Elevated numbers of infiltrating eosinophils accelerate the progression of Duchenne muscular dystrophy pathology in mdx mice

Eosinophils are key effector cells of the type 2 inflammatory response (Th2). Their development and activation are due to the production of IL5, IL3 and GM-CSF by type 2 innate lymphoid cells (ILC2) and CD4+ helper T cells (Dougan et al., 2019). Dysregulation of the Th2 response often results in blood and tissue eosinophilia, mediating chronic diseases such as asthma or inflammatory bowel disease (Bochner et al., 2012; Rodrigo-Muñoz et al., 2021). For example, in the intestine, eosinophils participate in the deposition of a fibrotic scar after irradiation (Takemura et al., 2018). In addition, aging has been associated with increased circulating eosinophils and their primary chemotaxin: CCL11 (eotaxin-1) (Brigger et al., 2020).

In healthy individuals, skeletal muscle regeneration is a multi-step process in which multiple cell types interact in a coordinated fashion (Oprescu et al., 2020). Fibrosis, in contrast, is a hallmark of severely impaired muscle repair, as seen in the incurable chronic disease Duchenne muscular dystrophy (DMD). DMD is an X-linked genetic disease affecting 1 in 3500 boys (Bladen et al., 2015) in which the DMD gene contains mutations that result in a lack of dystrophin protein, or in the production of a non-functional truncated form, leading to degeneration followed by chronic cycles of inflammation-regeneration (Dadgar et al., 2014). More recently, the lack of dystrophin in satellite cells (SC) has been shown to impair their behavior and contributes to the pathogenesis of DMD (Dumont et al., 2015). Degeneration is associated with the activation and differentiation of tissue-resident fibro-adipogenic progenitors (FAP) that drive matrix production and the formation of a fibrotic scar, as well as adipocyte differentiation and lipid deposition (Joe et al., 2010; Uezumi et al., 2010, 2011). Likewise, recruited inflammatory macrophages play an important role in the establishment of the fibrotic scar (Lemos et al., 2015; Juban et al., 2018; Chazaud, 2020; Juban, 2021). Intriguingly, an increase in eosinophil infiltration was reported in mdx mice, as well as in DMD patients compared with healthy adults (Cai et al., 2000; Kastenschmidt et al., 2021). This has been associated with the development of fibrosis in muscle, which is promoted by the release of major basic protein (MBP) and likely other toxic granule proteins (Cai et al., 2000; Wehling-Henricks, 2004; Wehling-Henricks et al., 2008; Schröder et al., 2013). Currently, corticosteroid treatment is the only effective clinical intervention to delay DMD progression and Prednisone, a well-known corticosteroid, drastically reduces the number of eosinophils present in muscle (Cai et al., 2000; Wehling-Henricks, 2004). On the other hand, eosinophil infiltration in myositis – a term that describes a number of muscle inflammatory diseases of diverse etiology – is not always associated with poor outcomes or with more severe pathology (Selva-O'Callaghan et al., 2014).

The regenerative capacity of skeletal muscle rests on SCs, also called muscle stem cells (MuSC), or muscle progenitors (MP) during regeneration (Lepper et al., 2011; Murphy et al., 2011; Sambasivan et al., 2011). After damage, MuSCs activate, proliferate and initiate the myogenic program to differentiate into myoblasts, which either fuse with existing myofibers or with each other to form new myofibers. A subset of MuSCs will self-renew and return to quiescence in order to maintain the pool of stem cells for further needs (Sacco et al., 2008; Giordani et al., 2018). Concomitantly, inflammatory cells (leukocytes and granulocytes) infiltrate the tissue and direct MP and FAP behavior by secreting various cytokines such as IL1β and TNFɑ (Arnold et al., 2007; Saclier et al., 2013; Otis et al., 2014). Infiltrating monocytes differentiate into Ly6C+ inflammatory macrophages, stimulating MP proliferation and inducing FAP death. Inflammatory macrophages then lose the expression of Ly6C as they skew to a pro-regenerative phenotype, stimulating MP differentiation, myofibre growth and FAP survival via the production of TGFβ, IL6 and NAMPT, among other factors (Arnold et al., 2007; Saclier et al., 2013; Varga et al., 2016; Ratnayake et al., 2021). Although this skewing process is well characterized in acute injury, recent studies suggest a misregulation of macrophage inflammatory function in mdx mice (Juban et al., 2018).

In acute injury, eosinophils are essential for efficient skeletal muscle regeneration, as ΔdblGATA mice, which lack eosinophils, display delayed regeneration (Heredia et al., 2013). Indeed, eosinophils participate in muscle regeneration via IL4/IL13 secretion, resulting in the activation of the trophic functions of FAPs towards MPs. Interestingly, although muscle eosinophilia has been described in mdx muscle, their depletion by crossing the mdx mice with a second strain of eosinophil-deficient mice (PHIL strain) did not improve the histopathology at an early age (Sek et al., 2019). Moreover, the authors also reported no changes in the histopathology of young mdx mice that overexpress IL5 (and display blood eosinophilia) (Sek et al., 2019). Importantly, this study only focused on mdx mice from 3 to 5 weeks of age, a very narrow period which corresponds to the acute inflammatory onset of the disease rather than its fibrotic progression (Pastoret and Sebille, 1995; Tidball et al., 1995; Muller et al., 2001). Thus, the long-term effect of hyper-eosinophilic infiltration on muscle regeneration, repair and fibrosis remains to be investigated. Based on these observations, we sought to confirm the potential toxic role of eosinophils in muscle using Tg(Cd3d-Il5)NJ.1638Nal mice (hereafter designated IL5-Tg; Lee et al., 1997). Here, we demonstrate that hyper-eosinophilia has an adverse effect on muscle repair, resulting in strong collagen deposition and eventually death.

As reported previously (Lee et al., 1997), IL5-Tg mice exhibit severe blood eosinophilia [>50% circulating white cells compared with <3% in wild-type (WT) mice, Fig. S1A,B]. To evaluate the effects of hyper-eosinophilia on muscle repair, notexin (NTX) was injected in the tibialis anterior (TA) muscle to elicit acute skeletal muscle injury, and muscle regeneration was analyzed histologically at 7, 14, and 28 days post damage (Fig. 1A-C). Quantification of the number of nuclei per fiber, as well as myofiber cross-sectional area (CSA) showed a significant delay in skeletal muscle regeneration as reflected first by a reduction in the number of nuclei per fiber at 7 days post damage (−13%; P=0.054; Fig. 1B). Furthermore, the size of newly formed myofibers was reduced at 14 and 28 days post damage (−17%, P=0.076; and −23%, P<0.001) (Fig. 1C). Interestingly, it has been reported that eosinophils stimulate tissue-resident FAPs, which, in return, enhance MP proliferation (Heredia et al., 2013). After confirming muscle eosinophilia (Fig. S1C,D), we analyzed FAP and MP proliferation by 5-ethynyl-2′-deoxyuridine (EdU) incorporation (Fig. S1E). Although we noticed an increase in EdU incorporation 1 day post injury, suggesting that MPs exit quiescence faster (+273.6%, P=0.058; Fig. S1F,G), MP proliferation was not affected at 3 and 5 days post injury (Fig. 1D,E) suggesting that MPs undergo an immediate exit of quiescence post injury. No change in FAP proliferation was detected in the IL5-Tg mouse model (Fig. 1F,G). Lastly, eosinophils have been described as a specific source of TGFβ and IL4, which could skew macrophages from a pro-inflammatory (Ly6C+) to a pro-regenerative (Ly6C−) phenotype. In our system, however, the proportion between these two subsets was not affected (Fig. 1H), but we did notice a decrease in the total number of macrophages (Fig. S1H) and especially a reduction in the total number of Ly6C− macrophages per muscle (Fig. 1I,J). Overall, these results suggest that hyper-eosinophilia affects MP early activation and macrophage infiltration, delaying MP fusion and myofiber growth.

As eosinophils have been described to stimulate FAP trophic function toward MPs (Heredia et al., 2013), we decided to co-culture single WT myofibers with activated FAPs from WT or IL5-Tg muscles. SC proliferation (EdU/MyoD) and differentiation (myogenin) were analyzed at 48 h and 72 h (Brun et al., 2018; Fig. S2A,D). As previously shown (Fiore et al., 2016), WT FAPs strongly enhance SC proliferation compared with the no FAP control condition (CT), shown by an increase in EdU+ cells at 48 h (P<0.05; Fig. S2B) and by an increase in the number of cells per clone at 72 h (P<0.05; Fig. S2E). Furthermore, by analyzing the different states of SC activation (Fig. S2C), we observed that co-cultivating myofibers with FAPs alters the proportional commitment as shown by the decreased percentage of MyoD+ cells (P<0.05 and P<0.01) and increased percentage of proliferating Pax7/EdU/MyoD+ cells (P<0.05 and P<0.01). Interestingly, the only difference noted between WT and IL5-Tg co-culture was an increase in MyoD/EdU+ cells, which is consistent with the slight increase in activation observed in vivo (Fig. S1F). However, this early difference did not result in defects in SC proliferation (Pax7+/EdU+), differentiation (myogenin+) or return to quiescence (measured indirectly by Pax7+/EdU−/myogenin− cells) at 72 h (Fig. S2D-F). These results demonstrate that the effect seen previously is likely due to the hyper-eosinophilia, and not to a defect in FAP trophic functions.

Although eosinophils are typically described as Th2 inflammatory cells, such as in adipose tissue (Rothenberg and Hogan, 2006; Rana et al., 2019), it is now accepted that they can also acquire different inflammatory profiles depending on their stimulation, and especially a Th1-antiviral-response (Drake et al., 2016). The authors demonstrated that, similar to macrophages, eosinophils from IL5-Tg mice were able to show a Th1 signature after treatment with E. coli, IFNγ or a combination of the two (Dolitzky et al., 2021). Interestingly, as skeletal muscle regeneration relies on both Th1 (IFNγ) and Th2 (IL4) responses (Arnold et al., 2007; Heredia et al., 2013), it is not surprising that eosinophils would display both Th1 and Th2 markers. To explore this possibility, WT and IL5-Tg eosinophils were sorted at 3 days post damage and subjected to RNA-sequencing (RNA-seq). Principal component analysis (PCA) indicated a clear transcriptomic difference between WT and IL5-Tg eosinophils (Fig. 2A). Compared with the WT samples, 1155 differentially expressed genes (DEGs) were discovered (676 upregulated and 479 downregulated DEGs; Table S3). By focusing on known cytokines and chemokines produced by eosinophils (Rothenberg and Hogan, 2006; Dolitzky et al., 2021), we were able to show that many of the upregulated DEGs in IL5-Tg eosinophils included both Th1 and Th2 markers such as Cd36, Cd274, Car4, Lair1, Cd101, Tgfb1 and Vegfa (Fig. 2B). Interestingly TGFβ1 has been recently shown to negatively impact MP fusion. Thus, an increase in this factor could explain the delay in skeletal muscle regeneration, and specifically the decrease in number of nuclei per fiber (Fig. 1B) (Girardi et al., 2021). Moreover, it is important to note the upregulation of Epx and Prg2 (MBP), which are enzymes known to contribute to the pathology of the DMD (Cai et al., 2000; Wehling-Henricks et al., 2008). Lastly, IL5-Tg eosinophils express less Il4, which could explain the observed decrease in macrophage numbers and, in particular, the pro-regenerative Ly6C− macrophages (Fig. S1H-J). It is important to note that macrophage skewing was not studied in the ΔdblGATA mice, as the authors state that both inflammatory and pro-regenerative macrophages infiltrate the tissue after damage (Heredia et al., 2013). However, the authors do not report defects in skeletal muscle regeneration in mice lacking the IL4R in myeloid cells (LysM^CRE:IL4R^flox mice), suggesting that IL4 is either not crucial for macrophage polarization, or that compensatory mechanisms exist in this model (Heredia et al., 2013). Overall, our observations are discordant with the paradigm that eosinophils (and hyper-eosinophilia) drive the generation of a Th2 environment. To highlight a relationship between hyper-eosinophilia and macrophage polarization, we analyzed the expression of Ly6C on blood monocytes (MO) and tissue-resident macrophages from skeletal muscle and white adipose tissue (WAT). As expected, the expression of Ly6C on blood MO was severely affected, with Ly6C+ MOs representing 56% of the population in WT mice but only 12% in the IL5-Tg mice (P<0.001; Fig. 2C,F). In contrast, we did not observe significant changes in the Ly6C+ population in skeletal muscle (Fig. 2D,F). Strikingly, we instead observed an increase of the proportion of Ly6C+ macrophages in WAT (P<0.01; Fig. S2E,F). In aggregate, these data suggest that the number and function of eosinophils must be carefully balanced in order to contribute to a restorative Th2 response, and that changing their levels or altering their activation status can result in homeostatic changes that lead to failed regeneration. Moreover, we confirm here that hyper-eosinophilia is not necessarily linked to healthy Th2 inflammation in all tissues.

After observing the impact of hyper-eosinophilia on regeneration during acute injury, we sought to determine whether it also plays a role in a mouse model of chronic disease. A murine model of DMD (D2-mdx) was generated, wherein IL5 was also overexpressed (mdx:IL5-Tg). Strikingly, mdx:IL5-Tg mice exhibited decreased survival, with over 50% dying by 6 months of age, whereas the mdx:WT mice survived up to 12 months of age (Fig. 3A). Muscle mass and histopathology were assessed at 6/7 and 12 months of age. Overall, mdx:IL5-Tg mice displayed a reduction in body mass (P<0.01), with a decrease of 22% and 28% (P<0.05) at 6/7 and 12 months (Fig. 3B). Upon calculating muscle to body mass ratio (Table S1), we found no significant changes in TA weight ratio, likely due to the small sample size (n=3-6) (Fig. 3C). However, we observed decreases in gastrocnemius (GC) and quadriceps (Quad) weight ratios between mdx:WT and mdx:IL5-Tg mice (P<0.05 and P<0.001; Fig. 3D,E). The histopathology of mdx:IL5-Tg mice and their control littermates was assessed with laminin staining and Picrosirius Red (PSR) to quantify myofiber size and fibrosis (Table S1). Interestingly, mdx:IL5-Tg mice did not exhibit any change in average myofiber size compared with mdx:WT in TA, GC, Quad and diaphragm muscles at 6/7 and 12 months, albeit no significant difference in size was found, likely due to low replicate numbers in the mdx:IL5-Tg cohort (Fig. 3F-I). However, we did notice a dramatic change in myofiber shape, with the emergence of hypertrophic myofibers in the GC and Quad of mdx:IL5-Tg mice at 12 months of age (Fig. 3G,H, white arrowheads). This could be a marker of myofiber branching, which is a known phenotype of DMD (Faber et al., 2014). This also indicates poor muscle repair, with few newly formed myofibers (small fibers). Through PSR staining, we noted that mdx:IL5-Tg mice exhibit an overall increase in fibrosis in the TA (P<0.01; Fig. 3J), that reached +44% (P<0.05) and +18.8% (P=0.08) at 6/7 and 12 months of age, respectively. Similarly, the GC displayed increased collagen deposition at 12 months (P<0.05; +54.8%; Fig. 3K), as well as the Quad (Fig. 3L) at 6/7 and 12 months of age. Finally, diaphragm muscles displayed more fibrosis overall (P<0.05; Fig. 3M) and we noted high variability in the fibrotic deposition of mdx:IL5-Tg diaphragms compared with mdx:WT mice, resulting in areas with very few muscle fibers, which is likely the cause of early death of mdx:IL5-tg mice (Fig. S3). Moreover, a possible role of eosinophilia in the development of cardiomyopathies has previously been described (Diny et al., 2017), although we did not observe any overt fibrotic scarring in the hearts of dead mdx:IL5-Tg mice (data not shown). In mdx muscles, where cycles of degeneration and repair occur, eosinophils have been shown to produce perforin, a substance that attacks muscle fibers through pore formation and causes necrosis, which we show to be increased in IL5-Tg eosinophils (Fig. 2) (Cai et al., 2000). In summary, although our results are limited due to a low number of mice, we found that mdx:IL5-Tg mice, compared with controls, clearly have a shortened lifespan and exhibit a worsening histopathology in muscle with age, suggesting that hyper-eosinophilia exacerbates the typical symptoms of DMD, including collagen deposition and, consequently, earlier death. Although Sek et al. (2019) described a moderate influence of hyper-eosinophilia on muscle histopathology, here we observed a dramatic effect on muscle fibrosis, which eventually led to death. One main difference is that here we have used the D2-mdx mice (D2.B10-Dmdmdx/J), unlike the previous authors who used the classic mdx (C57BL/10ScSn-Dmd mdx/J) mouse model. The D2-mdx mouse model is known to have a more robust phenotype closely mimicking human DMD (Fukada et al., 2010), which could explain the difference between studies. Also, Sek et al. focused exclusively on the early onset of the disease (3-5 weeks of age) corresponding to an acute inflammatory response, whereas we have focused more on the chronic phase of the disease.

Due to the low number of mdx:IL5-Tg mice reaching 6/7 to 12 months of age, we used microdamage (microD) to exacerbate the chronic damage caused by DMD and to induce the formation of fibrotic scarring in the TA muscle of younger mdx mice (Desguerre et al., 2012). At 3 months of age, mdx:IL5-Tg mice did not display loss in body weight (Fig. S4A). However, the mass of the microD TA was remarkably decreased compared with the non-damaged TA (−15%, P<0.05), and even more so when compared with the mdx:WT-microD TA (−32%, P<0.001; Fig. S4B; Table S1). In control mice, we observed that the microD protocol resulted in a 34% decrease in myofiber CSA (Fig. S4C,D), as expected. We saw a 49% decrease (P<0.01) in non-damaged TA CSA in mdx:IL5-Tg mice compared with undamaged controls and, upon microD, we observed a further decrease of 57% (P<0.01) in the CSA between the two groups (Table S1). This defect in myofiber size was associated with a strong increase in collagen deposition and scarring, as shown with PSR staining (+93%, P<0.001; Fig. S4E,F; Table S1). This last experiment confirmed that hyper-eosinophilia participates in the pathologic progression of the muscle histopathology, specifically decreasing myofiber growth and increasing fibrosis deposition within the muscle. Taken together, our results and the results from Sek et al. (2019) support a model where eosinophils do not participate in the early onset of the disease, but continuous hyper-eosinophilia will eventually affect chronic muscle repair and accelerate the fibrotic deposition. Muscle eosinophilia has previously been associated with myositis; however, it does not appear to be always associated with poor outcomes (Selva-O'Callaghan et al., 2014), and although muscle eosinophilia has been investigated in autosomal-recessive limb girdle muscular dystrophies (LGMD2A/2C), to our knowledge it has not been fully investigated in DMD (Schröder et al., 2013). Based on the data provided here, assessment of eosinophils in human DMD is clearly warranted and simple peripheral blood eosinophil counts could be explored as a biomarker for the advanced stage of the disease and guide the appropriate choice of treatment.

Mice were maintained in an enclosed and pathogen-free facility, housed in standard cages under 12 h light/dark cycles and fed ad libitum with a standard chow diet. All experimental procedures were approved by the University of British Columbia Animal Care Committee. IL5-Tg [Tg(Cd3d-Il5)NJ.1638Nal; Lee et al., 1997] and control littermate adult male mice (>8-weeks-old) were used for NTX injury. Acute muscle damage was induced by intramuscular injection of 0.15μg NTX snake venom (Latoxan), into the TA muscle. Then 0.5 mg of EdU was injected intraperitoneally the evening before harvesting (Click-IT plus EdU proliferation kit, Thermo Fisher Scientific). For co-culture experiments, TA and GC muscles of adult female WT and IL5-Tg littermates were damaged with NTX.

IL5-Tg male mice were bred for one generation with D2-mdx females (D2.B10-Dmdmdx/J, The Jackson Laboratory, #013141; Fukada et al., 2010). Only male littermates were used for experiments; micro-damage (microD) was performed as previously described (Desguerre et al., 2012) by injuring the TA muscle with 15 daily micro-needle pricks. TA muscles were harvested 1 week after the last day of injury.

Blood was collected from the saphenous vein: 1-2 drops of blood were collected in 1× PBS 2 mM EDTA and centrifuged at 300 g for 5 min at 4°C. Red blood cells were lysed using ACK lysis buffer (Gibco) and peripheral blood mononuclear cells (PBMCs) were stained with fluorophore-conjugated antibodies for detecting eosinophils as described below.

Mice were weighed, TA, GC, Quad and diaphragm muscles were harvested without fascia, weighed and fixed in 1% paraformaldehyde (PFA) overnight before being transferred to 70% ethanol. Muscles were embedded in paraffin and then sectioned at a thickness of 4 µm using a Leica microtome (RM 2255). Sections were stained with Hematoxylin and Eosin (H&E), PSR (Waxit) or laminin (Theret et al., 2021).

For immunostaining, damaged TAs were harvested, frozen in isopentane pre-cooled in liquid-nitrogen and stored at −80°C. We prepared 8 µm sections that were fixed with 4% PFA and saturated with blocking buffer (1× PSB, 3% goat serum, 0.3% Triton X-100). Primary antibodies (siglecF: Invitrogen, 14-1702-80, 1:200; laminin: Abcam, ab11575, 1:200) were incubated overnight at 4°C. Secondary antibodies (Invitrogen, A21247 and A11011; 1:1000) were incubated for 2 h at room temperature and DAPI (Invitrogen, D3571; 0.6 µM) was used to counterstain nuclei.

Images were acquired using a Nikon Eclipse Ni equipped with a dual-use monochromatic and color imaging camera (Nikon Digital Sight DS-U3 for brightfield, Qimaging Retiga EXi for fluorescence), operated via NIS-Elements software. We used 10× and 20× objectives for stitching complete tissue images from tiles and for representative images in figures. Collagen deposition of the whole muscle section (PSR) was calculated using Adobe Illustrator CS6 and Fiji (ImageJ, version 2.0.0-rc/69/1.52n). CSA of the whole muscle section (laminin) was calculated using Fiji and Open-CSA macro (Desgeorges et al., 2019). Images were assembled using Adobe Illustrator CS6.

Mice were anesthetized using avertin, followed by cardiac puncture, and perfused with 20 ml of 1× PBS 2 mM EDTA. Muscles were harvested without fascia, minced and digested for 1 h in 1.5 U/ml collagenase D (Millipore Sigma, 11 088 882 001), 2.4 U/ml Dispase II (Millipore Sigma, 04 942 078 001) and 10 mM CaCl₂ at 37°C. The cell suspension was diluted in FACS Buffer [1× PBS, 2% fetal bovine serum (FBS), 2 mM EDTA] and passed through a 40 µm filter. After a 20-min incubation in FACS buffer with 1:1000 Fc Receptor-blocking antibody (AbLab), cells were stained with antibodies listed in Table S2. For the proliferation assay, cells were stained for EdU incorporation as described in the manufacturer's manual (Click-iT Plus EdU kit; Thermo Fisher Scientific). Cell populations were acquired using LSRII (BD Biosciences) and Diva software. Flow data was then analyzed using FlowJoTM (BD Biosciences).

Briefly, 3 days after NTX, muscles were harvested and digested as described above and stained for FAPs (antibodies in Table S2). FAPs (CD31−/CD45−/α7int− SCA1+ cells) were sorted using Astrios cell sorter (Beckman Coulter) and plated at 52,000 cells/cm² in six-well plates in 4.5 g/l glucose Dulbecco's modified eagle medium (DMEM; Invitrogen) supplemented with 1% sodium pyruvate, 10% FBS and 2.5 ng/ml bFGF (Invitrogen).

FAPs and myofibers were co-cultured as previously described (Fiore et al., 2016). Briefly, 4 days after FAP sorting, single myofibers from WT extensor digitorum longus muscles were extracted as described by Brun et al. (2018) and placed in FBS-coated Bowden chambers (Falcon, 353102). Chambers containing 50-60 single myofibers were placed on top of previously sorted FAPs in 4.5 g/l glucose DMEM supplemented with 5% FBS and 0.5% chick embryo extract. After 47 h or 71 h of co-culture, myofibers were incubated for 1 h with 20 µM EdU before being harvested, fixed with 4% PFA and stained for Pax7/EdU/MyoD (48 h) or Pax7/Edu/myogenin (72 h), as previously described in Brun et al. (2018). Quantification was carried out using a Nikon Eclipse Ni equipped with a dual-use monochromatic Qimaging Retiga EXi camera and operated via NIS-Elements software. Representative images were taken using a 40× objective.

Three days after NTX damage, mice were anesthetized using avertin, followed by cardiac puncture and perfusion with 20 ml of 1× PBS 2 mM EDTA. Muscles were carefully dissected and digested as described above. The cell suspension was stained using a panel of antibodies listed in Table S2, and eosinophils (CD45+ Ly6G−/NK1−/CD3− CD11b+ SiglecF+ cells) were directly collected into RNAzol using the Influx sorter (Becton Dickinson). RNA was extracted using the Direct-zol RNA Miniprep (Zymo Research).

Sample quality was assessed using the Agilent 2100 Bioanalyzer. RNA samples were prepared in accordance with the protocol for the TruSeq Stranded mRNA library kit (Illumina) on the Illumina Neoprep automated microfluidic library prep instrument. Paired-end sequencing was performed on the Illumina NextSeq 500 using the High Output 150 cycle Kit (Illumina). Illumina base call files were de-multiplexed using bcl2fastq2 (v 2.20) on BaseSpace. Adaptor sequences were trimmed and low-quality reads (<35 base pairs) were discarded. In addition, Bowtie was used to remove read pairs that aligned against abundant sequences. Demultiplexed read sequences were then aligned to the mm10 genome reference using STAR aligner. The number of aligned reads to each annotated gene was tallied with RnaReadCounter to generate read-count matrices for all samples, which were used as inputs for downstream analyses. Bowtie, STAR and RnaReadCounter are tools built under RNA-Seq Alignment (v 1.1.1). Downstream analyses of read-count data were performed in R (v 3.6.2). Genes with less than 2 counts per million (CPMs) in at least three samples were filtered out. Batch correction was applied using RUVSeq (v 1.20.0; Risso et al., 2014) where 3000 least-significant DEGs generated from a first-pass differential expression analysis were used as ‘in-silico’ controls to estimate unwanted factors of variation. Batch-adjusted counts were normalized by upper-quartile normalization with EDASeq (v 2.20.0; Risso et al., 2011) before being subjected to PCA. Filtered counts were reprocessed and analyzed using DESeq2 (v 1.26.0 (Love et al., 2014)), incorporating two factors of unwanted variation (k=2) from the RUVg function in the design formula. Analyses in DESeq2 included count normalization and differential expression analysis. Volcano plot shown in Fig. 2B was generated using the EnhancedVolcano package (v 1.12.0; https://github.com/kevinblighe/EnhancedVolcano). All P-values were adjusted using the Benjamini-Hochberg correction.

Graphs were plotted and statistical tests performed using Prism 8 (GraphPad Software). Unpaired two-tailed Student's t-test or two-way ANOVA were performed, followed by post-hoc testing (Šidák multiple comparison). P<0.05 was considered statistically significant. Gaussian distribution was not assumed. Sample size for each experiment is indicated in the figure legends. Data represented in graphs are mean values with each individual replicate value plotted±standard error of the mean. Figures were assembled using Adobe Illustrator CS6.

The authors thank Justin Wong and Andrew Johnson from the UBC Flow Core, as well as Ryan Vander Werff, Tara Stach, Bernie Zhao and Dennis Okello from the BRC-Seq.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.200112.