Endoplasmic reticulum oxidoreductin 1‐alpha deficiency and activation of protein translation synergistically impair breast tumour resilience

Background and Purpose Endoplasmic reticulum (ER) stress triggers an adaptive response in tumours which fosters cell survival and resilience to stress. Activation of the ER stress response, through its PERK branch, promotes phosphorylation of the α‐subunit of the translation initiation factor eIF2, thereby repressing general protein translation and augmenting the translation of ATF4 with the downstream CHOP transcription factor and the protein disulfide oxidase, ERO1‐alpha Experimental Approach Here, we show that ISRIB, a small molecule that inhibits the action of phosphorylated eIF2alpha, activating protein translation, synergistically interacts with the genetic deficiency of protein disulfide oxidase ERO1‐alpha, enfeebling breast tumour growth and spread. Key Results ISRIB represses the CHOP signal, but does not inhibit ERO1. Mechanistically, ISRIB increases the ER protein load with a marked perturbing effect on ERO1‐deficient triple‐negative breast cancer cells, which display impaired proteostasis and have adapted to a low client protein load in hypoxia, and ERO1 deficiency impairs VEGF‐dependent angiogenesis. ERO1‐deficient triple‐negative breast cancer xenografts have an augmented ER stress response and its PERK branch. ISRIB acts synergistically with ERO1 deficiency, inhibiting the growth of triple‐negative breast cancer xenografts by impairing proliferation and angiogenesis. Conclusion and Implications These results demonstrate that ISRIB together with ERO1 deficiency synergistically shatter the PERK‐dependent adaptive ER stress response, by restarting protein synthesis in the setting of impaired proteostasis, finally promoting tumour cytotoxicity. Our findings suggest two surprising features in breast tumours: ERO1 is not regulated via CHOP under hypoxic conditions, and ISRIB offers a therapeutic option to efficiently inhibit tumour progression in conditions of impaired proteostasis.

ISRIB offers a therapeutic option to efficiently inhibit tumour progression in conditions of impaired proteostasis.

K E Y W O R D S
breast cancer, endoplasmic reticulum stress, ERO1 alpha, ISRIB (integrated stress response inhibitor), PERK pathway, UPR (unfolded protein response)

| INTRODUCTION
Endoplasmic reticulum (ER) client proteins are folded and post-translationally modified in the ER before being exported in the secretory pathway (Sun & Brodsky, 2019). The high rate of proliferation of cancer cells, together with cancer-associated conditions such as hypoxia and shortage of nutrients, imposes stress on the ER, a process referred to as ER stress, which impairs its ability to fold and export proteins. As a consequence, a plethora of corrective measures are triggered, collectively known as UPR (Unfolded Protein Response), which increases resistance to stress and adaptation, and contributes to the thriving and survival of tumour cells (Cubillos-Ruiz et al., 2017;Fels & Koumenis, 2006;Wang et al., 2012).
UPR, through its activated protein kinase RNA-like endoplasmic reticulum kinase (PERK) branch, promotes the phosphorylation of eukaryotic initiation factor 2 alpha (p-eIF2α) with consequent downregulation of global protein synthesis, thus reducing the protein load of the ER and relieving the stress. However, if protein synthesis restarts under conditions of impaired proteostasis, adaptation is shattered and cell death occurs (Han et al., 2013). Downstream from p-eIF2α, ATF4, a pro-survival factor, is selectively translated and leads to the transcription of genes involved in the ER functions (Guan et al., 2017). It also activates the transcription factor C/EBP homologous protein (CHOP). These two steps of attenuation of protein translation and ATF4 induction are also triggered by other pathways and therefore are part of the integrated stress response (ISR) (Guan et al., 2017).
High levels of ERO1 are associated with different cancers and are predictive of their malignant phenotype and worse clinical outcome (Julian Cornelius et al., 2021;Tanaka et al., 2015;Tanaka et al., 2016;Yang et al., 2018;Zhang et al., 2020;Zhou et al., 2017). Recently, we and others characterized the potential for ERO1 to promote angiogenesis and breast cancer metastasis in hypoxia (Manuelli et al., 2021;May et al., 2005;Tanaka et al., 2016;Varone et al., 2021;Zilli et al., 2021). Our analysis of the secretome of breast tumour cells genetically deleted for ERO1 indicate that ERO1 promotes the secretion of different angiogenic factors, among them the master angiogenic regulator VEGF. Consequently, the inhibition of ERO1 in tumours could be a promising therapeutic strategy to impair angiogenesis and curtail tumour growth and metastasis (Varone et al., 2021).
Unfortunately, the currently available ERO1 inhibitors EN460 and QM295 suffer off-target effects and prevent their use in vivo to test their ability to inhibit breast cancer growth and metastasis Hayes et al., 2019). However, ISRIB, an inhibitor of p-eIF2 alpha activity which rescues the repression of the protein translation, has no off-target effects and its safety profile in preclinical cancer models suggests the possibility of using it in humans (Rabouw et al., 2019;Schoof et al., 2021;Sidrauski et al., 2015).
In this study, we set out to investigate whether ISRIB, by repressing p-eIF2 alpha, which is upstream to ERO1 in the PERK branch of the ER stress response, also inhibits ERO1 activity, hence inhibiting tumour angiogenesis in breast cancer. Surprisingly, although ISRIB inhibited CHOP, it had no direct effect on either ERO1 expression or its angiogenic activity, suggesting that ERO1 expression is not regulated through CHOP in highly metastatic MDAMB231 breast tumours under hypoxic conditions. However, ISRIB, together with the genetic deficiency of ERO1, acts on the inhibition of protein translation, which is an adaptive feature of some tumours, and promotes proteotoxicity, thus synergistically limiting tumour growth.  (Jaffe et al., 1973) and grown in 1% gelatin-coated flasks in M199 supplemented with 10% FBS, 10% newborn calf serum, 20 mM Hepes, 2 mM glutamine, 6 UÁml À1 heparin, 50 μgÁml À1 endothelial cell growth factor, penicillin, and streptomycin. Cells were used between the third and fifth passages.

| Cell culture and transfection
Highly metastatic human MDAMB231* breast cancer cells were transfected with ERO1-Lα CRISPR-Cas9 KO plasmids (SC-401747 for human, Santa Cruz Biotechnology) with three target-specific guide RNAs (gRNA) of 20 nt. The plasmids were co-transfected with homology-directed repair HDR plasmids (SC-401747-HDR for human, Santa Cruz Biotechnology), which led to the insertion of a puromycin resistance gene and a red fluorescent protein (RFP) gene. Wild-type, heterozygous and knock-out clones are analysed by SDS-PAGE and Sanger sequencing. ERO1 knock out (KO) HeLa cells and FLAG-VEGF 121 are described elsewhere (Varone et al., 2021).

| Detergent-insoluble and detergent-soluble VEGF 121
Detergent-insoluble and -soluble VEGF 121 and BIP were prepared as described earlier (Rai et al., 2021). FLAG Immunoblot was used to detect VEGF 121 . BIP was detected by a KDEL antibody.

| Hypoxic chamber
Cells were transferred into a hypoxic chamber (Ruskinn Invivo2 400, UK) at 37 C and maintained in deoxygenated culture medium at the following gas concentrations: O 2 0.1%, CO 2 5% for 48 h. Control cells were maintained in standard culture medium in a normoxic incubator.

| Motility assay
Conditioned media from equal numbers of WT and ERO1 KO MDAMB231* cells were used as an attractant to stimulate HUVEC migration. HUVECs were suspended in DMEM, 0.1% BSA at a concentration of 0, 75 Â 10 6 ml À1 , and added to the upper compartment of Boyden chamber. The assay was carried out in 5% CO 2 at 37 C for 6 h. At the end of the incubation, filters were fixed and stained with Diff-Quik (Marz-Dade, Dundingen, Switzerland) to detect cells adhering to the lower surface. Thereafter, migrated cells were counted in 10 high-power fields for each filter.

| VEGF ELISA
Secreted VEGF was measured in the conditioned media of MDAMB231* cells with human VEGF Quantikine ELISA Kit (DVE00, R&D Systems).

| Real-time quantitative RT-PCR analysis
Total RNA was isolated using the RNeasy Mini Kit (Qiagen) following the manufacturer's instructions. One microgram of total RNA was reverse-transcribed and analysed using the Applied Biosystems' realtime PCR System and the ΔΔCt method. Relative gene expression in cells was normalized to GAPDH or cyclophilin mRNA levels. The primer sequences are described in Varone et al. (2021).

| Animals
Eight-to 10-week-old female SCID mice were obtained from Charles River Laboratories (Calco, Italy) and maintained under specific-pathogen-free conditions. SCID mice were housed in isolated vented cages, and handled using aseptic procedures. Animal studies are reported in compliance with the ARRIVE guidelines Paclitaxel was injected intravenously (IV) at the dose of 15 mgÁkg À1 , Q7x2 (every 7 days for 2 weeks) and stopped 2 weeks before the end of the experiment. ISRIB (trans-isomer, Aurogene s.r.l) was dissolved in 10% DMSO (Sigma-Aldrich) in corn oil. It was injected interaperitoneally (IP), at the dose 2.5 mgÁkg À1 every 2 days for 3 weeks.
Forty-eight hours after the last dose of ISRIB (see Figure 3a, scheme of the pharmacological treatment), mice underwent volatile anaesthesia with isoflurane, analysed by bioluminescence imaging (BLI) and then killed by cervical dislocation. Metastases were quantified by BLI, where, mice injected with D-luciferin (150 mgÁkg À1 IP, Caliper Lifescience) were scanned after 10 min with IVIS Lumina Series III XRMS (Perkin Elmer). Images were analysed with the Living Image software (Perkin Elmer) and the metastasis burden was expressed as total flux (photonsÁs À1 ). The analyses were not blinded for practical constraints, because they were done by the same two researchers who performed the treatments. Primary tumours were randomly selected for further analysis: of which five for RNA sequencing and quantitative real-time and five for histopathological analysis.
2.14 | RNA sequencing RNA was extracted from WT and ERO1 KO (vehicle, paclitaxel, ISRIB and the combination paclitaxel, ISRIB-treated) xenografts (four separate samples for each condition) with the Qiagen RNeasy kit and quantified with Nanodrop; quality was measured using Qubit. The overall quality of sequencing reads was evaluated using FastQC (v.0.11.9). Sequence alignments of total-RNA (stranded) to the reference human genome (GRCh38) were performed using STAR (v.2.7.9a) in two-pass mode. Specifically, gene expression was quantified at the gene level by using the comprehensive annotations made available by Gencode (v38 GTF File). Samples were adjusted for library size and normalized with the variance stabilizing transformation in the R statistical environment using DESeq2 (v1.28.1) pipeline. When performing differential expression analysis between groups, we applied the embedded Independent Filtering procedure to exclude genes that were not expressed at appreciable levels in most of the samples con- For immunohistochemistry (IHC) analysis, paraffin was removed with xylene and the sections were rehydrated in graded alcohol. Antigen retrieval was carried out using a preheated target retrieval solution for 35 min. Tissue sections were blocked with FBS serum in PBS for 60 min and incubated overnight with primary antibody. The antibody binding was detected using a polymer detection kit (GAR-HRP, Microtech) followed by a diaminobenzidine chromogen reaction (Peroxidase substrate kit, DAB, SK-4100; Vector Lab). All sections were counterstained with Mayer's haematoxylin and visualized using a bright-field microscope. Samples were analysed to count the CD31-immunopositive areas.

| Survival
Survival of breast cancer patients was analysed using the KMPlotter tool, which is publicly available at https://kmplot.com/analysis. (PERK) or their ratio (EIF2AK3/ERO1A). Statistical significance was assessed using a log rank test.

| Statistics
The data and statistical analysis comply with the recommendations of the British Journal of Pharmacology on experimental design and analysis in pharmacology (Curtis et al., 2018). Data are the mean ± SD, and were analysed with Prism 7 (GraphPad). Statistical analysis was based on the number of independent samples/experiments (n), as indicated in the figure legends. Statistical significance was evaluated using the unpaired t-test for two-group analysis, one-way ANOVA or two-way ANOVA for multiple comparison tests for three or more group analysis. Significant differences were set to P < 0.05. Results of The increased BIP in ERO1 KO cells under hypoxic conditions suggested ER stress. ER stress triggers a homeostatic response, the so-called UPR, which is involved in cancer thriving. One of the features of this ER stress-related homeostatic response is to promote the attenuation of protein translation via p-eIF2 alpha and through its PERK arm. In hypoxia, ERO1 KO cells had higher p-eIF2 alpha indicating repression of the protein translation ( Figure S1).
The two steps of attenuation of protein translation and selective ATF4 induction also are triggered by other pathways and are therefore part of the integrated stress response (ISR). Recently, ISRIB, an inhibitor of ISR, was shown to rescue p-eIF2 alphamediated attenuation of protein translation (Nguyen et al., 2018;Rabouw et al., 2019;Sidrauski et al., 2015). Furthermore, hypoxia induces a reduction in protein synthesis by activating the ISR arm of the UPR (Leprivier et al., 2015;Wouters & Koritzinsky, 2008). Therefore, to investigate the difference in this signalling between highly metastatic WT and ERO1 KO MDAMB231* breast cancer cells, we analysed their levels of protein translation in a puromycinbased assay (SUnSET) (Schmidt et al., 2009)  protein translation (lanes 6-7 vs. lanes 2-3, Figure 1c), which was more efficiently recovered by ISRIB (lanes 6-7 vs. lanes 8-9, Figure 1c). These results suggest lower protein translation in ERO1 KO cells under hypoxia and a stronger recovery with ISRIB treatment ( Figure 1d). Furthermore, a viability assay (MTS) pointed to reduced viability of ISRIB-treated ERO1 KO cells, which had undergone hypoxia (Figure 1e).
These findings suggest that sustained activation of the adaptive arm of ISR, inducing attenuation of protein translation in proteostasisimpaired ERO1-devoid cancer cells under hypoxia, is counteracted by ISRIB, thus affecting their viability.

| ISRIB does not inhibit ERO1 and the functionally-related VEGFA signal
Next, we examined whether ISRIB inhibits ERO1. ISRIB was reported both to reactivate protein translation and to inhibit ATF4, which is upstream of CHOP (Rabouw et al., 2019). CHOP also was repressed by ISRIB (Zyryanova et al., 2021). Since CHOP regulates ERO1, we wondered whether ISRIB inhibited ERO1 (Marciniak et al., 2004). To tackle the impact of ISRIB on the levels of secreted VEGF and its ability to promote angiogenesis, we exploited CM collected from equal numbers of WT and ERO1 KO MDAMB231* cultured in normoxic and hypoxic conditions, and treated or not with ISRIB to induce migration of HUVECs, which are primary endothelial cells with proangiogenic potential. We confirmed our previous findings that showed a lower angiogenic potential of ERO1 KO MDAMB231* CM from normoxic conditions, but we could not detect any angiogenic effect of ISRIB in both normoxic and hypoxic conditions (Figure 2d). These results suggest that ISRIB impairs CHOP but has no effect on either ERO1 levels or ERO1-related angiogenic activity.

| ISRIB selectively impairs the growth and spread of ERO1-deficient breast cancer
The detrimental effect of ISRIB on ERO1 KO MDAMB231* cells

| ERO1 KO breast cancer up-regulates PERK branch of UPR
To identify pathways that might account for the different responses of WT and ERO1 KO MDAMB231* breast cancer to ISRIB, we used RNA sequencing analysis for the transcriptional profiling of these tumours from mice, which received either ISRIB, or paclitaxel, or the combination of the two drugs.
Data analyses identified genes that were differently regulated in WT and ERO1 KO MDAMB231* breast tumours: 3580 genes decreased, and 3857 genes increased in ERO1 KO tumours. We then F I G U R E 2 ISRIB inhibits ATF4 and CHOP signal but not that of ERO1.  with their WT counterparts, we found a mammalian/mechanistic target of rapamycin complex 1 (mTORC1) stress signalling, which is part of the integrated response arm of the UPR and involved in activation of protein synthesis (Wouters & Koritzinsky, 2008), and UPR itself  these tumours. However, ATF4, a pro-survival effector of the ISR arm of UPR (downstream to the PERK branch), is down-regulated in the treated ERO1 KO tumours compared with the WT counterparts ( Figure 5d). These findings suggest up-regulation of UPR, and specifically of the PERK arm, in ERO1 KO breast tumours and downregulation of UPR after the combination of ISRIB and paclitaxel treatment in the same tumours.

| ERO1/PERK in breast cancer patients
Previously we reported that the levels of ERO1 correlate with breast tumour aggressiveness (Varone et al., 2021), whereas others have shown that low PERK levels positively correlate with better overall survival (Jewer et al., 2020).

| DISCUSSION
ERO1 is a protein disulphide oxidase that participates in protein oxidative folding of nascent proteins in the endoplasmic reticulum (Zito, 2015). Although its activity in mammals is compensated by other enzymes such as peroxiredoxin (PRDX4), ERO1 deficiency impairs VEGFA folding and secretion in highly aggressive triplenegative breast tumours (MDAMB231*), curtailing the tumour angiogenesis and metastasis (Varone et al., 2021;Zito, 2013;Zito et al., 2012;Zito, Melo, et al., 2010). ERO1's effect, in its capacity as a protein disulphide oxidase, is not restricted to VEGF but also to other angiogenic factors; thus, the consequence of its deficiency on the restraint of tumour angiogenesis might be highly effective (Manuelli et al., 2021;Varone et al., 2021). Unfortunately, the two ERO1 inhibitors currently available cannot be employed in vivo on account of potential off-target effects . Many influential reports have suggested that CHOP regulates ERO1 in disparate ER stress conditions (Li et al., 2009;Marciniak et al., 2004;Pozzer et al., 2018), so we have focused on ISRIB, a small molecule that inhibits the integrated stress response by reactivating protein translation, and also inhibits the downstream CHOP signal (Zyryanova et al., 2021).
In view of the CHOP-induced ERO1 levels and the beneficial cytotoxic activity of ISRIB in some cancers, we tested whether ISRIB also inhibited ERO1 in preclinical models of breast cancers (Ghaddar et al., 2021;Jewer et al., 2020;Nguyen et al., 2018). We employed triple-negative breast cancer MDAMB231 cells which, through serial in vivo passages, acquire a more aggressive phenotype in terms of proliferation and metastases, and refer to them as MDAMB231*.
F I G U R E 6 ERO1/PERK cooperation in breast tumours. Kaplan-Meier plotter depicting relapse-free survival of breast cancer patients (n = 948) stratified for gene expression levels of ERO1 (a), EIF2AK3 (PERK) (b) and the ratio EIF2AK3/ERO1 (c). In panel (c), the upper (n = 237) and lower quartile (n = 237) of the ratio are represented. Statistical significance was assessed using a log-rank test.
WT and ERO1 KO MDAMB231* cells responded differently to hypoxia, which is a common stress condition in solid tumours and their micro-environment. Indeed, we saw that ERO1 KO cells reduce protein synthesis more effectively by activating the ISR arm of UPR (Leprivier et al., 2015;Wouters & Koritzinsky, 2008).
We therefore tested ISRIB's effects on highly aggressive WT and ERO1 KO MDAMB231* under hypoxic conditions. ISRIB was more effective in reactivating the protein translation in ERO1 KO MDAMB231* under hypoxia, which experience proteotoxicity and therefore, impaired their cell viability. In accordance with a previous report, we confirmed the ability of ISRIB to inhibit CHOP expression (Zyryanova et al., 2021). However, under these conditions, ISRIB does not impair either ERO1 expression or its activity, indicating that ERO1 is not regulated through CHOP in hypoxic breast cancer, and therefore ISRIB only blunts the CHOP signal but not that of ERO1, or the functionally related tumour angiogenesis. This outcome corroborates recent findings, pointing instead to the regulation of ERO1 by the transcription factor nuclear factor IB (NFIB) in breast cancer (Zilli et al., 2021).
Irrespective of this, there was a synergistic effect in restraining tumour growth and metastasis in ERO1 KO MDAMB231*-bearing mice treated with ISRIB, despite the lack of any significant response in ISRIB-treated WT-tumour bearing mice. Importantly, the RNA sequencing data on ERO1 KO MDAMB23* breast cancer indicate an increase in UPR and specifically in ER-resident kinase PERK (eIF2AK3) branch, suggesting activation of the PERK arm of the UPR.
The PERK pathway connects ER stress to repression of protein translation and, by up-regulating enzymes and chaperones, fosters protein folding (Walter & Ron, 2011). Our findings highlight an adaptive mechanism whereby the lack of ERO1 in breast tumour cells converges on the PERK pathway of the UPR and, by attenuating protein translation, limits the proteotoxicity. The repression of protein synthesis which, while giving ERO1-deficient cells a significant pro-survival benefit, renders them more susceptible to ISRIB. ISRIB reactivates protein translation by binding eIF2B, a guanine nucleotide exchange factor for eIF2, which becomes resistant to the inhibitory effect of p-eIF2alpha, and thus weakens the prosurvival effect of PERK-mediated repression of the protein translation, which results detrimental in the context of the impaired proteostasis imposed by ERO1 deficiency (Han et al., 2013;Zyryanova et al., 2021). Therefore ERO1 deficiency from one side impairs proteostasis, while from the other side it represses the protein translation through PERK activation, in ER stress conditions relevant for tumours such as F I G U R E 7 ERO1 deficiency in breast tumours up-regulates PERK and dictates the ISRIB-mediated cytotoxic effect. ERO1 is a protein disulphide oxidase in the endoplasmic reticulum, whose expression is regulated by CHOP in a variety of ER stress conditions. Previously, we reported that the lack of ERO1 in highly metastatic breast tumours impairs secretion of angiogenic factors, among which VEGFA, and angiogenesis, hence acting on the tumour resilience. In this study, we employed ISRIB, a small molecule which, by reactivating protein translation, enfeebles the adaptive PERK-mediated mechanism of protein repression. In breast tumour (MDAMB231*) cells under hypoxia, ISRIB inhibits CHOP but has no effect on ERO1 activity, suggesting that under hypoxia CHOP does not regulate ERO1. However, ISRIB is synergistic with ERO1 deficiency in terms of impairment of the tumour burden. Mechanistically, ERO1 deficiency up-regulates the PERK branch of UPR, repressing protein translation, which renders ISRIB more effective to restrain tumour growth in a context of impaired proteostasis. In ERO1 KO tumours, ISRIB-dependent reactivation of protein translation together with the impairment of angiogenesis constitutes a double-hit which weakens tumour resilience to stress. hypoxia. Consequently, the low load of protein translation upon ERO1 deficiency predisposes ERO1 KO MDAMB231* breast cancer cells to become susceptible to ISRIB, which is then toxic by increasing protein translation in a context of impaired proteostasis.
The outcome of ISRIB's selective effect on ERO1 KO MDAMB231* breast tumours with up-regulated PERK pathway is in line with other reports of the effectiveness of ISRIB on mutant KRAS lung cancer with high PERK/p-eIF2alpha (Ghaddar et al., 2021;Jewer et al., 2020). These findings suggest that activation of the PERK arm of the UPR, imposing low levels of protein translation together with impaired proteostasis, is a prerequisite for the cytotoxic effect of ISRIB on cancer cells.
Resistance to paclitaxel, one of the first-line drugs for breast cancer, was suggested to be due to UPR (Lee et al., 2011). This proposal might suggest that a resistance to paclitaxel may be counteracted by drugs that weaken UPR. Our RNA sequencing data suggest an opposite response of WT and ERO1 KO breast cancer to the combination of ISRIB and paclitaxel on the UPR pathway: UPR is up-regulated in WT tumours treated with the combination, but is down-regulated in ERO1 KO tumours treated with the same pharmacological combination. These different UPR responses of WT and ERO1 KO tumours to the combination paclitaxel and ISRIB, together with the impairment of ERO1 KO tumour burden, indicate that UPR weakening correlates with a cytotoxic response of cancer to these two drugs.
In conclusion, our study supports the notion that the PERK arm of UPR with the downstream attenuation of protein translation is an important adaptive mechanism of tumourigenesis and that ISRIB impairs this mechanism of cancer adaptation by reactivating protein translation. Furthermore, our findings on ERO1-deficient breast tumours suggest that ISRIB restrains the growth of tumours with high PERK and a low load of protein translation/low ERO1, possibly because the rapid ISRIB-dependent increase in protein translation results in a toxic action in cells deficient in an enzyme with protein folding activity, hence, with impaired proteostasis (Figure 7). Our analysis of breast tumour patients indicated better relapse-free survival of those with a high PERK/ERO1 ratio, through a cooperation between ERO1 and the PERK pathway in these tumours. To conclude, ISRIB may be a valid drug in the pharmacological armamentarium for breast tumours with high PERK and low ERO1 levels and, thus impaired proteostasis.

CONFLICT OF INTERESTS
The authors declare no competing interests.

AUTHOR CONTRIBUTIONS
EV and AD conducted the experiments. MCB, MB, and LD performed library preparation and next generation sequencing. FP conducted histochemical staining. RG designed in vivo experiments. EZ acquired funding, designed and oversaw the experiments, and wrote the manuscript.

RIGOUR
This Declaration acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research as stated in the BJP guidelines for Research, Design and Analysis, Immunoblotting and Immunochemistry, and Animal Experimentation, and as recommended by funding agencies, publishers and other organizations engaged with supporting research.

DATA AVAILABILITY STATEMENT
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Raw

RNA-sequencing data have been deposited in European Nucleotide
Archive under the accession E-MTAB-11313.