 Figure 1 543 544 Figure 1: Macrophage sub-populations sorted by FACS and confirmation of 545 intracellular bacteria number by traditional visible colony count. (a) Schematic 546 overview of the process of sorting RAW 264.7 cells for RNA-seq and viable count analysis. 547 There were 4 independent biological repeats, each repeat includes two sorts: one was 548 sorted into an RNAlater solution, enabling later RNA extraction; another sort was used for 549 confirming the number of intracellular bacteria. (b) Three populations of cells were 550 determined according to GFP intensity. (c) The number of intracellular bacteria from different 551 populations was calculated after their recovery by plating it onto LB agar plate and CFU 552 counting. 553 554 555 556 557 558 559 560 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Figure 2561 562 Figure 2: Sample clustering and differentially expressed genes (DEGs) between 563 different macrophage populations with differing bacterial burdens. (a) Principal 564 component analysis (PCA) of expression data, the first two components. Dots represent 565 replicates and are coloured by condition (red=Control, green=High, blue=Low, purple=No). 566 The % variance is given. (b) Expression heatmap of all DEGs (adjusted p 1) in any of 6 comparisons (Control vs No, Control vs Low, 568 Control vs High, No vs Low, No vs High, Low vs High). Axis are hierarchically clustered. 569 Expression values are per gene Z-scores with low=blue and high=red. 570 571 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Figure 3 572 573 Figure 3: Characteristics of unique DEGs in the comparison of Control vs High. 574 (a) Venn diagram showing the number of overlapped or unique DEGs (adjusted p 1) in the three comparisons: Control vs No, Control vs Low 576 and Control vs High. There are 310 unique DEGs in the comparison of Control vs High. (b) 577 Heatmap of enriched KEGG pathways (adjusted p < 0.05) for the 310 unique genes in (a). 578 The heatmap shows mean expression across all genes in the enriched pathways, with the 579 rows being pathways and columns individual samples. Red indicates relative pathway 580 activation and green represents relative pathway suppression. 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Figure 4 599 600 Figure 4: Heatmap of changes in gene expression levels of cytokine and chemokine 601 genes in three groups infected with LF82 (No, Low and High) alongside the Control 602 uninfected group. Heatmap of differentially expressed cytokines (adjusted p 1) between each of No, Low and High vs Control. Rows represent 604 cytokines and columns samples. The y-axis is hierarchically clustered. Expression values 605 are per gene Z-scores with low=blue and high=red. 606 607 608 609 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Figure 5 610 611 Figure 5: Signature gene expression among 4 populations and their relevant enriched 612 GO-BP pathways. (a) Signature analysis for genes that are elevated (adjusted p 613 1) in all groups (No, Low, High) vs control. Showing: (left) metagene 614 violin plot, with the mean expression z-score on the y-axis and group on the y-axis; (right) 615 expression heatmap for all genes in the signature, showing genes by row and samples by 616 column. The y-axis is hierarchically clustered. Expression values are per gene Z-scores with 617 low=blue and high=red; (bottom) ten most enriched GO biological processes (p.adj < 0.05) 618 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint for the signature genes. Showing the -log10p adjust value on the x axis and the number of 619 DEGs in each enriched pathway as the data label. (b) as (a) however for the genes that are 620 downregulated (adjusted p < 0.05 log2fold < -1) in all groups (No, Low, High) vs control. 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Figure 6 656 657 Figure 6: Gene expression levels of five candidate host DEGs selected for further 658 testing. Genes Adcy1, Pik3cb, Mlkl, Map2k1 and Itch were selected from the signature 1 659 gene list involved in pathways; cell-cell adhesion, TNF signalling, necrotic cell death, MAPK 660 pathways and NF-κB. Boxplots show expression of genes of interest in four groups: Control 661 in red, No in green, Low in blue and High in purple. Black dots denote individual samples. 662 Error bars represent SEM. 663 664 665 666 667 668 669 670 671 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Figure 7 672 673 Figure 7: Evaluation of the effects of different chemical inhibitors on intracellular 674 bacterial load in RAW 264.7 cells. RAW 264.7 cells were infected with LF82 for 1 hour 675 followed by treatment with different chemical inhibitors for a further 6 or 24 hpi; ST034307 676 (a), Clomipramine (b), Necrosulforamide (c), Trametinib (d), GSK2636771 (e). Bacterial 677 recovery is displayed as CFU/g of protein. Data points represent the mean of three technical 678 repeats plus the standard deviation at a timepoint of 6 or 24 hpi. Each treatment was 679 compared to the untreated control group. Statistical significance was determined by two-way 680 ANOVA. *, P <0 0.05. **, P < 0.01. ***, P < 0.0001. 681 682 683 684 685 686 687 688 689 690 691 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Figure 8 692 693 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Figure 8: Quantification of intracellular LF82 burden post-inhibitor treatment using 694 imaging flow cytometry. RAW 264.7 cells infected with LF82::rpsMGFP were treated with 695 1 μM ST034307, 1 uM GSK2636771, 1 uM Necrosulforamide, 100 nM Trametinib or 1 uM 696 Clomipramine for 24 hpi (a), 48 hpi (b) and 72 hpi (c). Infected cells treated with DMSO were 697 used as a control. Intracellular LF82::rpsMGFP was counted via IFC. The spot count profile 698 separated cells into those with no bacteria, cells containing 1-5 bacteria, or cells containing 699 over 5 bacteria. The sub-populations of a graph represent the mean of three biological 700 repeats. Error bars represent SEM. The number of portions of sub-populations represents 701 the mean of three biological repeats. 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint 731 732 Figure 9 733 734 Figure 9: TNFα secretion by RAW 264.7 cells measurement post-inhibitor treatment. 735 RAW 264.7 cells were stimulated overnight by 100 ng/ml LPS. Activated RAW 264,7 cells 736 were then infected with LF82 at MOI of 100 or treated with bacteria-free medium (as 737 uninfected RAW 264.7 cells) for 1 hour. One hour post-infection, infected or uninfected RAW 738 264.7 cells were washed and treated with different chemical inhibitors at two different 739 concentrations for further indicated times. Infected or uninfected cells in absence of chemical 740 treatment was regarded as a control. Graph (a) represents uninfected RAW 264.7 cells that 741 were treated with or without chemical treatments for 6 hpi. (b) As described for (a) but for 742 infected RAW 264.7 cells. (c) Uninfected RAW 264.7 cells were treated with chemical 743 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint inhibitors for 24 hpi. (d) As for (c) but for infected RAW 264.7 cells. Statistical significance 744 was determined by one-way ANOVA. *, P <0 0.05. **, P < 0.01. ***, P < 0.0001. 745 746 747 Tables 748 Table 1 749 Term P value p.BH value GO Regulation of nitric oxide biosyntheic process 1.07E-08 1.35E-06 GO Positive regulation of cell adhesion 5.89E-08 5.41E-06 GO Regulation of NF Kappab import into necleus 7.08E-07 3.76E-06 GO Phagocytosis 3.39E-06 1.31E-04 GO Apoptotic signaling pathway 7.77E-06 2.48E-04 GO Positive regulation of phagocytosis 1.28E-05 3.59E-04 GO Regulation of tumor necrosis factor biosynthetic process 2.72E-05 6.46E-04 GO Regulation of calcium ion transport 3.65E-05 8.34E-04 GO Protein autophosphorylation 1.04E-04 1.85E-03 GO Regulation of fatty acid transport 1.85E-04 2.90E-03 GO Negative regulation of meiotic cell cycle 5.07E-03 3.63E-02 Table 1: Go enrichment analysis for signature 1 genes (516 genes) in bp terms. 750 751 752 753 754 Table 2 755 Term P value p.BH value GO rRNA metobolic process 8.96E-10 3.62E-06 GO Ribosome assembly 2.38E-08 4.81E-05 GO DNA Conformation change 2.18E-05 1.26E-02 GO DNA packaging 6.05E-05 2.44E-02 GO Cellular amino acid metabolic process 6.87E-05 2.52E-02 GO Alpha amino acid metabolic process 1.36E-04 4.30E-02 756 Table 2: GO enrichment analysis for signature 2 genes (222 genes) in BP terms. 757 758 759 760 761 762 763 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Table 3 764 Gene Symbo l Protein name Prosed Functions Involved both GObp and KEGG Pathways Adcy1 Adenylate Cyclase 1 Catalyses the formation of the signalling molecule cAMP in response to G-protein signalling Calcium signalling pathway Bcl3 B-Cell Lymphoma 3-Encoded Protein the regulation of transcriptional activation of NFκB target genes TNF signalling pathways Ccl2 C-C Motif Chemokine Ligand 2 mobilization of intracellular calcium ions Phagocytosis; Calcium signalling pathway Cd44 CD44 Molecule Cell-surface receptor that plays a role in cell-cell interactions, cell adhesion and migration Cell adhesion Hif1A Hypoxia Inducible Factor 1 Subunit Alpha transcriptional regulator of the adaptive response to hypoxia Autophagy Itch Itchy E3 Ubiquitin Protein Ligase targeting specific proteins for lysosomal degradation TNF signalling pathways, Apoptosis Lyn Src Family Tyrosine Kinase the regulation of innate and adaptive immune responses Calcium signalling pathway Map2k 1 Mitogen-Activated Protein Kinase Kinase 1 Involvement in the ERK pathway by activation of ERK1 and ERK2 Cell adhesion; TNF signalling pathways Mlkl mixed lineage kina se domain-like key role in TNF-induced necroptosis, a programmed cell death process Apoptosis; TNF signalling pathways Myl2 Myosin Light Chain 2 plays a role in heart development and function Cell adhesion Pik3cb Phosphatidylinosito l-4,5-Bisphosphate 3-Kinase Catalytic Subunit Beta activation pathway in neutrophils Cell adhesion; TNF signalling pathways; Autophagy Ptpn2 Protein Tyrosine Phosphatase Non- Receptor Type 2 Regulate cell growth, differentiation, mitotic cycle, and oncogenic transformation Cell adhesion Vegfa Vascular Endothelial Growth Factor A proliferation and migration of vascular endothelial cells Cell adhesion; Calcium signalling pathway Table 3: Gene of interests from Signature 1. 765 766 767 768 769 770 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Table 4 771 Targeting gene Chemical inhibitor Reference Adcy1 ST034307 Watts, V.J., 2018 Mlkl Necrosulfonamide Rübbelke, M., 20201 Map2k1 Trametinib (GSK1120212) Khan, Z.M., 2020 Pik3cb GSK2636771 Vanhaesebroeck, B., 2021 Itch Clomipramine Rossi, M., 2014 772 Table 4: Selected Signature 1 genes and relevant chemical inhibitors. 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Supplementary Figures: 801 Figure S1 802 803 Figure S1: Gating strategy for isolation of LF82::rpsMGFP infected RAW 264.7 cells 804 for three different populations (No, Low and High). (a) For the isolation of a highly pure 805 RAW 264.7 population, cells were gated on their forward scatter area (FSC-A) and side 806 scatter area (SSC-A), excluding debris from the live gate. (b) Dead cells were further 807 excluded based on FSC-A versus the intensity of 7AAD. (c) This was followed by gating out 808 three sub-populations of living RAW 264.7 cells according to GFP intensity, resulting in 809 sorting final three populations including cells with no bacterial burden, low bacterial burden 810 and high bacterial burden. 811 812 813 814 815 816 817 818 819 820 821 822 823 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Figure S2 824 825 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint Figure S2: Effects of the different chemical inhibitors on LF82 growth and cytotoxicity 826 to RAW 264.7 cells. (a) Growth curve of LF82 after treatment with DMSO or the five 827 chemical inhibitors (10 μM ST034307, 10 μM GSK2636771, 10 μM Necrosulforamide, 10 μM 828 Trametinib or 10 μM Clomipramine). (b) Lactate dehydrogenase (LDH) cytotoxicity assay 829 was undertaken for uninfected or LF82 infected RAW 264.7 cells treated with different 830 concentrations of chemical inhibitors. DMSO, a diluent for the inhibitors, was used as a 831 control. Experimental groups were compared to the control. Statistical significance was 832 determined by one-way ANOVA. Statistical significance was determined by one-way 833 ANOVA. *, P <0 0.05. **, P < 0.01. ***, P < 0.0001. 834 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 10, 2024. ; https://doi.org/10.1101/2024.01.10.575028doi: bioRxiv preprint