Our original strategy entailed targeting exons 3 and 4 of the hnRNPD gene to debilitate the RRMs present in all four AUF1 isoforms. Mutant hnRNPD^{flx3,4/flx3,4} mice containing functional hnRNPD alleles amenable to loxP-mediated recombination ( Figure S1A ), were derived via gene-targeting manipulations of embryonic stem cells, germline removal of antibiotic selection cassettes and inbreeding to a C57Bl6/J background for 12 generations. To generate a mutant hnRNPD^Δx3,4 allele, we crossed hnRNPD^flx3,4/+ mice to a mouse line expressing germline-active Cre recombinase. Subsequently, hnRNPD^Δx3,4/+ mice were intercrossed to yield homozygous hnRNPD^{Δx3,4/Δx3,4} mice. Examination of the F2 progenies at post-natal day 10 (P10) revealed a skew in mendelian segregation relating to a 30% loss in the hnRNPD^{Δx3,4/Δx3,4} genotype ( Figure S1B ). However, hnRNPD^{Δx3,4/Δx3,4} embryos were properly detected till the embryonic day E14.5, suggesting that their post-natal loss was not due to early embryonic lethality. Moreover, the majority of hnRNPD^{Δx3,4/Δx3,4} mice identified post birth displayed a delay in weight gain ( Figure S1C ), and 50% succumbed during the two months of age ( Figure S1D ). The remaining that survived to adulthood displayed fertility issues and perished progressively past the 6-months of age. The analyses of AUF1-encoding mRNAs and proteins in extracts from hnRNPD^+/+ , and hnRNPD^{Δx3,4/Δx3,4} mouse embryonic fibroblasts (MEFs) indicated the near-complete loss of all isoforms ( Figure 1A ). Thus, the genetic removal of exons 3 and 4 did not yield an RRM deficient mutein but instead led to the diminished synthesis of AUF1s.

To identify whether AUF1s have specific functions in innate immune cells, we restricted the Δx3,4 mutation in the mouse myeloid linage via crossing the hnRNPD^{flx3,4/flx3,4} mice to a LysMCre⁺ line. Contrary to whole-body mutants, LysMCre⁺ hnRNPD^{flx3,4/flx3,4} mice (termed hereafter as M-hnRNPD^Δx3,4 or M-Δx3,4 mice) appeared phenotypically normal through a 12-month observation period. As in the case of hnRNPDΔ^x3,4/Δx3,4 MEFs, the reduction of AUF1 proteins in M-hnRNPD^Δx3,4 mice was verified in bone marrow-derived (BMDMs), thioglycolate-elicited peritoneal (TEMPs) macrophages and peritoneal Gr1⁺ polymorphonuclear cells (PMNs) but not in other myeloid derivatives like e.g. Siglec-F⁺ peritoneal eosinophils ( Figure 1B and Figure S1E ).

Published data on obligatory AUF1 deficient mice demonstrated their increased susceptibility to the systemic administration of lipopolysaccharides (LPS; endotoxins) from Gram-negative bacteria. LPS activates the canonical TLR4 pathway in immune and non-immune cells triggering an excessive cytokine storm (15). This leads to a variety of cytokine-induced danger signals in tissues and changes in physiology. Depending on the LPS dose, such changes range from hypothermia to systemic organ failure and lethality. To test whether our mutant mice responded to endotoxemia in an exaggerated manner, we challenged them first with a low dose of LPS (4mg/kg) and monitored changes in body temperature. In hnRNPD^+/+ mice, this dose elicited a temperature drop till the 4^th hour post treatment which then recovered by the 6^th hour ( Figure 1C ). Under the same conditions, the hypothermic response of hnRNPD^{Δx3,4/Δx3,4} mice verified their enhanced susceptibility. The drop in their body temperature was not transient but continued past the 4^th hour leading to a moribund state which necessitated the termination of the experiment. Surprisingly, the hypothermic response of M-hnRNPD^Δx3,4 mice to the same treatment was as transient as in the case of their control and of hnRNPD^+/+ mice ( Figure 1D ). Similarly, the susceptibility of M-hnRNPD^Δx3,4 mice to higher doses of LPS prolonging hypothermia (12mg/kg) or leading to sub-lethality (20mg/kg) was comparable to that of their corresponding controls ( Figure 1D ).

The different response of the whole-body to the myeloid mutants could reflect changes in the spectrum of inflammatory mediators elicited by LPS. To address this, we measured 10 prototypical mediators in the sera of all mice challenged with the low dose of LPS for 90mins ( Figure 1E ). Eight of these mediators (TNF, IL6, IL12b, IL1β, IL10, and the chemokines CCL2, 17&22) were detected in a similar range between the different control groups; whereas two (TGFβ1 & CXCL1) showed intergroup variations and were thus excluded from further comparisons.

In our analysis we noted selective quantitative differences between the mutants. hnRNPD^{Δx3,4/Δx3,4} mice possessed higher pro-inflammatory IL1β, whereas this was not observed in M-hnRNPD^Δx3,4 mice. The secretion of IL1β is elicited by inflammasomes which can sense signals from degenerating tissues (28, 29). Given the previous foreground on AUF1s control over senescent or homeostatic programs (17), the exacerbated response of the whole-body mutants to LPS might be a secondary consequence of inflammatory tissue degenerations enhanced by the loss of such programs in afflicted tissues.

Notably, the sera from the challenged M-hnRNPD^Δx3,4 mice revealed augmentations in pro-inflammatory TNF and IL12b. The mRNAs encoding these cytokines are subject to ARE-mediated control, and their augmented presence connects to the enhanced LPS-response of mouse mutants lacking other ARE-BPs like Zfp36/TTP, TIA1, or Elavl1/HuR (4, 11). Unlike those mutants, the M-hnRNPD^Δx3,4 mice did not display an exacerbated response to LPS despite the presence of higher TNF and IL12b; whereas the quantity of these cytokines was not altered in hnRNPD^{Δx3,4/Δx3,4} mice. These observations may be taken to suggest that (a) AUF1s have a more limited role upon the secretion of TNF and IL12b which may be masked by the pleiotropic events underlying the response of the whole-body knockouts, and (b) the activities of TNF and IL12b in our myeloid mutants could be compensated due to the effects of AUF1s’ loss upon other myeloid-derived effectors or inhibitors acting downstream of these cytokines.

The invariable, yet puzzling, systemic response of our myeloid AUF1 mutants to endotoxin prompted us to screen them for more restricted, cell-mediated and organ specific inflammatory responses. We started with two well-established models of intestinal inflammation elicited by chemical irritants. In the first model, inflammation in the colon is induced by the addition of dextran sodium sulfate (DSS) in the drinking water of mice for 7 days (30). This disrupts the colonic epithelial barrier allowing entry of commensal bacterial antigens and activation of underlying innate immune cells. The inflammatory stimulus is subsequently removed to allow for the resolution of inflammation and the restitution of the epithelium. Clinical symptoms (i.e. weight loss, diarrhea and bloody stools) were monitored daily for 15 days after the initial exposure to DSS, with endoscopic evaluation of the colon on day 8 and histological assessment on days 3, 6 and 13 representing the initiation, peak and restitution phases, respectively. The evaluation of the clinical and endoscopic parameters revealed that the disease activity of the M-hnRNPD^Δx3,4 mice, was indistinguishable from that in the controls ( Figures 2A, B ). Similarly, the histological assessment verified that the inflammatory, degenerative and restitution phases in the afflicted colons progressed invariably between mutant and control groups ( Figure 2C ). To test whether M-hnRNPD^Δx3,4 mice could be differentiated via the cytokine variations we detected in their LPS response, we sought for disturbances in IL1β, TNF, IL6 & IL10 that are commonly engaged in systemic and mucosal inflammation. However, we failed to detect any statistical differences in the secretion of these cytokines by colonic explants isolated from mice on the 13^th day of the challenge ( Figure S2 ). In the second model, colitis was induced by the haptenating agent 2,4,6-trinitrobenzene sulfonic acid (TNBS). This model is more complex because it relies on Type I adaptive immune responses encompassing the functions of Th1 T-helper subsets (31). Yet, innate immunity is paramount in this model as some form of clinical disease can occur even in the absence of lymphocytes. For the model, TNBS in ethanol was administered intrarectally in mice 7 days after a peripheral skin pre-sensitization, whereas a parallel group received only the ethanol vehicle (32). Weight changes were monitored for 5 days after administration of TNBS ( Figure 2D ), with parallel endoscopic evaluation on day 3 ( Figure 2E ) followed by histopathology at day 5 ( Figure 2F ). The kinetics of weight loss and recovery, the endoscopic scoring and the histological analysis of afflicted colonic tissue were nearly identical in control and M-hnRNPD^Δx3,4 mice ( Figure 2E ).

Collectively our data indicate that myeloid AUF1s are dispensable for the proinflammatory and Type I cell-mediated immune reactions which facilitate intestinal inflammation in models of chemically-induced colitis.

Next, we screened for changes in Type II cell-mediated responses such as those observed in allergy and hypersensitivity reactions (33). To do so we explored an animal model of human asthma where lung inflammation is elicited post the systemic sensitization of mice to Ovalbumin (OVA) and subsequent to a local challenge with aerosolized OVA. This leads to a skewed Th2 response, the production of OVA-specific IgE, the elicitation of eosinophilic lung inflammation, and airway obstruction. In this model, immunoregulatory innate immune cells promote the recruitment of eosinophils and lymphocytes; and sustain the Th2 feedback loop while counteracting pro-inflammatory, neutrophilic and Th1 responses.

Relative to mice challenged with PBS, both Control and M-hnRNPD^Δx3,4 mice challenged with OVA mounted the allergic response. However, the lungs of challenged M-hnRNPD^Δx3,4 mice displayed less peribronchial (PBI) and perivascular (PVI) inflammation, reduced fibrosis, and decreased bronchial mucus metaplasia relative to the lungs of control mice ( Figures 3A, B ). This correlated with a significant reduction in OVA-specific IgE in the sera of M-hnRNPD^Δx3,4 mice ( Figure 3C ) and a dramatic decrease in cells infiltrating their Bronchioalveolar Fluid (BALF; Figure 3D ). Flow cytometric analysis of BALF exudates from the M-hnRNPD^Δx3,4 lungs revealed that this numeric reduction was not only restricted to eosinophils but extended to all central myeloid and lymphoid infiltrates ( Figure 3E and Figure S3 ).

The assessment of BALFs for T-helper 2 lymphokines ( Figure 3F ) revealed a significant reduction of IL-4 in the lungs of challenged M-hnRNPD^Δx3,4 mice whereas other lymphokines engaged in the allergic response were either augmented (e.g. IL-5, IL-13) or remained unchanged (e.g. IL-9). However, we detected significant elevations in the Th1-lymphokine IFNγ and of pro-inflammatory TNF. Together, our data indicate that the loss of AUF1s functions in myeloid cells restricts the progression of allergic airway infiltration and skews T-helper functions.

The resistance of M-hnRNPD^Δx3,4 mice to allergic airway inflammation as opposed to their normotypic response to endotoxemia and colitis, suggested that in innate cells, AUF1 does not act as an anti-inflammatory factor. Instead it appears to be engaged in signal-induced programs balancing cell-mediated immune reactions.

To gain further insight into this, we focused our subsequent analyses on macrophages as a prototypical innate-immune subset. Given the prior connection of AUF1s to pro-inflammatory control, we sought for comprehensive changes in RNA expression incurred by the loss of this RBP in macrophages activated by the predominant TLR4 ligand -i.e., LPS- for 4hrs. We decided to assess only mature mRNAs associated with the response without considering immature, fragmented and non-coding RNAs. To do so, we used 3′ end sequencing of transcripts (Quantseq) and performed our analysis on transcript rather than on gene level. Our search for changes (with a log2 Fold Change of 1 and p value<0.05) between the profiles of M-hnRNPD^Δx3,4 and control BMDMs showed an unexpectedly small number of differentially expressed transcripts (424 transcripts corresponding to 204 genes) ( Figure 4A and Table S1 ). Moreover, 2/3 of the transcripts appeared as downregulated in AUF1-deficient macrophages ( Figure 4B ), contrasting the original hypothesis on AUF1s’ as primary decay-promoting factors. By inspecting the dataset for changes in hnRNPD RNAs we verified the reduction of three RNA isoforms; and in three RNAs derived from Nonsense-mediated decay (NMD) ( Figures 4A, C ). Notably, the RNAs of other isoforms including that of the larger p45 ( Figure 4C ) were poorly expressed in macrophages suggesting the predominance of the p37, p40 & p42 isoforms in these cells.

To explore further the biological consequences of these differences in macrophage activation, we performed functional analysis using WebGestalt (2019). Ten weighted terms were identified as significant with an FDR<0.05 ( Figure 4D ). As expected, the term “negative regulation of gene expression” was significantly enriched in the dataset and included RNA regulators involved in nuclear and cytoplasmic events besides hnRNPD (e.g. Elavl1, Eif4a3, hnRNPs A2/B1 & U, Pum1). This implied that AUF1s may act in the apex of several regulons for RNA control. However, these regulons did not connect to changes in cytokines. Rather they connected to (a) metabolic changes such as those relating to nitrogen containing biomolecules or ATP; (b) negative effects of stress in protein modification or use and (c) extracellular signals transmitted by biotic stimuli such as those produced by the same or other cells.

We postulated that AUF1s may regulate innate immune responses depending upon signal-induced changes in cellular metabolism as in those inducing the wide spectrum of polarized macrophage phenotypes. Extremes of this spectrum can be elicited in vitro through different signaling combinations supporting either the classical pro-inflammatory phenotype (M1-like); or a multitude of alternative phenotypes (M2-like; M2a,b,c,d) (34). M1-like macrophages are driven by Toll-like receptor (TLR) and interferon signals and use Nitric Oxide Synthase 2 (NOS2) to metabolize arginine to nitric oxide (NO) which can be further metabolized to downstream reactive nitrogen species. Functionally, they mount pro-inflammatory type-I and Type III immune responses against bacteria, intracellular pathogens, and tumor cells and support pathologic tissue damage like in sepsis and IBD and cellular transformation if uncontrolled. Differently, the major subset of M2 macrophages is driven by IL-4 (and/or IL13; M2a) and use Arginase 1 (ARG1) to hydrolyze arginine to urea and ornithine for the support of polyamine and proline synthesis. Functionally, M2 macrophages inhibit M1-like pro-inflammatory responses, promote vascular remodeling, tissue regeneration and helminth control; but also support pathologic Type II immune responses (e.g., as in allergy), tumor vascularization and growth (35, 36).

In response to signals promoting a classical M1 phenotype (i.e., LPS+IFNγ) the total levels of hnRNPD transcripts were reduced ( Figure 4E ). Still, M1 polarized M-hnRNPD^Δx3,4 BMDMs secreted TNF, IL6, IL12, IL10 and expressed the Nos2 and several chemokine mRNAs at levels comparable to control macrophages and only under the specific signaling regime ( Figure 4F and Figure S4 ). Intriguingly, and in response to the M2-promoting cytokine IL-4, hnRNPD transcripts were also reduced ( Figure 4E ); and M-hnRNPD^Δx3,4 BMDMs elevated properly the characteristic Arg1, Mrc1, Retnla/Fizz1, Chi3l3/Ym1 mRNAs whereas M1 markers remained silent ( Figure 4F and Figure S4 ).

Together, our ex vivo data indicate that AUF1s may not be required to establish central macrophage activation and M1 or M2a-like polarization programs. However, their connection to regulatory catabolic programs may be linked instead to their requirement for transition between programs occurring in other polarization settings.

In vivo, M1- and M2- like macrophages co-exist in inflammatory settings, and their plasticity allows them to switch their phenotype in response to microenvironment derived cues. Apart from the highly polarized macrophages induced by LPS ± IFNγ or IL4, several different alternative subsets have been described in vitro. These subsets presumably mimic phenotypes switching in response to particular tissue signals that block one phenotype for another to occur (25, 37). For example, when LPS-primed macrophages are exposed to the combination of homeostatic TGFβ and anti-inflammatory IL10 they “deactivate” proinflammatory mediators and acquire an M2-like identity (also known as M2c) (38). In response to TGFβ and IL10, the expression of hnRNPD mRNAs was as low as in the case of macrophages treated with LPS; still, LPS-activated M-hnRNPD^Δx3,4 BMDMs reduced their secretion of TNF, IL-6 and IL-12 while upregulating the M2-associated mRNAs of Arg1, Chi3l3 and Retnla ( Figure 5A ) in a manner similar to control macrophages.

Another macrophage switching effect is observed, when the M1 macrophages are exposed to purinergic signals stimulated by adenosines (described loosely as M2d) (39). Adenosines bind to A2A receptors to suppress pro-inflammatory mediators while enhancing the expression of M2 markers; but also augment the expression of vascular remodeling factors like VEGFα and one of its downstream regulators, Thrombospondin-1 encoded by the Thbs1 mRNA (40, 41). To explore the response of LPS-activated macrophages to adenosines, we used the specific adenosine A2A receptor agonist, CGS-21680 (CGS). In control BMDMs primed with LPS, CGS suppressed the expression of TNF, IL-12b and IL10 but not of M1-related chemokines; induced the secretion of VEGF and the expression Arg1, Chil3l3, Retnla and the Thbs1 mRNAs; and -strikingly- augmented the expression of hnRNPD transcripts ( Figure 5B and Figure S5 ). On the contrary, M-hnRNPD^Δx3,4 BMDMs exposed to LPS+CGS failed to downregulate specifically TNF and IL12 and were unable to augment properly VEGFα and the, Thbs1 and Arg1 mRNAs ( Figure 5B and Figure S5 ).

The compromised expression of VEGFα & Thrombospondin as opposed to the enhanced TNF by M-hnRNPD^Δx3,4 BMDMs suggested that AUF1s may be involved in the macrophage specific support of vascular permeability of inflamed tissues. Indeed, we noted that the reduction of hnRNPD expressing cells in the BALF of OVA challenged M-hnRNPD^Δx3,4 mice related to comparable reductions in the Vegfa and Thbs1 mRNAs ( Figure 5C ).

Three pieces of evidence suggested that AUF1s control selective transcripts rather than adenosine receptor signaling. First, the induction of Chil3l3 and Retnla mRNAs and the reduction of secreted IL10 in M-hnRNPD^Δx3,4 BMDMs were comparable to those in control macrophages ( Figure 5B ) Second, the expression of the Adora2a mRNA encoding the A2A receptor was enhanced 150-fold in both control and AUF1 lacking macrophages exposed LPS+CGS and 50-fold in those treated with LPS whereas it was only minimally affected by other macrophage signals ( Figure 5D ). Similarly, the loss of AUF1s did not affect the macrophage expression of Adora2b mRNA encoding the A2B receptor that may act synergistically to A2A and which was exclusively enhanced 5-fold by LPS+CGS ( Figure 5D ). Finally, the examination of downstream PKA activity which is commonly activated by LPS and LPS+CGS signals was not significantly altered by the loss of AUF1s ( Figure S6 ).

These data indicate that AUF1s are required downstream of adenosine signalers to control specific transcripts during the switch of pro-inflammatory macrophages to those supporting vascular remodeling.

M2-like pro-angiogenic macrophages usually appear as tumor-associated infiltrates (TAMs) to support tumor angiogenesis and growth (42). To connect the functions of myeloid AUF1s to tumor growth we used the model of B16 melanoma tumors transplanted subcutaneously in the flanks of syngeneic mice. Contrary to tumors in control mice, the tumors in M-hnRNPD^Δx3,4 mice grew to smaller volumes at the endpoint ( Figure 6A ), posing difficulties in their comparative evaluation of vessel density. In an effort to gain insight on tumor vascularization, we selected whose volume was proximal to the lower range of control tumor volumes compare their vessel density as marked by endothelial marker, CD31 (PECAM). As shown in Figure 6B , a reducing trend in the number of CD31⁺ vessels were observed in the tumors from M-hnRNPD^Δx3,4 mice but their low number precluded statistical validation. To gain indirect insight we assessed the inflammatory infiltration of all tumors following dissociation and analyses of their immune infiltrates via flow cytometry. The percentage of total immune CD45+ tumor infiltrates in M-hnRNPD^Δx3,4 mice was reduced 4-fold relative those in the tumors of control mice indicating a possible defect in tumor vasculature impeding immune infiltration ( Figure 6C ). We also noted reductions amongst the myeloid CD45+ cells expressing high levels of Gr1+ which includes neutrophils and myeloid-derived suppressive cells; and on CD11b+ Gr1 negative subsets suggesting limitations in the general myeloid influx.

To get a direct answer as to whether the loss of AUF1s in myeloid cells limits their effect on blood vessels, we exploited an in vivo angiogenic model measuring neovascularization directly (43). In this model, inert biomaterials in a sponge-like form were engrafted subcutaneously in the flanks of mice and were injected either with PBS or recombinant VEGFα every 2 days. At the 14^th day sponges were excised and processed for the immunohistochemical enumeration of blood vessels using the endothelial marker endomucin. As demonstrated in Figure 6D , the vessel density of PBS-treated sponges engrafted in M-hnRNPD^Δx3,4 mice was nearly half of those in the grafts of control mice. In contrast, the VEGFα-treated sponges possessed an equally elevated blood vessel density in both the grafts in control and mutant mice.

Our analysis on macrophages suggested that AUF1s may affect a transcript specific regulon to facilitate the functional switch from the pro-inflammatory to the pro-angiogenic state. To gain insight on how AUF1s modulate such transcripts, we focused on those that appear as most affected by the loss of AUF1s’ functions, namely the Tnf, Il12b, Vegfa and Thbs1 mRNAs (44). First, we immunoprecipitated AUF1-containing RNPs (RIP) from cytoplasmic extracts of resting, LPS or LPS+CGS treated BMDMs for 4hrs using a specific antibody recognizing all four AUF1 isoforms. To estimate background, we performed parallel immunoprecipitations with an isotype-matched antibody. The efficiency of the RIP was monitored using immunoblots for the AUF1 proteins as antibody-bound (e.g. Figure S7A ). Subsequently, the RNA content of the RNPs was analyzed via qRT-PCR and values were normalized to the levels of expression of each mRNA ( Figure 7A ). Enrichments above 2-fold were considered as significant. This analysis demonstrated that AUF1:RNPs associate only with the Vegfa mRNA in resting macrophages; with the Tnf mRNA in LPS stimulated macrophages; and most strongly with the Il12b, Vegfa and Tnf mRNAs in adenosine-stimulated macrophages. In contrast we did not observe associations with the Ccl2 mRNA which we used as a relevant control; neither with the Thbs1 mRNA indicating that its reduction could be due to the loss of eliciting VEGF (45).

Further comparison of the levels of these mRNAs in extracts from Control BMDMs revealed that CGS suppressed the LPS-induced expression of the Il12b mRNA whilst augmenting that of the Vegfa mRNA; yet it failed do so in M-hnRNPD^Δx3,4 macrophages ( Figure 7B ). These data correlated to the differences we observed in the secretion of IL12b and VEGF by these cells ( Figure 5B ). Notably, and in contrast to its protein ( Figure 5B ), the Tnf mRNA was properly suppressed by CGS in both control and M-hnRNPD^Δx3,4 BMDMs ( Figure 7B ). We postulate that AUF1s either control the translation of the Tnf mRNA or the post-translational release of the TNF protein.

The analysis of the decay of these mRNAs following the arrest of their transcription by Actinomycin D was most revealing ( Figures 7C , S7B ). CGS had a potent destabilizing effect upon the il12b mRNA which was ablated in the absence of AUF1s. In contrast the decay of the Vegfa mRNA was not affected by the loss of AUF1s suggesting that AUF1s are required for its nuclear synthesis as promoted by CGS. Finally, the decay of the Tnf mRNA was not affected neither by the loss of AUF1s nor by CGS. Taken together these data indicate that AUF1s are involved in various transcript specific RNPs that respond to adenosine signaling which independently promote mRNA decay (e.g. Il12b mRNA), nuclear synthesis (e.g. Vegfa mRNA) or translation (e.g. Tnf mRNA).

The targeting of the hnRNPD locus in murine ES cells was performed by Regeneron Pharmaceuticals using their proprietary Velocigene recombineering technology (66). Modifications included the flanking of exons 3 and 4 with loxP sites; the inclusion of and FRT-flanked neomycin resistance gene for selection proximal to the loxP of exon 3. These ES cells were integrated onto blastocysts from C57Bl6/J mice. Derivative mice where then crossed to mice expressing a germline encoded Flipase gene to remove the neomycin cassette and yield mice containing the hnRNPD^flx3,4 allele. To generate a mutant hnRNPD^Δx3,4 allele, hnRNPD^flx3,4/+ mice were crossed to a mouse line expressing germline-active Cre recombinase (67). Subsequently, hnRNPD^Δx3,4/+ mice were intercrossed to yield homozygous hnRNPD^{Δx3,4/Δx3,4} mice. For the myeloid restriction of the mutation hnRNPD^{flx3,4/flx3,4} mice were crossed to the LysM-MCre⁺ line. All mouse lines were maintained in a C57BL/6J background and in the animal facilities of the Biomedical Sciences Research Center (BSRC) “Alexander Fleming” under specific-pathogen free conditions. All experiments were performed with 8 to 16 weeks old mice and in accordance to the recommendations of the BSRC’s Institutional Committee for Protocol Evaluation, the Veterinary Authorities of the Prefecture of Attika, the national legislation and the European Union Directive 63/2010. Protocols were approved by the Veterinary Authorities of the Prefecture of Attika (licenses # 4376/2014, 262781-22/04/2020, 26497-13/01/2020, 6197-21/11/2017, 4401-5/7/2016, 2823-13/06/2018, 531-13/2/2019, 26488-13/01/2020).

Mice 6- to 8-week-old were injected intraperitoneally (i.p.) with the indicated amount of LPS (LPS from Salmonella enterica serotype enteritidis; Sigma-Aldich, L6011). Core body temperature and survival were monitored through a period of 7 days to determine mice sensitivity to endotoxemia. For measurement of inflammatory mediators, hnRNPD^{Δx3,4/Δx3,4} and hnRNPD^Δx3,4 mice were injected i.p. with LPS at 4mg/kg of body weight. Sera were collected 90 minutes later by cardiac puncture and cytokines were quantified using LEGENDplex™ Mouse Th Macrophage Panel (Biolegend, Cat. No.740846).

Mice 6- to 8-week-old were fed ad libitum for one cycle with water containing 2% (wt/vol) DSS (MW 40,000 kDa; MP Biomedicals Inc.) for 7 days followed by 8 days with regular water without DSS (68). Mice were monitored daily for changes in body weight, stool consistency and rectal bleeding. These values were used to calculate the Disease Activity Index (DAI). Endoscopy was performed using Coloview endoscopic system (Karl Storz, Germany) at day 8. Colonic inflammation was scored according to MEICS (Murine Endoscopic Index of Colitis Severity) as described (69). Mice were sacrificed at indicated time points or at the end of the protocol for the isolation of colonic tissue. Histological scoring was performed on hematoxylin and eosin (H&E) stained colon sections as previously described (70). Scores of inflammation and epithelial damage were summed to produce the overall histological score. All histological assessments were performed in a blinded fashion, by two independent investigators.

Mice 6- to 8-week-old were pre-sensitized applying TNBS solution (1% w/v in acetone and olive oil solution 4:1) epidermally at day -7. On day 0, mice were inoculated intra-rectally with 100μl TNBS solution (2.5% w/v in absolute ethanol). Weight loss was daily monitored from day 0 to 4. Endoscopy was performed the 3rd day when severe symptoms peaked as described above (32).

Airway inflammation was induced by ovalbumin (OVA-sensitization) as previously described (54). Briefly, sex matched mice were pre-sensitized on days 0 and 15 via intraperitoneal injection of 10μg chicken egg-derived OVA (Sigma Aldich, A5503) mixed with 1mg of aluminum hydroxide (Alum) as an adjuvant dissolved in 300 μl phosphate buffered saline (PBS) on day 0 and 15. Subsequently, mice inhaled 1% OVA aerosol for 30 min/day from day 21 to 25. At the end point, bronchoalveolar lavage fluid (BALF) was collected from the lungs by gently washing airway lumina with 0.5 mM EDTA in PBS via a tracheal cannula. BALF cells were analyzed by flow cytometry (strategy of analysis is presented in Figure S3 ), while supernatants were used for cytokines analysis using the LEGENDplex™ Mouse Th Cytokine Panel (Biolegend, 741044). To determine the effectiveness of the immunization, anti-OVA IgE levels were measured in the serum of PBS and OVA challenged mice using Anti-Ovalbumin IgE Elisa Kit (Cayman, 500840), according to the manufacturer’s instructions. Lungs were collected, fixed in 10% formalin (VWR, 9713.500) and proceeded for histopathological examination. General morphology and inflammation were examined in H&E, Periodic acid Schiff (PAS)-H&E, and Trichrome Masson stained lung sections. Perivascular (PVI) and peribronchiolar (PBI) inflammation were assessed using a four-tier score system (0=none, 1=mild, 2=moderate, 3=severe, 4=tissue destruction) with an increment of 0.5 if the inflammation fell between two integers.

Mice were injected subcutaneously in the flank with 10⁶ B16-F0 melanoma cells in 100ul PBS. Tumors were grown for 14 days. Post-mortem volume was determined using all 3 dimensions (m1×m2×m3×π/6). To prepare Single cell suspensions, tumor tissues were gently removed, cut into small pieces and incubated with 50 μg/ml Collagenase/Dispase (Roche, 10269638001) and 200 Units/ml DNAse I (Sigma-Aldrich, D5025) in RPMI 1640 (Gibco, 31870025) for 30 min at 37°C. Enzymatic dissociation was stopped with 10% FBS. Cells were washed with PBS and stained with the anti-mCD45, anti-mCD11, anti-mGr1, anti-mCD3 and anti-CD45R/B220 flow cytometry antibodies to identify immune subsets infiltrated in tumors. FACS analysis strategy is presented in Fig 6B.

Growth factor induced angiogenesis assay was performed as described previously (43). Briefly, subcutaneously implanted sponges of 1 cm³ size were injected with 100μl of either PBS alone (as negative control) or PBS containing 10 ng/ml^-1 VEGF (VEGF 164, R&D,493-MV), every 2 days. Sponges were excised, and paraffin embedded after 14 days.

For immunohistochemistry, all tissues were embedded in paraffin blocks. Thick sections (4μm) of embedded colon and lung tissues were stained with H&E. PAS staining and Masson trichrome were performed with standard protocols to determine fibrosis and mucus secretion in lung tissues. Blood vessels were identified by immunofluorescence staining using primary antibody against Endomucin antibody [santa cruz, sc65498/V.7C7.1] followed by secondary anti-rat 546 Alexa (A11081) and vessel density was calculated as the number of blood vessels per mm² of sponge section area using ImageJ software. Representative images were obtained on a Nikon ECLIPSE E200 microscope with a Nikon Digital Sight DS-5M digital camera.

Bone Marrow Derived Macrophages (BMDMs) were isolated from tibiae and femora of 6-8 week-old mice. Isolation was performed as previously described by Warren and Vogel (1985) and then differentiated to Bone Marrow Derived Macrophages in RPMI 1640 medium (Gibco, 31870-025) supplemented with 5% FBS (Gibco, 31870-025), 1% penicillin/streptomycin (Gibco, 15140122), 1% L-glutamine (Gibco, 25030024) and 10% L-medium (produced by L929 cells) incubated for 10 days at 37°C with 5% CO_2. Exudate macrophages, neutrophils and eosinophils were isolated from mice 48 or 72 hours upon intraperitoneal injection of 1 ml Brewer thioglycollate medium (4%; Becton-Dickinson).

BMDMs were activated with LPS (100 ng/ml, Salmonella enterica enteritidis; Sigma L6011), IFN-γ (10 ng/ml, Peprotech 315-05), IL-4 (10ng/ml, Peprotech 214-14), CGS (10μM, Sigma, 21680), IL-10 (10ng/ml, Peprotech, 210-10) and/or TGFβ1 (10 ng/ml, Cell-signaling, 231LC) for 24 hours.

For surface Single cell suspensions were prepared from either in vivo or in vitro experiments. For surface antigens, staining was performed using standard procedures. Antibodies were from BD Biosciences CD11b (M1/70); CD34 (RAM34); CD4 (H129.19); CD8 (H129.19), siclecF (E50-2440); eBioscience (F4/80), (CD16/32 (93), ScaI (D7), Biolegend GR1 (RB6-865), CD11C (N418); CD3e (145-2C11), CD206 (CO68C2), CCR3 (144505), Zombie NIR; B220 (RA3-6B2), Lin (mouse lineage panel, BD Pharmigen-559971). Cell viability was assessed by Propidium Iodide Staining. Cells were detected using FACSCantoII flow cytometer. Data were analyzed using FACSDIVA (V.6; BD Biosciences) and V10; Treestar, FlowJo).

The cells were lysed in ice-cold lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.25% SDS, 10% Glycerol, 1% NP-40 and one tablet of complete protease inhibitors) for 15 min on ice. Lysates were analyzed on SDS-polyacrylamide gels (SDS-PAGE; 12-14%), along with protein molecular weight markers (Nippon MWP03, Fermentas, Thermo Scientific, SM0431) and blotted onto nitrocellulose membrane (GE Healthcare). After blocking with 5% milk in TBS-Tween 20 buffer, membranes were incubated with primary and HRP-conjugated secondary antibodies (Biotech, anti-rabbit HRP 1030-05 or anti-mouse HRP 4050-05) signals were visualized by enhanced chemiluminescence (ECL, GE Healthcare) using films or a ChemiDocTMXRS+ System with Image LabTM software. Primary antibodies used: anti-AUF1 (milipore, 07-260) and anti-GAPDH (Ambion, 6C5). Cell culture Supernatants or serum were analyzed via specific ELISA kits of TNFa, IL-10, IL-6, VEGF, IL12 (Peprotech 900-K54, 900-K53, 900-K50, 900-K99), IL-1b (Thermo-Fisher, 88-7013-88), Anti-Ovalbumin IgE (Cayman 500840) or cytometric bead-based assay panels following manufacturer’s instructions Legendplex Panels (Biolegend, Cat. No.741044 and 740846). Protein levels were normalized to cell number using crystal violet staining.

BMDMs were stimulated with LPS or LPS and CGS for 4 hours. Cells were harvested and lysed in 300μl lysis buffer containing 100 mM KCl, 25 mM EDTA, 5 mM MgCl2, 10 mM HEPES, 0.5% NP40, 2 mM DTT, 0.2%, protease inhibitors cocktails, Vanadyl Ribonucleoside Complex (Invitrogen) and 100 U/ml RNAse OUT (Invitrogen). A small quantity of samples was kept aside as Input (total cytosolic) for Western Blot analysis and for total cytosolic RNA (RNA-input). Antibody uncoated beads (protein G), (Dynabeads™ Protein G Immunoprecipitation kit, ThermoFisher), were washed and maintained in 600 µl of NT-2 buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM MgCl2, and 0.05% NP40). For IP, 400 µl of cell lysate were loaded onto the beads and incubated overnight on a rotary at 4°C. Subsequently, beads were washed four times with NT-2 buffer, one of which contained 0.7M urea and finally resuspended in NT2 buffer. Fifteen microliters of the total volume of samples were removed for immunoblot verification while the remaining volume was used for RNA isolation via TRI reagent. For immunoblotting, polyclonal anti-AUF1 antibody (Millipore 07-260) and a IP conformation-specific secondary antibody (anti-mouse IgG -HRP-conjugated Veriblot for IP Detection Reagent, Abcam, ab131366) were used.

Stability of the mRNAs of interest was assessed by the addition of RNA polymerase II inhibitor, actinomycin D (10 μg/ml, Sigma) into cell culture for a period of 240 min. Total RNA was isolated at the indicated time points using TRI reagent and quantified using qRT-PCR. The half-live of transcripts was calculated using a fitted exponential curve. High undetermined values are arbitrarily defined.

Total RNAs were extracted from cells using TRI Reagent (MRC, TR 118) according to the manufacturer’s protocol. For cDNA synthesis 0.5-2μg of total RNA was reverse transcribed using MMLV-RT (Promega, M1705) and Oligo dT (NEB, S1316S) at 37°C for 1h. qRT-PCR was performed using EvaGreen SsoFast mix (Bio-Rad, Hercules, CA, USA, 172-5201) on a RotorGene 6000 machine (Corbett Research, Qiagen). Relative mRNA expression was normalized to β2-microglobulin (B2M) and calculated as the difference between the test values and the control values assigned as 1, using Bio-Rad RelQuant (Bio-Rad). Primers are listed in Supplementary Materials.

RNA Total RNA in expansion phase was collected using TRI Reagent (Invitrogen, Life technologies). RNA was extracted according to the manufacturer’s instructions and resuspended in 30μls of nuclease free water. Total RNA was quantitated using the NanoDrop ND1000 Spectrophotometer. Samples were diluted accordingly to a mean concentration of approximately 150 ng/μl, on an area of 235.0 and their quality assessed on a Bioanalyzer (Agilent Technologies) using the Agilent RNA 6000 Nano Kit reagents and protocol (Agilent Technologies). Sequencing libraries were prepared by using the QuantSeq 3’ mRNA-Seq Library Prep Kit FWD (QuantSeq-LEXOGEN™, Vienna, Austria) and RNA-seq libraries were sequenced on an Ion Proton PI™ V2 chips, an IIon Proton™ System. Library preparation and alignment strategies are provided in Supplementary Material . Mapped sequences Quant-Seq BAM files were analyzed on a transcript level base with the Bioconductor package metaseqR2 (71) v.1.3.14 which has built-in support for Quant-Seq data. Differential expression analysis for the contrasts MΔχ3,4 versus Control was performed using DESeq2, edgeR, NOISeq, limma, NBPSeq, ABSSeq and DSS algorithms and a combined meta-analysis procedure by the PANDORA algorithm implemented in metaseqR2. 3’ UTR areas (and their corresponding transcripts) presenting a PANDORA p-value less than 0.05 and an absolute fold change (for each contrast) greater than 1 in log2 scale were considered as differentially expressed. Categorical enrichment analysis of differentially expressed transcripts was implemented with WebGestaltR6 version 0.4.4 R package. The method that was used was the Over-Representation Analysis (ORA) against the Biological Process functional database of Gene Ontology and Kyoto Encyclopedia of Genes Genomes (KEGG) pathways and the terms retrieved were identified at a significance level of FDR ≤ 0.05. Multiple testing adjustment was carried out with the Benjamini-Hochberg procedure. All data extracted from sequencing datasets and gene sets were plotted in R (v.4.1.0) using ggplot27 (v.3.3.5) package. Z-scores were calculated as a derivative of the normalized counts.

Statistical analysis was performed on Graphpad Prism 6.01 (Graphpad Software). Statistical significance was determined using unpaired Student’s t-test or One Way Anova when comparing three or more groups. Survival analysis was performed using Kaplan-Meier statistics. Results with a P-value 0.05 considered to be statistically significant.