The strand-specific regulation of miR-155-3p in response to lip- opolysaccharide and interleukin-10 stimulation requires FUBP1 protein

The microRNA-155 (miR-155) promotes inflammatory responses in macrophages. Activating macrophages with lipopolysaccharide (LPS) elevates miR-155, while the anti-inflammatory cytokine interleukin-10 (IL10) reduces miR-155 levels. MiR-155 exists in two forms, miR-155-5p and miR-155-3p, produced from the precursor of miR-155 (pre-miR-155). LPS stimulation of macrophages results first in elevation of miR-155-3p levels, followed by increases miR-155-5p. We previously identified the CELF2 protein to interact with pre-miR-155 and impair miR-155-5p expression. We now show that CELF2 only regulates the miR-155-5p expression and that another protein called FUBP1 controls miR-155-3p levels in response to LPS and IL10.


Introduction
MicroRNAs (miRNAs) are short regulatory non-coding RNAs that post-transcriptionally regulate the expression of genes through sequence-specific targeting of mRNA [1]. Initially discovered as involved in the cellular development of larval Caenorhabditis elegans, miRNAs have now been shown to regulate many genes in mammalian cells [2,3], including the host immune cell response to pathogens. For example, an innate immune cell called a macrophage [4] becomes activated to produce inflammatory cytokines by exposure to bacterial cell wall components such as lipopolysaccharide (LPS) to help eliminate the pathogen [4][5][6][7][8]. O'Connell et al. identified miRNA-155 (miR-155) as a key miRNA induced in the inflammatory response [7] and required for macrophage expression of the inflammatory cytokine tumor necrosis factor-alpha (TNFα) [7].
We now describe the role of FUBP1 in the control of miR-155-3p expression. First, we show that FUBP1 increases its association with pre-miR-155 in response to LPS and IL10. Second, we examined the role of FUBP1 in macrophage function by creating FUBP1 deficient cells using CRISPR-Cas9-mediated gene targeting. We found that in LPS stimulated macrophages, IL10 inhibition of TNFα requires FUBP1. Interestingly, FUBP1 affects the production of the two alternative strands of miR-155, miR-155-5p, and miR-155-3p oppositely: FUBP1 promotes miR-155-3p expression but inhibits miR-155-5p. Finally, we show that FUBP1's KH domain 3 mediates FUBP1 binding to pre-miR-155, suggesting an important new role for FUBP1 in controlling miR-155 biogenesis affecting the inflammatory response of macrophages.

Interleukin-10 induces association of FUBP1 to pre-miR-155
We and others have shown that IL10 inhibits LPS-induced miR-155-p expression in macrophage cells [18,25]. We further showed that IL10 inhibits the maturation of pre-miR-155 to mature miR-155-5p [18] and that the RNA binding protein CELF2 contributes to the process [25]. However, our mass spectrometry-based examination pre-miR-155 associated proteins had also identified FUBP1 as another protein that might interact with pre-miR-155 in an LPS and IL10 dependent manner [25].
To follow up on the mass spectrometry data, we analyzed pre-miR-155 pull-down samples for the presence of FUBP1 protein. The amount of FUBP1 observed in cell lysates remained constant regardless of whether the cells were stimulated or not, suggesting that FUBP2 expression levels do not change in response to stimulations. Instead, it is the association of FUBP1 with pre-miR-155 that changes. FUBP1 in the pull-down was quantified and normalized to the total FUBP1 in cell lysates. As seen in Figure 1, LPS treatment of cells increased the amount FUBP1 associated with pre-miR-155 is with no additional effect of IL10 stimulation. RAW264.7 cells were transfected with biotinylated pre-miR-155 oligonucleotide and stimulated with LPS ± IL10 for 2 hours before collecting the pull-down samples. Expression levels of FUBP1 protein interacting with pre-miR-155 oligonucleotide were determined by immunoblotting. The graph shows the FUBP1 band intensities in the pull-down sample normalized to the FUBP1 protein in total cell lysate. The significance in the difference between the LPS ± IL10 stimulations to unstimulated sample or comparison indicated was calculated by One-Way ANOVA with Tukey's correction. * p < 0.05, ns = not significant.
After confirming the reduction of FUBP1 protein in the FUBP1 KD cells, we stimulated the parental RAW264.7, FUBP1 KD, and CELF2 KD cells with LPS ± IL10 for 4 hours, isolated RNA, and quantified the levels of pri-miR-155, pre-miR-155, miR-155-5p, and miR-155-3p via qPCR. The data were all normalized to the LPS-stimulated sample of the parental RAW264.7 cells. Figure 2b shows that knocking down CELF2 does not alter the basal (unstimulated) expression of pri-miR-155, pre-miR-155, miR-155-5p and miR-155-3p. However, the LPS-stimulated levels of all four increased in CELF2 KD cells, consistent with our previous finding that CELF2 is a negative regulator of miR-155-5p expression [25]. Also, as we previously reported, IL10 inhibition of miR-155-5p is impaired in CELF2 KD cells [25]. However, we now show that IL10 inhibition of miR-155-3p is normal in the CELF2 KD cells (Figure 2b, rightmost panel).
Previous studies report that miR-155 deficiency results in increased LPS-induced inflammatory cytokine production [25,28]. Therefore, we examined IL10's action on TNFα production in FUBP1 KD cells. Cells were stimulated with LPS ± 0.5 ng/mL or 10 ng/mL IL10 for 1 hour, and TNFα levels in the supernatant were quantified by ELISA. Figure 3 shows that in parental RAW264.7 cells, IL10 inhibits LPS-induced TNFα expression in a dose-dependent manner. However, in FUBP1 KD cells, IL10 could not decrease LPS-induced TNFα levels at either concentration of IL10 tested. This impairment of IL10 action suggests IL10R signaling requires FUBP1 for IL10 inhibition of LPS-induced TNFα, perhaps by reducing the level of miR-155-3p in response to IL10. FUBP1 KD cells were stimulated with 1 ng/mL LPS ± indicated concentration of IL10 for 1 hour before collecting the cell culture supernatant. The level of TNFα in the supernatants was determined by ELISA. (Two-Way ANOVA with Tukey's correction. *** p < 0.001, ** p < 0.01, ns = not significant).

Figure 4. STAT3 dependence of miR-155-5p and miR-155-3p expression
Expression level in STAT3 WT and KO cells of pri-miR-155, pre-miR-155, miR-155-5p or miR-155-3p was determined by qPCR and normalized to GAPDH or snoRNA202 levels. (a) Data plotted to represent the RNA expression normalized to LPS-stimulated sample of the STAT3 WT. (b) The data in panel A replotted with RNA expression normalized to each cell's own LPS-stimulated sample. (c) Kinetics of miR-155-5p and miR-155-3p expression normalized to the STAT3 WT 0 hour sample. The significance of the difference in values (if any) between the LPS ± IL10 stimulated sample at the same time point was determined by Two-Way ANOVA with Tukey's correction. **** p < 0.0001, *** p < 0.001, ** p < 0.01, ns = not significant.
We then examined the kinetics of miR-155-5p and miR-155-3p expression (Figure 4c). In STAT3 WT cells stimulated with LPS, miR-155-3p rises rapidly with a peak at 2 hours, while miR-155-5p rises more slowly but is still rising at 4 hours. The presence of IL10 (LPS + IL10) reduced the levels of both miR-155-5p and miR-155-3p. These kinetics are similar to those observed by Simmonds et al. in human peripheral blood-derived macrophages [26]. In STAT3 KO cells stimulated with LPS, miR-155-3p levels also rise before miR-155-5p. Remarkably, the presence of IL10 (LPS + IL10) enhanced rather than inhibited the ex-pression of both miR-155-5p and miR-155-3p. We did not look at longer time points because the effect of autocrine cytokines will contribute to gene expression at longer stimulation times.
We next examined the levels of pri-miR-155-, pre-miR-155, miR-155-5p, and miR-155-3p in SHIP1 WT and KO cells at 4 hours of stimulation. Perimacs were isolated from mice as described above. The Figure 5a data are normalized to the LPS-stimulated sample of the SHIP1 WT cell and show that SHIP1 deficiency leads to elevated basal (unstimulated) levels of pri-miR-155, pre-miR-155, miR-155-5p and miR-155-3p. Paradoxically, the LPSstimulated expression of all four RNAs is decreased in the SHIP1 KO compared to the SHIP1 WT cells. To see if IL10 could inhibit the LPS-induced expression of the four RNAs in the SHIP1 KO cells and account for impaired LPS stimulation of the four RNAs, we normalized the RNA expression data to each cell's own LPS-stimulated sample. As Figure  5b shows, IL10 can inhibit LPS-induced pri-miR-155, pre-miR-155, and miR-155-5p in both SHIP1 WT and KO cells, but IL10 inhibition of miR-155-3p is impaired in the SHIP1 KO as compared to the SHIP1 WT cells. These observations suggest IL10 control miR-155-3p expression is more dependent on SHIP1 signaling than of miR-155-5p. sample at the same time point was determined by Two-Way ANOVA with Tukey's correction. **** p < 0.0001, *** p < 0.001, ** p < 0.01, * p < 0.05, ns = not significant.

Figure 5c
shows the kinetics of miR-155-5p and miR-155-3p expression in the SHIP1 WT and SHIP1 KO cells. In SHIP1 WT cells, miR-155-3p can be detected at 2 hours, which is earlier than miR-155-5p which is seen only at 4 hours. The kinetics of both miR-155-3p and miR-155-5p in the SHIP1 WT cells (Figure 5c) is slightly delayed compared to the STAT3 WT cells (Figure 4c). This difference might reflect the different genetic backgrounds of SHIP1 WT/KO (BALB/c) vs. STAT3 WT/KO mice (C57BL/6). IL10 could inhibit miR-155-5p in both SHIP1 WT and KO cells. However, IL10 inhibited miR-155-3p expression less well in SHIP1 KO as compared to SHIP1 WT cells.

FUBP1 interacts with pre-miR-155 via KH domain 3
Our data suggest that the regulation of miR-155-3p expression in response to LPS and IL10 is dependent on FUBP1. Previous studies have shown the RNA-binding protein KSRP interacts with pri-miR-155 and pre-miR-155 to regulate miR-155 expression [28,30]. In addition, Zhou et al. described KSRP as required for regulation of miR-155-5p; in KSRP deficient cells, miR-155-5p levels in response to LPS stimulation in dendritic cells are decreased, while miR-155-3p levels are increased [30].
Both KSRP and FUBP1 proteins have a conserved architecture of 4 tandem KH domains to interact with RNA/DNA [34]. However, KSRP mainly interacts with RNA via its third KH domain (KH3) and substituting the key GXXG residues of KH domain with GDDG significantly decreased the interaction of KSRP to RNA [32]. Therefore, to test whether FUBP1 interacts with pre-miR-155 through its KH3 domain, we generated recombinant WT and KH3 domain GDDG mutants for FUBP1 and KSRP (Figure 6a). We then measured the interaction of these recombinant proteins to pre-miR-155 using biolayer interferometry (BLI). Figures 6b and 6c shows that both KSRP WT and FUBP1 WT interact with pre-miR-155, but the KH3 GDDG mutants of KSRP and FUBP1 do not. (c) Data plotted to represent the BLI response of indicated proteins with biotinylated pre-miR-155. Unpaired Student's t-test determined the comparison between the WT and KH3 GDDG mutant. **** p < 0.0001, *** p < 0.001.
In addition to FUBP1's control of pri-miR-155 and pre-miR-155, FUBP1 also regulates miR-155-3p levels (Figure 2b). The FUBP1 knockdown studies suggest that FUBP1 participates in LPS stimulation and IL10 inhibition of miR-155-3p. However, FUBP1 knockdown appears to have no further effect on miR-155-5p independent of its impact on pri-miR-155 and pre-miR-155. IL10 inhibition of miR-155-5p is impaired in CELF2 KD cells, as we have previously described [25]. However, our current data suggest this is partly due to the requirement for CELF2 for IL10 inhibition of pre-miR-155 expression.
IL10 signaling downstream of the IL10R involves both STAT3 and SHIP1:STAT3 pathways [14]. Therefore, we examined the STAT3 and SHIP1 dependency of IL10 regulation of pri-miR-155, pre-miR-155, miR-155-5p and miR-155-3p. We found that IL10 required only STAT3 to inhibit pri-miR-155, pre-miR-155 and miR-155-5p, and miR-155-3p (Figure 4b). The STAT3-only requirement for IL10 inhibition of miR-155-5p contrasts with our previous observation that IL10 inhibition of miR-155-5p expression required both SHIP1 and STAT3 [18]. In that study, we used perimacs allowed to rest in culture overnight before the stimulations. An overnight rest period better lets us observe the responses of the cells independent of the environment those cells might have experienced in the mouse. For example, the SHIP1 KO mouse has elevated specific cytokines in its circulation [36]. However, we now use perimacs within 2 hours of extraction from the mouse, so the cells more resemble their native state in the mouse. With these 2-hour rested perimacs, we found: (i) basal levels of miR-155-5p and miR-155-3p are higher in the SHIP1 KO than SHIP1 WT cells, and (ii) LPS induction of miR-155-5p and miR-155-3p is reduced in the KO as compared to the WT cells. The increased basal miR-155 expression may reflect the lack of SHIP1 in these cells since SHIP1 is a negative regulator of miR-155 expression [37]. On the other hand, the reduced miR-155-5p and miR-155-3p levels induced by LPS in the SHIP1 KO cells may be due to network-dosage compensatory mechanisms [38] developed in the cell to blunt stimulation-dependent elevation of miR-155 levels above the basal level.
With these caveats in mind for interpretation of the data from the SHIP1 KO cells, we examined the ability of IL10 to inhibit the LPS-induction of miR-155-5p and miR-155-3p in SHIP1 WT and SHIP1 KO cells (Figure 5b, normalized to each cell's own LPS-stimulated samples). As Figure 5b shows, IL10 can inhibit LPS-induced miR-155-5p in both SHIP1 WT and KO cells, but SHIP1 deficiency impairs Il10 inhibition of miR-155-3p. These data suggest the IL10 control of miR-155-5p requires only STAT3 but control of miR-155-3p requires both SHIP1 and FUBP1 (summarized in Figure 7).
We had previously examined the kinetics of IL10 inhibition of LPS-induced expression of TNFα [14]. We showed that the early (within 2 hours) response required a SHIP1:STAT3 complex, while the late phase (>2 hours) only needed STAT3. In the current study, we found that miR-155-3p becomes expressed earlier than miR-155-5p, and inhibition of miR-155-3p requires SHIP1 and STAT3, while inhibition of miR-155-5p requires only STAT3 (Figure 7). These observations suggest that IL10 inhibition of miR-155-3p may be necessary for IL10 inhibition of the early (SHIP1/STAT3 dependent) phase of TNFα production. In contrast, miR-155-5p participates in inhibition of the late (STAT3 dependent) phase of TNFα expression. However, the mechanism by which STAT3 or SHIP1:STAT3 complexes control miR-155-5p and miR-155-3p remains to be determined. One possibility is that both lead to proteins' expression that participates in the control of pre-miR-155 processing. In fact, since the absence of STAT3 results in IL10 enhancing rather than reducing LPS-induced miR-155-5p and miR-155-3p levels, we predict these STAT3-induced proteins may suppress pre-miR-155 processing. FUBP1 has not been previously described to participate in miRNA processing. Ruggiero et al. reported that KSRP binds pre-miR-155 and is required to mature pre-miR-155 to miR-155-5p [28]. The effect of KSRP KD [30] (decrease LPS-induced miR-155-5p and increase miR-155-3p) is opposite to that of FUBP1 KD (Figure 2c, increase LPS-induced miR-155-5p and decrease LPS-induced miR-155-3p), suggesting the role of FUBP1 and KSRP as miR-155-3p and miR-155-5p enhancers, respectively (summarized in Figure 8a). Our work indicates the KH3 domain of FUBP1 participates in binding to pre-miR-155, analogous to the requirement of the KSRP KH3 domain for miR-155-5p binding [28]. The KH3 domain of FUBP1 and KSRP share 79% sequence homology (Figure 8b). Thus KSRP and FUBP1 may recognize the same or similar sequences in pre-miR-155. The consensus recognition motif for the KSRP KH3 domain (GGGG) [35] and FUBP1 KH3 domain (UUGUG) [41] both appear at the 3΄ end of the miR-155-5p sequence of pre-miR-155 (Figure 8c). Future studies will examine whether FUBP1 and KSRP compete to interact with pre-miR-155 in this region. KSRP processes and controls the expression of miRNAs (miR-145, miR-150, let-7a) [29,42,43] other than miR-155, and the proposed mechanism involves recruitment of proteins such as Drosha, DGCR8, and Dicer [29]. Interestingly, all the miRNAs controlled by KSRP have the 5p strand as the dominantly expressed strand (guide strand) and the 3p strand as the passenger strand [44][45][46]. Thus, KSRP may be participating in strand selection by bringing specific proteins into the Dicer complex that enhances the thermodynamic stability [47] of the 5p strand binding to the Dicer protein. The increased strength may result in selective processing of the 5p strand into the mature guide strand. Conversely, the association of FUBP1 with pre-miR-155 may bring in proteins that favor the processing of the 3p strand.
We found that FUBP1 is required for both LPS induction of miR-155-3p and IL10 dependent inhibition of miR-155-3p expression (Figure 7). However, although the amount of FUBP1 protein associated with pre-miR-155 increases in LPS-stimulated cells (compared to unstimulated cells), the amount of FUBP1 protein observed in the pre-miR-155 pulldown from LPS or LPS + IL10 treated cells are about the same (Figure 1). Similar amounts of FUBP1 bound suggests the difference in the outcome of FUBP1-dependent, LPS vs. IL10 action on miR-155-3p must be mediated by a difference other than FUBP1 simply binding to pre-miR-155. Post-translational modification or association with other proteins might regulate FUPB1 in LPS vs. LPS + IL10 treated cells.
Future studies involve characterizing the mechanisms of FUBP1 and CELF2 control of pri-miR-155 and pre-miR-155 levels, FUBP1 regulation of miR-155-5p vs. miR-155-3p expression, and the possible competition of FUBP1 with KSRP. It will also be essential to identify the 3p and 5p target genes relevant to LPS and IL10 control of macrophage function. These investigations will also provide insight into the regulation of other miRNAs' expression, especially those also described to be regulated by KSRP [29,42,43,48].

Mouse colonies
SHIP1 WT or SHIP1 KO (in the BALB/c background) mice were provided by Dr. Gerald Krystal (BC Cancer Research Centre, Vancouver, BC). STAT3 WT and KO mice (in the C57BL/6 background) were generated as described [14]. All mice were maintained in accordance with the animal care protocols approved by the University of British Columbia Animal Care Committee.

Construction of the FUBP1-pLX-sgRNA targeting vector
The FUBP1 sgRNA targeting sequence (5' GCTAAATCCGACCATCCCATC) was designed using the CRISPR Gold online tool [51]. Oligonucleotides corresponding to this sequence were cloned into the pLX-sgRNA vector, using overlap-extension PCR as described [25]. The pLX-sgRNA vector with target-specific insert was transformed into chemically competent Stbl3 E. coli cells and colonies selected using ampicillin. The resulting FUBP1-pLX-sgRNA construct was confirmed by sequencing.
Clarified cell lysates were added to streptavidin magnetic beads (1164786001, Millipore-Sigma, Oakville, ON) in 1.5 mL microfuge tubes and incubated for 90 minutes at 4 o C on a nutator. The tubes were then briefly centrifuged at 5,000 RPM, and magnetic beads were immobilized using a magnetic tube stand (12321D, ThermoFisher Scientific, Nepean, ON). Lysates were removed, and the beads were resuspended in the wash buffer (0.1% Tween-20 containing PSB) and rocked for 5 minutes at 4 o C on a nutator. The washing was repeated 3 times. The proteins were eluted by boiling 2x SDS-PAGE sample buffer (0.125 M Tris, pH 6.8, 5% 2-mercaptoethanol, bromophenol blue, 13.5% glycerol, 4.5 % SDS) for immunoblot analysis.

Isolation of and stimulation of mouse peritoneal macrophages
Primary peritoneal macrophages (perimacs) were isolated from mice by peritoneal lavage with 3 mL of sterile PBS. Perimacs were seeded at 2.0 x 10 6 cells per well in a 6-well tissue culture plate or 1.18 x 10 6 cells per well in a 24-well tissue culture plate in Iscove's Modified Dulbecco's Medium (IMDM, SH30228, HyClone, Logan, UT) supplemented with 10% FCS. Cells were allowed to adhere for 2 hours, rinsed with room temperature PBS to remove non-adhered cells, and fresh media added. The media was changed after 1 hour, and the cells were stimulated with 1 ng/mL LPS ± 1 ng/mL IL10 for 1, 2, or 4 hours. Triplicate wells were used for each stimulation condition.

RNA extraction and qPCR
Total RNA was extracted using Tri-Reagent (T9424, Millipore-Sigma, Oakville, ON) according to the manufacturer's instructions. 1-3 µg of RNA was treated with RNase-free DNase I (04716728001, Millipore-Sigma, Oakville, ON) for 20 minutes at 37 o C, followed by the addition of 0.1 M EDTA to a final concentration of 8 mM to inactivate DNase I.
For measurement of pri-miR-155, pre-miR-155, and GAPDH, 200 ng of DNase I treated RNA were reverse transcribed using SuperScript™ IV reverse transcriptase and random hexamers (18090050, ThermoFisher Scientific, Nepean, ON). qPCR quantification of pri-miR-155, pre-miR-155, and GAPDH were achieved with primers for pri-miR-155, pre-miR-155, and GAPDH in conjunction with the SYBR Green master mix (100029284, ThermoFisher Scientific, Nepean, ON). RNA levels were analyzed using the comparative CT method with GAPDH as the normalization control.

Molecular cloning of FUBP1 and KSRP
The open reading frame (ORF) of mouse FUBP1 and KSRP was obtained by PCR on cDNA generated from RNA isolated from SHIP1 WT perimacs. The FUBP1/KSRP ORF was inserted into the Gateway entry vector (pENTR1A, Invitrogen, Burlington, ON) via restriction digest and ligation. The products were transformed into DH5α E. coli chemically competent cells, and colonies were selected using 50 µg/mL of kanamycin. The sequences of the FUBP1/KSRP pENTR1A vectors were confirmed by sequencing. The GXXG sequences in the third KH domain of FUBP1 and KSRP were mutated to the GDDG sequence using site-directed mutagenesis. Briefly, PCR with non-overlapping primers (5΄ phosphorylated forward primer with desired mutation and non-overlapping reverse primer) were used to generate FUBP1/KSRP daughter plasmids containing the desired mutation. The resulting PCR product was extracted with phenol-chloroform, treated with Dpn I to remove the parental vector, daughter plasmids ligated with T4 ligase, and transformed into DH5α chemically competent cells. The FUBP1/KSRP WT and KH3 GDDG mutant pENTR1A vectors were transferred to lentiviral vector FUGWBW using Gateway LR reactions as we have described [14].

Measurement of TNFα production
Cells were seeded at 2.0 x 10 4 cells per well in a 96-well tissue culture plate and allowed to adhere overnight. The media was changed the next day 1 hour before stimulation. Cells were stimulated with 1 ng/mL of LPS ± indicated concentration of IL10 for 1 hour. Triplicate wells were used for each stimulation condition. The supernatant was collected, and secreted TNFα protein levels were measured using a BD OptEIA Mouse TNFα Enzyme-Linked Immunosorbent Assay (ELISA) kit (558534, BD Biosciences, Mississauga, ON).

Biolayer interferometry(BLI)
The binding affinity between the FUBP1 or KSRP proteins to pre-miR-155 was examined using biolayer interferometry with super-streptavidin (SSA) biosensor tips (18-5057, ForteBio, Fremont, CA). SSA biosensor tips were hydrated in BLI assay buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.5 mM TCEP, 0.2% Tween-20) prior to coating with biotinylated pre-miR-155 and blocking with 0.1% BSA. The kinetic measurements were done at 30 o C with an orbital flow of 1000 rpm. A 60 second baseline was established using BLI assay buffer. The pre-miR-155 coated biosensors were then dipped into wells containing KSRP or FUBP1 protein and association monitored for 600 seconds. The sensors were then transferred to wells containing only BLI assay buffer, and protein dissociation was monitored for 600 seconds. The raw data were analyzed using the Octet Red Data Analysis software (ver. 8.2).

Statistical analysis
Band intensities were quantified in immunoblots using LI-COR Odyssey imaging system and Image Studio™ Lite software (LI-COR Biosciences, Lincoln, NE). Graph-Pad Prism 6 (GraphPad Software Inc., La Jolla, CA) was used to perform all statistical analyses. Statistical details can be found in figure legends. Values are presented as means ± standard deviations. Unpaired Student's t-tests were used where appropriate to generate two-tailed P values. Two-way ANOVA was performed where required with appropriate multiple comparisons tests. The differences were considered significant when p ≤ 0.05.

Ethical statement
The perimacs were derived from mice in accordance with the animal care protocol (A21-0203) approved by the University of British Columbia Animal Care Committee. The cell culture experiments are done in accordance with UBC Biosafety requirements (B16-0206). Institutional Review Board Statement: The perimacs were derived from mice in accordance with the animal care protocol (A21-0203) approved by the University of British Columbia Animal Care Committee.