Immunopathological and microbial signatures of inflammatory bowel disease in partial RAG deficiency

The immune system shapes microbial composition and controls the maintenance of host–microbe symbiosis (Ansaldo et al., 2021). Likewise, the microbiota influences the development and function of the host’s immunity (Hooper and Macpherson, 2010). This reciprocal relationship plays an important role in a variety of diseases with potential therapeutic implications (Gilbert et al., 2018). The gastrointestinal (GI) tract represents a unique interface environment to study interactions between the host’s immune system and the microbiota (Fan and Pedersen, 2021; Rook et al., 2017). In particular, the surface expression of the T cell receptor (TCR) endows T cells with the capacity to mediate specific microbial recognition in the context of both mutualistic coevolution and pathology (Ivanov et al., 2022). The characteristics of a successful T cell response to a given microbe (whether pathogenic or commensal) are antigen specificity and appropriate functional differentiation into T helper (Th), T effector (Teff), or T regulatory (Treg) cells. While pro-inflammatory T cell responses are mainly directed at clearing pathogens, Treg cells survey the ecosystem to modulate immune response, maintaining peripheral tolerance while dampening tissue inflammation (Ivanov et al., 2022). The induction of specific peripheral Treg (pTreg) populations has been shown to be driven by microbial-derived metabolites, essential to establish a tolerogenic environment in the GI tract (Takeuchi et al., 2024).

The RAG1 and RAG2 genes encode for lymphoid-specific proteins that initiate the process of V(D)J recombination, allowing the assembly of variable (V), diversity (D), and joining (J) genes at the TCR and immunoglobulin loci (Fugmann et al., 2000). While functionally null (or quasi-null) variants in the RAG genes result in severe combined immunodeficiency with the absence of T and B lymphocytes (T⁻B⁻NK⁺ SCID) (Schwarz et al., 1996) or Omenn syndrome (OS) (Villa et al., 1998), hypomorphic variants allowing for residual protein expression and function are responsible for various clinical phenotypes, globally defined as partial RAG deficiency (pRD) (Csomos et al., 2022) and characterized by partial development of T and B cells with a restricted repertoire.

Inflammatory bowel disease (IBD)–like colitis has been reported in up to 15% of pRD patients, being the most frequent manifestation of immune dysregulation in this condition, after autoimmune cytopenias (Delmonte et al., 2020). Here, to better characterize the pathophysiology of IBD associated with pRD, we studied a cohort of sixteen pRD patients, evaluated their clinical and immunological features, analyzed inflammatory signatures both in peripheral blood and in the colonic tissue, and characterized fecal microbiota composition. Moreover, we applied a multiomics strategy (including lamina propria [LP] CD4⁺ T cell single-cell RNA-sequencing [scRNA-seq] and TCR repertoire analysis, and fecal microbiome and metabolome analysis), to study a mouse model of hypomorphic Rag1 deficiency (Rag1^R972W/R972W, from now on referred to as Rag1^w/w mice) that presents with spontaneous development of colitis. This mouse model recapitulates the immunological phenotype seen in pRD patients, with a low number of naïve T cells, and a variable number of activated/memory T cells with a restricted TCR repertoire (Ott de Bruin et al., 2018). We show that Rag1^w/w mice have an inflammatory Th1/Th17 signature already present at weaning, associated with abundance of T cells expressing the gut-homing α4β7 integrin in mesenteric lymph nodes (MLN). This inflammatory signature is accompanied by a reduction in microbiota diversity, with depletion of microbial species producing short-chain fatty acid (SCFA) both in pRD patients and in Rag1^w/w mice, and defective induction of pTreg cells in the LP of Rag1^w/w mice. We evaluated the capacity of various treatment modalities to correct intestinal and systemic inflammation. The use of antibiotics improved intestinal inflammation in Rag1^w/w mice, but did not correct the Th1/Th17 skewing. The use of vedolizumab (targeting α4β7 integrin) in Rag1^w/w mice and of ustekinumab (targeting IFN-γ and IL-17 production) in a pRD patient with IBD were ineffective in correcting intestinal inflammation. In contrast, both systemic and intestinal inflammatory signatures were corrected upon bone marrow transplantation (BMT), suggesting that this therapeutic approach might be beneficial in restoring complex interactions between the immune system and microbiota in pRD and potentially also in patients affected by other severe monogenic forms of immune dysregulation.

We studied a cohort of 16 pRD patients carrying biallelic RAG variants. All patients (11 males and 5 females, mean age 18.1 years, age range 16 mo–44 years) presented with a pRD phenotype showing a combination of severe and recurrent infections and clinical manifestations of immune dysregulation. Patients’ demographics, genotype, clinical phenotype, and treatments for immune dysregulation are reported in Table 1. One third of the patients (P1–P5, 4 males and 1 female, mean age 11.9 years, age range 16 mo–37 years) presented with IBD-like colitis with protracted diarrhea and weight loss.

The patients exhibited variable degrees of T and B cell lymphopenia, associated with a significant reduction in the proportion of naïve T cells, while normal serum immunoglobulin A (IgA) and IgM levels were present in most of the patients (Fig. 1 A). Increased levels of CXCL9, IFN-γ, CXCL10, IL-17A, IL-10, and IL-6 were detected in the plasma of pRD patients irrespective of the IBD status (Fig. 1 B), consistent with the systemic inflammatory signatures recently described by our group in patients carrying RAG mutations (Bosticardo et al., 2025). Moreover, no statistically significant difference was observed in the residual recombination activity of RAG variants among pRD patients with or without IBD (Fig. S1 A).

Hematoxylin–eosin (H&E) staining of colonic samples of pRD patients with IBD showed active chronic colitis with mucosal architectural distortion, and immunohistochemistry and in situ hybridization revealed prominent CD3⁺ T cell infiltrates and increased levels of IFNG and CXCL9 transcripts, respectively (Fig. 1 C and Fig. S1 B). Together, these data demonstrate that pRD is characterized by systemic and tissue-specific inflammatory signatures with up-regulation of IFN-γ signaling and skewing toward Th1/Th17 immune responses. Based on these findings, we aimed at further dissecting the pathogenesis of chronic colitis in Rag1^w/w mice.

We have previously demonstrated that Rag1^w/w mice have impaired generation of T and B cells in the thymus and bone marrow, resulting in peripheral lymphopenia, with markedly reduced proportion of naïve T cells and increased percentage of activated/memory T cells, restricted TCR repertoire, hypogammaglobulinemia, impaired specific antibody production, and signs of immune dysregulation, as indicated by elevated serum IgE and production of a broad range of autoantibodies (Ott de Bruin et al., 2018). These features recapitulate the immunological phenotype seen in pRD patients.

Analysis of intestinal pathology in Rag1^w/w mice revealed spontaneous colitis, macroscopically characterized by substantial thickening of the colon throughout its entire length. H&E staining of colonic samples from 10- to 12-wk-old Rag1^w/w mice showed colonic architectural disruption with epithelial hyperplasia, crypt elongation, and prominent inflammatory cell infiltrate consisting mainly of CD3⁺ T cells in the colon LP (Fig. 2 A). To quantify the histologic abnormalities, we used a colitis score that evaluates the degree of architectural disruption, inflammatory infiltrate, and muciparous gland activity, ranging from zero (normal) to 10, corresponding to the most severe alterations (Rigoni et al., 2016). Comparing the colitis score of male and female mice at 12 wk of age, we observed that colitis was more severe in female mice (Fig. S2 A). For this reason, further studies were focused on female Rag1^w/w mice. We observed that the colitis score of female Rag1^w/w mice increases over time as mice age, with the histologic changes becoming evident only after weaning (Fig. 2 B).

The increased expression of Ifng and Cxcl9 transcripts was detected by in situ hybridization in Rag1^w/w colonic specimens at 10–12 wk of age (Fig. 2 C), similar to what observed in pRD patients with IBD. In addition, real-time quantitative polymerase chain reaction (qPCR) analysis showed higher levels of Ifng and Il17 transcripts in the colonic tissue from Rag1^w/w mice (Fig. S2 B). Besides an increased frequency of activated, pro-inflammatory CD4⁺ T cells, also the frequency of myeloid cells (especially neutrophils and monocytes) was increased in the LP of Rag1^w/w mice (Fig. S2, C and D).

Since the colonic inflammation was associated with prominent T cell infiltrates, we aimed at studying the T cell phenotype in locally draining MLN. A higher proportion of activated Teff cells (identified as CD44^highCD62L^low CD3⁺ T cells) was observed in MLN from 12-wk-old Rag1^w/w mice (Fig. S2 E). In addition, a significantly higher percentage of MLN T cells expressed the integrin α4β7 (Fig. S2 F), which is responsible for T cell homing into gut-associated lymphoid tissues through its binding to the mucosal addressin cell adhesion molecule (Kurmaeva et al., 2014). Finally, to assess systemic inflammation, we measured serum levels of several cytokines, and demonstrated a progressive increase over time in IFN-γ, CXCL9, CXCL10, IL-2, TNF-α, IL-12p70, and IL-17A/F levels in Rag1^w/w mice (Fig. S2 G), mirroring what was observed in pRD patients.

Based on these findings, we focused on characterizing the colonic inflammation in Rag1^w/w mice. Flow cytometry analysis of LP immune cells at different ages revealed that the T lymphocytic infiltrate, determined as TCRβ⁺ cells, is quantitatively similar in Rag1^w/w and wild-type mice at 3 wk of age, but is significantly increased at 12 wk of age in Rag1^w/w mice (Fig. 2 D), paralleling the worsening of the colitis score. Flow cytometry analysis of LP Th lymphocytes (identified as TCRβ⁺, CD4⁺, Foxp3⁻ cells) demonstrated that Rag1^w/w mice have an increased percentage of CD4⁺ T cells expressing IFN-γ and/or IL-17 compared with wild-type controls (Fig. 2, E and F). Noteworthily, an increased proportion of IFN-γ⁺/IL-17⁺ double-positive T cells, frequently identified in the inflamed LP of patients and murine models of IBD (Annunziato et al., 2007), were observed in Rag1^w/w mice. This T cell phenotype was already present in Rag1^w/w mice at 3 wk of age (Fig. 2, E and F), even before development of a significant T cell infiltrate (Fig. 2 G), suggesting that the Th1/Th17 skewing of Rag1^w/w mouse CD4⁺ T cells is genetically determined.

To further study the T cell–mediated immune dysregulation, we performed scRNA-seq of sorted LP CD4⁺ T from Rag1^w/w mice and wild-type controls at 3 and 12 wk of age. Differential gene expression identified 11 transcriptionally distinct CD4⁺ T cell subsets (Fig. S3 A). Based on the differential gene expression and known markers from the literature, we were able to annotate these different subsets (Fig. S3 B). In particular, Tregs were the more abundant population and were identified by the expression of Foxp3; naïve CD4⁺ T cells expressed high levels of Ccr7 and Sell, while activated T cells were positive for Nkg7. We identified T follicular helper (Tfh) cells expressing Tcf7 and Cxcr5, and a small population of Th2 cells expressing the Il4 gene. Several populations expressing high levels of Il17, Ifng, and interferon-responsive genes were identified, similar to what was seen by flow cytometry. When we looked at the different sample contribution to these subsets, an enrichment in CD4⁺ T cells expressing the Ifng and Il17 transcripts was present in Rag1^w/w mice already at 3 wk of age, along with a marked reduction in the proportion of naïve CD4⁺ T cells (Fig. 3, A and B). Flow cytometry demonstrated that FOXP3⁺ T cells represented ∼20% of total LP CD4⁺ cells in both wild-type and Rag1^w/w mice; however, due to the increased abundance of total TCRβ⁺ cells, the absolute count of FOXP3⁺ cells was significantly increased in Rag1^w/w mice (Fig. S3 C).

To better define transcriptional signatures characteristic of T cell subtypes, we performed gene set enrichment analysis comparing the LP CD4⁺ T cell expression profiles in Rag1^w/w mice and wild-type controls at 12 wk of age. Pathways enriched in IFN-γ response and Tfh subsets of Rag1^w/w mice included “T cell activation,” “response to interferon-gamma,” “response to interferon-alpha,” “response to chemokine” and “regulation of leukocyte activation” (Fig. 3 C). Furthermore, pathway analysis demonstrated enrichment in “negative regulation of immune system process” and “negative regulation of cell activation” in the IFN-γ⁺ and IL-17⁺ clusters of LP CD4⁺ T cells in wild-type compared with Rag1^w/w mice (Fig. S3 D). These data are consistent with active LP inflammatory processes associated with chronic colitis in Rag1^w/w mice. Finally, inflammatory signatures of cell activation in Rag1^w/w mice were more prominent at 12 wk of age than at 3 wk of age (Fig. S3 E).

On the same samples, we also performed single-cell TCR sequencing with the goal of defining the TCR repertoire of LP CD4⁺ T cells. A severe restriction of the TCR-β (TRB) repertoire of sorted LP CD4⁺ T cells was observed in Rag1^w/w mice, with nearly 20% of hyperexpanded clonotypes (abundance >100 per clonotype, each of which is defined by a unique TCR-α [TRA]-TRB pairing) already present at 3 wk of age (Fig. 3 D). Hyperexpanded clonotypes were observed especially among CD4⁺ T cells expressing Ifng and among cells in the IFN-γ response cluster (Fig. 3 E). Analysis of the amino acid composition of the complementarity-determining region 3 (CDR3) of TRB clonotypes in LP CD4⁺ T cells of Rag1^w/w mice revealed increased abundance of hydrophobic amino acids (Fig. 3 F), known molecular biomarkers of self-reactive T cells (Stadinski et al., 2016).

Overall, these data indicate that the colitis observed in Rag1^w/w mice is characterized by a CD4⁺ T cell infiltrate in the LP with an immunological phenotype and transcriptional profile consistent with Th1/Th17 inflammatory signatures, T cell activation, and expansion of T cell clonotypes with characteristics of self-reactivity. This phenotype is already present early in life, whereas the colonic inflammation histologically manifests only after weaning, a time of profound remodeling (and reactivity) of the host–microbiota relationship (Lubin et al., 2023), suggesting a potential role of altered interactions between the host’s immune system and the microbiota in driving and sustaining colonic inflammation.

To characterize the composition of intestinal microbiota in pRD patients, we performed 16S rRNA gene sequencing analysis on fecal samples from 15 pRD patients (P1, P3-P16) and 23 healthy controls (HC). Firstly, pRD patients, independently from the presence of IBD, showed a significant reduction of fecal microbiota alpha diversity (Fig. 4 A), a signature commonly associated with several pathological conditions including IBD (Ott et al., 2004). This reduced gut microbiota diversity in IBD is especially manifested in a loss of gut bacteria capable of producing SCFA, microbial metabolites that dampen inflammation (Arpaia et al., 2013; Smith et al., 2013) and bolster gut barrier integrity (Wang et al., 2020). The loss of these taxa is thought to contribute to IBD progression, though the dynamics of these metabolites and the taxa that produce them in pRD remains poorly understood.

To understand whether reduced alpha diversity in pRD was associated with altered microbiota composition, we studied the differential abundance of microbial species in pRD patients and HC. A significant increase in the relative abundance of oral bacteria, including the Streptococcus genus, was detected in the gut microbiota of pRD patients (Fig. 4 B and Fig. S4 A). In this context, there is increasing evidence that microorganisms of the oral cavity, ectopic to the gut, may trigger intestinal inflammation (Atarashi et al., 2017; Kitamoto et al., 2020; Read et al., 2021). Further analysis of microbial differential abundance between pRD and HC identified a significant reduction in the relative abundance of the Ruminococcaceae family, which contains numerous SCFA producers (Fig. 4 C), and, at the genus level, a significant reduction of SCFA producers Butyricicoccus and Coprococcus (Fig. 4 D) in pRD patients.

Gut mucosal IgA is a host molecule that promotes host–microbe symbiosis by binding gut bacteria and tethering them to host mucus to prevent their penetration into host tissues (Pabst and Slack, 2020; Yang and Palm, 2020). This tethering can additionally enhance colonization of targeted bacteria (Donaldson et al., 2018). To assess whether reduced abundance of taxa in pRD patients was associated with diminished IgA against the same taxa, we performed IgA-sequencing (IgA-seq) profiling on pRD patients and HC.

Firstly, when examining the total antimicrobial mucosal IgA repertoire, pRD and HC tended to cluster separately (Fig. 4 E). Moreover, concomitant with diminished relative abundance of Ruminococcaceae and Coprococcus observed in pRD patients, we found that Coprococcus was less IgA-coated in pRD patients as compared to HC, as was the prominent SCFA-producing genus Faecalibacterium (Fig. 4 F).

To evaluate whether similar microbial signatures were observed also in the mouse model, we performed 16S rRNA gene sequencing on murine fecal samples collected from Rag1^w/w mice and wild-type controls. Firstly, we analyzed the fecal microbiota from co-housed Rag1^w/w and wild-type mice. Co-housing represents a standard approach in mouse facilities, and it is known to normalize the microbiota composition of mice kept in the same cage. Analysis of the fecal microbiota in co-housed 12-wk-old Rag1^w/w and wild-type mice revealed a significant reduction of microbiota alpha diversity in Rag1^w/w mice (Fig. 4 G). The observation that Rag1^w/w mice manifested a low alpha diversity in their microbiota, even when co-housed with wild-type mice, indicates a very strong effect of genotype on microbial composition. In order to assess whether worsening of colonic inflammation over time correlates with changes in microbiota composition, we performed a longitudinal time course analysis on fecal samples collected from Rag1^w/w mice and wild-type controls at different time points (3, 5, 10, and 14 wk of age) and housed separately. Significantly reduced alpha diversity of the microbiota was observed in Rag1^w/w mice, which became more evident particularly at 10 and 14 wk of age (Fig. 4 H). Beta diversity principal component analysis showed that Rag1^w/w and wild-type mice cluster separately, with completely independent clustering at 14 wk of age (Fig. S4 B).

Further analysis of microbial differential abundance between Rag1^w/w and wild-type mice identified a significant progressive reduction in the relative abundance of the Ruminococcaceae family in mutant mice (Fig. 4 I) as was seen in pRD patients, and, at the genus level, a significant reduction of Intestinimonas, Lachnospiraceae, and Oscillibacter that became more prominent over time (Fig. S4, C and D). Moreover, concordant with human data, IgA-seq profiles of Rag1^w/w and wild-type mice cluster separately (Fig. 4 J); in addition, the Ruminococcaceae family was significantly less IgA-coated in mutant mice as compared to wild-type mice (Fig. 4 K).

The Ruminococcaceae family belongs to the Firmicutes phylum and is known to be enriched for SCFA-producing bacteria (Deleu et al., 2021; Fusco et al., 2023). To investigate for possible differences in the representation of metabolic pathways between Rag1^w/w and wild-type mice, we utilized phylogenetic investigation of communities by reconstruction of unobserved states (PICRUSt) analysis. PICRUSt can predict microbial community function by inferring from marker gene sequencing using curated reference genome databases (Langille et al., 2013). Pathways related to SCFA production (“pyruvate fermentation to butanoate”; “acetyl-CoA fermentation to butanoate II”; “L lysine fermentation to acetate and butanoate”) were significantly depleted in Rag1^w/w mutant mice (Fig. 4 L). Subsequent fecal metabolomics assessment of Rag1^w/w mice showed broad metabolic abnormalities, with decreased fecal SCFA levels in Rag1^w/w mice (Fig. 4 M), in agreement with the altered microbiota landscape.

SCFAs are known to play an important role in the generation of pTreg cells in the colon and in supporting their suppressive function (Smith et al., 2013). Therefore, we investigated whether the depletion of SCFA was associated with altered Treg composition in the colon LP CD4⁺ cells of Rag1^w/w mice. While Foxp3 is the master regulator of Treg development (Fontenot et al., 2003), Helios and Rorc have been used to identify thymic/central Treg (tTreg) and pTreg subpopulations, respectively, in the GI tract (Sefik et al., 2015; Thornton et al., 2010; Xu et al., 2018). In addition, a substantial proportion of Helios⁺ GI Tregs also express Gata3 (Wohlfert et al., 2011) and the surface molecule Neuropilin-1 (Singh et al., 2015), whereas Rorc⁺ Tregs largely co-express c-Maf and IL-10 (Xu et al., 2018). Single-cell RNA-seq data of sorted LP CD4⁺ T cell showed no significant depletion on the Treg cluster in Rag1^w/w mice (Fig. 3, A and B). To better characterize the transcriptomic profile and subset distribution of LP Treg, we performed a subcluster analysis of the Treg cluster and identified six subpopulations (Fig. 5 A). Rorc and Il10 expression was detected in clusters 2 and 3, identifying these cells as effector pTreg. In contrast, clusters 0, 1, 4, and 5 were characterized by high levels of tTreg markers such as Gata3 and Ikzf2 (the latter encoding for Helios) (Fig. 5 B). When we investigated whether these clusters were equally represented in wild-type and Rag1^w/w mice, we observed a marked depletion in microbiota-induced Rorc⁺/Maf⁺Il10⁺ pTreg, and enrichment of Ikzf2⁺Gata3⁺ tTreg in Rag1^w/w mice at 12 wk of age (Fig. 5 C). Moreover, flow cytometry analysis confirmed the reduction of RORγt⁺ Treg cells in the LP of Rag1^w/w mice at 12 wk of age (Fig. 5 D).

Together, these data indicate that the reduction of SCFA-producing microbial species that becomes evident after weaning in Rag1^w/w mice is associated with a reduction in SCFA-dependent Rorc⁺/Maf⁺Il10⁺ effector pTreg cells, possibly altering mechanisms of immune homeostasis and promoting colonic inflammation. In line with recent data in wild-type mice (Gu et al., 2024), we found little co-expression of Rorc and Gata3 also in Rag1^w/w mice (1.99% and 1.31%, respectively) (Fig. 5 E). Finally, an increased representation of hyperexpanded clonotypes was detected in the Ikzf2⁺Gata3⁺ tTreg compartment of Rag1^w/w mice at 12 wk of age (Fig. 5 F).

To assess the relative role of host and microbiota abnormalities in the pathogenesis of chronic colitis, we evaluated different treatments targeting either the microbiota or the immune cells in Rag1^w/w mice. In particular, considering the higher percentage of MLN T cells expressing the gut-homing integrin α4β7 in Rag1^w/w mice, we tested the effects of vedolizumab, a monoclonal antibody targeting the α4β7 integrin and currently approved in humans for the treatment of Crohn’s disease and ulcerative colitis (Wyant et al., 2016). 10-wk-old Rag1^w/w mice received 200 μg of vedolizumab or of isotype control every 2 days for 2 wk and then every 5 days until sacrifice after 45 days from treatment initiation (Fig. 6 A), following a protocol previously published (Lindebo Holm et al., 2012). Moreover, we tested the role of antibiotics in depleting gut microbial species and possibly dampening the microbiota-induced inflammation. To this purpose, we chose the broad-spectrum antibiotic cocktail with ampicillin, neomycin, metronidazole, and vancomycin (ANMV), as it has been widely used in murine studies to ablate microbiota (Sim et al., 2022). ANMV were administered with three different modalities (Fig. 6 B): (1) for 4 wk (8–12 wk of age), (2) starting at 3 wk of age, or (3) to pregnant mothers, to expose the mice to antibiotics since fetal life, and continuing ANMV until 12 wk of age in Rag1^w/w offspring. Finally, another group of Rag1^w/w mice received BMT at 7 wk of age after lethal total body irradiation (TBI), and were analyzed 20 wk after transplantation (Fig. 6 C). Treatment conditions were compared with wild-type mice and untreated Rag1^w/w mice.

While vedolizumab proved ineffective in ameliorating colitis, broad-spectrum ANMV antibiotics (independently from the schedule of administration) and BMT significantly improved the colonic inflammation, with a marked reduction in the colitis score at histologic evaluation (Fig. 6 D), normalization of the LP T cell numbers by flow cytometry analysis (Fig. 6 E), and significant reduction of the Cxcl9 and Ifng inflammatory signature (Fig. 6 F). However, both vedolizumab and ANMV-treated mice retained a high percentage of LP CD4⁺ T cells expressing IFN-γ and/or IL-17 (Fig. 6 G and Fig. S5 A). Moreover, scRNA-seq on sorted CD4⁺ T cells infiltrating the colon LP in ANMV-treated Rag1^w/w mice showed persistence of cells expressing the Ifng and Il17 transcripts (Fig. 6 H). The use of broad-spectrum antibiotics was associated with a redistribution of Treg subsets in ANMV-treated compared with untreated Rag1^w/w mice (compare Fig. S5 B and Fig. 5 C). However, while the pattern of Il10-expressing Tregs mirrored what observed in wild-type mice, defective Il10 expression was not rescued by ANMV treatment (Fig. S5 C).

Among Ikzf2⁺Rorc⁻Il10⁻ tTreg cells, analysis of TCR repertoire demonstrated a significant overlap of TRA clonotypes between ANMV-treated and untreated Rag1^w/w mice at 12 wk (Fig. S5 D). We also observed that the LP tTreg repertoire of Rag1^w/w mice was already established at 3 wk (as shown by clonotype overlap of Ikzf2⁺Rorc⁻Il10⁻ cells between 3-wk- and 12-wk-old mice). A similar effect (although less pronounced) was seen also in wild-type mice (Fig. S5 D). These data suggest that the shaping and the restriction of the TCR repertoire of tTreg cells in Rag1^w/w mice are not driven by the microbiota, but are genetically established since early in life. Notably, no overlap was seen in the TRA repertoire of wild-type and Rag1^w/w mice within either the tTreg or the pTreg compartments.

Among the various treatment modalities attempted, BMT was the only one that allowed resolution of the Th1/Th17 colonic inflammatory signature in Rag1^w/w mice (Fig. 6 G). In addition, BMT was associated with a significant reduction in systemic inflammation, as shown by the normalization of IFN-γ, CXCL10, CXCL9, TNF-α, IL-2, and IL-10 plasma levels (Fig. 6, I and J; and Fig. S5 E). In contrast, ANMV treatment was only partially effective in dampening systemic inflammation, and minimal changes of plasma levels of inflammatory biomarkers were observed upon treatment with vedolizumab (Fig. 6, I and J; and Fig. S5 E).

BMT had also prominent effects on microbiota diversity and composition, as demonstrated by normalization of microbiota alpha diversity (Fig. 6 K) and increased relative abundance of Intestinimonas, Lachnospiraceae, and Oscillibacter genera (Fig. 6 L). Finally, analysis of fecal metabolome in BMT-treated Rag1^w/w mouse composition showed an increase in SCFA levels compared with untreated Rag1^w/w mice (Fig. 6 M).

We also investigated the effects of BMT and of ustekinumab (targeting IFN-γ and IL-17 production) in pRD patients. Among nine pRD patients who are alive after BMT, plasma levels of inflammatory biomarkers before and after definitive treatment were measured in seven of them, including two with IBD (P2 and P5 in Table 1). Plasma levels of IFN-γ, CXCL10, and IL-17A normalized after BMT (Fig. 7 A). Comparison of gut microbiota diversity and composition in two pRD patients (P8, P12) in which fecal samples were collected before and after BMT showed a trend toward normalization of alpha diversity (Fig. 7 B), and relative abundance of oral bacteria (Fig. 7 C) and of Firmicutes (Fig. 7 D). Four of the five patients with IBD (P1, P2, P4, and P5) received BMT. P1 and P4 died of respiratory failure and sepsis, and of disseminated cytomegalovirus infection, respectively, a few months after BMT. The two patients who survived (P2 and P5) had full resolution of GI symptoms after BMT; fecal calprotectin levels before and after BMT were measured in P2, documenting resolution of GI tract inflammation (1,549 μg/g before BMT versus 9 μg/g after BMT; normal value: <50 μg/g). In contrast, treatment with ustekinumab was ineffective to treat IBD in patient P3, in spite of an increase in the frequency of administration of this drug after 6 mo of treatment. After 9 mo of treatment with ustekinumab, P3 continued to have diarrhea and abdominal pain, and fecal calprotectin levels remained elevated (223 μg/g compared with 114 μg/g before treatment). A repeat colonic biopsy performed after 2 years of treatment with ustekinumab showed worsening of T cell infiltrates and persistence of inflammation (Fig. 7 E), prompting referral for BMT.

In this study, we sought to identify disease-specific immunopathological and microbial changes driving severe and drug-resistant chronic colitis in pRD patients.

The spectrum of RAG deficiency phenotypes correlates with recombination activity of the mutant RAG alleles, with minimal levels in patients with SCID and OS but residual functional activity in pRD patients (Bosticardo et al., 2025). By applying a multiomics approach to a large group of RAG-mutated patients, we have recently shown that pRD is associated with a type 1 immune profile both in blood and in tissues (Bosticardo et al., 2025). Taking advantage of our Rag1^w/w mouse model, which recapitulates the immunological phenotype seen in pRD patients (Ott de Bruin et al., 2018), we were able to dissect the gut immunopathological signatures, and demonstrated an LP CD4⁺ T cell infiltrate with an immunological phenotype and transcriptional profile consistent with Th1/Th17 inflammatory signatures, T cell activation, and oligoclonal T cell expansions with molecular signatures of self-reactivity. Of note, IFN-γ– and IL-17–mediated inflammation has been extensively described in IBD pathogenesis, with synergistic action of Th1 and pathogenic Th17 cells in promoting disease development and facilitating aggravation of the mucosal inflammation (Cao et al., 2023). A similar systemic inflammatory signature has been previously demonstrated in patients with OS (a condition characterized by the oligoclonal expansion of activated T cells) (Rigoni et al., 2020) and in the GI tract of a mouse model of this disease (Rigoni et al., 2016). Our study extends this phenotype to pRD and suggests a role for dysregulated T cells in driving organ-specific inflammation. Interestingly, although systemic and tissue-specific inflammatory signatures are already present early in life in Rag1^w/w mice, their colonic inflammation manifests histologically only after weaning. It has been demonstrated that at weaning, the gut microbiota induces a vigorous immune response, the so-called ‘‘weaning reaction,’’ that is fundamental for the physiological immune maturation (Al Nabhani et al., 2019). In particular, this weaning reaction is associated with induction of pTregs and development of peripheral tolerance, and its perturbation has been linked to increased susceptibility to immune-mediated diseases later in life (Al Nabhani et al., 2019). We hypothesize that the genetically determined restriction of TCR repertoire in mice and humans with hypomorphic RAG variants may cause an aberrant weaning reaction, ultimately leading to altered composition of gut microbiota. Indeed, Rag1^w/w mice showed a significantly reduced alpha diversity of gut microbiota composition even when co-housed with wild-type and heterozygous littermates, thus overriding the partial normalization of gut microbiota composition and diversity that is typically observed in co-housing (Robertson et al., 2019). Analysis of the fecal microbiota of pRD patients revealed a relative increase in oral pathogenic bacteria that can drive Th1 induction and severe gut inflammation in the context of a genetically susceptible host (Atarashi et al., 2017). The reduction of SCFA-producing bacteria with concomitant lower levels of SCFA in stool specimens represents the major link between the microbial and immunopathological signatures (Mann et al., 2024). In particular, butyrate, propionate, and acetate, three fermentation by-products of fiber digestion, signal through G protein–coupled receptors (GPCRs); activation of GPR43 by SCFAs promotes pTreg generation and suppressive function (Smith et al., 2013). Furthermore, butyrate-mediated activation of the GPCR GPR109A expressed by CD103⁺ dendritic cells induces the development of GI tract Rorc⁺ Treg cells, promoting oral tolerance and protecting from food allergy (Tan et al., 2016). Finally, butyrate and propionate inhibit histone deacetylase and at low concentration can enhance histone acetylation (Lund et al., 2022), thereby inducing epigenetic changes at the Foxp3 locus that facilitate intestinal Treg cell differentiation (Arpaia et al., 2013; Furusawa et al., 2013). It has been recently demonstrated that the GI LP is the key microniche that supports generation and function of Il10⁺Rorc⁺ effector Treg cells (Gu et al., 2024). In particular, commensal antigens induce the generation of colonic Rorc⁺ Tregs, which have been shown to dampen IFN-γ– and IL-17–mediated pro-inflammatory responses of Teff cells (Sefik et al., 2015; Xu et al., 2018) and to suppress Th2-mediated responses to dietary antigens (Abdel-Gadir et al., 2019) and parasites (Ohnmacht et al., 2015). Accordingly, the loss of Rorc⁺/Maf⁺ Tregs in mice is associated with dysbiosis, lethal oxazolone-induced colitis, increased levels of IgE, and severe allergy (Abdel-Gadir et al., 2019; Ohnmacht et al., 2015). IL-10 has a fundamental role as tolerogenic cytokine, and its depletion has been associated with colitis development (Murai et al., 2009). It has been recently shown that a subpopulation of effector Tregs stably expressing IL-10 is indispensable to maintain colonic health (Dikiy et al., 2025). Our data show that the decrease in SCFA-producing microbial species that becomes manifest after weaning in Rag1^w/w mice is associated with a reduction in SCFA-dependent Rorc⁺/Maf⁺Il10⁺ peripheral effector Treg cells, thereby altering mechanisms of immune homeostasis and promoting colonic inflammation. In this pro-inflammatory environment, the exposure to colonic microbes may favor the development of inflammation with an increased number of Th1/Th17 cells infiltrating the LP, as observed in Rag1^w/w mice. Additionally, analysis of the TCR repertoire has shown that LP CD4⁺ T cells from Rag1^w/w mice are enriched in clonotypes containing hydrophobic amino acids at the center of the TRB CDR3 region, a biomarker of self-reactive T cells. Purging of T cells expressing TCR chains with increased frequency of hydrophobic amino acids in their CDR3 occurs in the thymus as a result of high-strength interaction between the TCR and self-peptide/MHC (Stadinski et al., 2016). We have previously shown that peripheral blood conventional CD4⁺ T cells from RAG-mutated patients and Rag1^w/w mice contain a high frequency of clonotypes with hydrophobic amino acids in the TRB CDR3 (Daley et al., 2019; Rowe et al., 2017). These data suggest that pRD is characterized by failure of negative selection and increased thymic output of oligoclonal and self-reactive T cells, which may trigger peripheral tissue damage. The demonstration that oligoclonal expansions of Th1/Th17 cells can be detected in the LP of Rag1^w/w mice already at weaning strongly indicates that this pro-inflammatory signature is microbiota-independent and rather genetically driven.

Several lines of evidence indicate that alterations of TCR repertoire (with restricted diversity and oligoclonal expansion of inflammatory and dysreactive clonotypes) play a critical role in the development of IBD. First, contrary to what observed in Rag1^w/w mice, no spontaneous development of inflammatory gut disease is observed in Rag^−/− mice (Aranda et al., 1997), which can be induced upon transfer of naïve CD4⁺CD45RB^high T cells from wild-type mice into Rag1^−/− or Rag2^−/− mice (Reinoso Webb et al., 2018). Moreover, adoptive transfer of total CD4⁺ cells from MLN of Rag2^R229Q/R229Q mice (mimicking OS) into Rag1^−/− recipients reproduced the IBD observed in Rag2^R229Q/R229Q mice, whereas adoptive transfer of wild-type Treg cells into Rag2^R229Q/R229Q mice improved bowel inflammation (Rigoni et al., 2016). To further confirm the importance of T cells in maintaining homeostatic interactions with the microbiota, intestinal inflammation was a prominent feature of a mouse model of partial Lck deficiency affecting T, but not B, cell development and function (Lui et al., 2024). Of note, consistent with our findings, Lui et al. demonstrated that complete versus partial loss-of-function LCK in mouse models causes disease with differing phenotypes. While both mouse models showed an arrest in thymic T cell development and profound T cell lymphopenia, only mice with partial loss of function exhibited residual T cell proliferation and intestinal inflammation. Furthermore, the intestinal disease in mice with partial loss of function was prevented by CD4⁺ T cell depletion or by transfer of regulatory T cells. These findings suggest that partial loss-of-function LCK spares sufficient T cell function to allow the maturation of some conventional T cells but not regulatory T cells—leading to intestinal inflammation. Finally, in humans, IBD is not observed in patients with RAG-deficient SCID, but is frequently detected in patients with OS due to oligoclonal expansion of activated autologous T cells (Rigoni et al., 2020).

Defects of humoral immunity may also contribute to the immune dysregulation of pRD. We have confirmed that patients with pRD manifest B cell lymphopenia. While several of them had normal or low-normal IgM and IgA levels, previous studies from our group and others have shown that restriction of the BCR repertoire is another cardinal feature of pRD (Csomos et al., 2022; Lee et al., 2016). Similar features have been observed in Rag1^w/w mice (Ott de Bruin et al., 2018). Furthermore, the B cell compartment of both pRD patients and Rag1^w/w mice is enriched in dysreactive B cells (Bosticardo et al., 2025; Csomos et al., 2022; Ott de Bruin et al., 2018), which may further contribute to autoimmune manifestations. To investigate the possible contribution of impaired humoral immune responses to intestinal inflammation, we have performed IgA-seq profiling of fecal microbiota in pRD patient and Rag1^w/w mice. Production of IgA against small intestine–resident bacteria has been shown to promote colonization with bacteria promoting homeostasis and eubiosis. IgA-seq profiling demonstrated the reduced presence of IgA-coated Ruminococcaceae both in pRD patients and in Rag1^w/w mice compared with controls. This may affect the capacity of Ruminococcaceae to tether to the mucus and hence to colonize the gut, accounting for the reduced relative abundance of this family in the fecal microbiome.

In this study, we explored multiple approaches to prevent or treat IBD associated with pRD. Despite the efficacy of vedolizumab in treating ulcerative colitis and Crohn’s disease (Wyant et al., 2016), the use of this drug was ineffective in Rag1^w/w mice. It has been previously shown that vedolizumab may be beneficial to prevent IBD in a SCID T cell transfer model of colitis if administered starting at the same day of T cell transfer, but did not improve intestinal inflammation if administered 3 wk after T cell transfer (Lindebo Holm et al., 2012). Negative results were also observed upon administration of ustekinumab, targeting both IFN-γ and IL-17 production, in one patient with pRD. In this case, despite an increase in the frequency of administration of the drug, worsening of intestinal inflammation was observed. Although limited to a single case, this observation suggests that targeting of IFN-γ and IL-17 production is insufficient to control inflammation. We speculate that interfering with cytokine production or with T cell trafficking without correcting T cell function and diversity of the TCR repertoire may be detrimental, and may in fact aggravate the inability to adequately control the microbial burden.

Interestingly, the use of antibiotics was associated with improvement of inflammatory signatures in the GI tract of Rag1^w/w mice. We speculate that by depleting pathogenic bacteria, the use of antibiotics may have beneficial effects on dysbiosis, and hence also on the induction of pathogenic T cell responses. However, antibiotics do not correct the genetically determined skewing of Th profile nor the restriction of TCR repertoire. While they may help dampen inflammation in patients with pRD, they are insufficient to correct the disease.

BMT is the treatment of choice to normalize adaptive immune responses in pRD (Schuetz et al., 2023). Our study demonstrates that it is also successful in correcting the genetically determined immune dysregulation in the GI tract and dysbiosis associated with hypomorphic forms of the disease. The use of BMT in Rag1^w/w mice led to resolution of T cell infiltrate in the colon, normalization of the microbiome composition, and restoration of SCFA levels. Our preliminary results confirm the beneficial effects of BMT also in pRD patients, allowing complete resolution of clinical and laboratory features of IBD. The outcome of BMT in pRD is superior when performed early in the course of the disease, and especially in patients with positive newborn screening, whereas organ damage is predictive of worse outcome when transplantation is performed in adulthood (Schuetz et al., 2023).

Despite the caveats that should be acknowledged when comparing human and murine gut microbiota (Nguyen et al., 2015), our data demonstrate that pRD patients and Rag1^w/w mice share common microbial signatures, indicating that Rag1^w/w mice may represent a valuable model to study the pathophysiology of the chronic colitis identified in pRD patients.

Overall, by applying and integrating an array of multiple approaches to a cohort of pRD patients and a corresponding mouse model, we dissected the immunopathological, microbial, and metabolomic signatures associated with IBD in pRD. Our findings shed new light in the pathophysiology of intestinal disorders in patients with hypomorphic RAG mutations and establish a curative role for BMT in fully resolving the disease phenotype, which could possibly extend to other genetically determined immune dysregulation disorders characterized by a restricted TCR repertoire.

The study included 16 patients with biallelic RAG variants causing pRD, manifesting as combined immune deficiency (13/16, 80%) or leaky SCID (3/16, 20%). Characteristics of the patients’ clinical and immunological phenotype are reported in Table 1. Moreover, serum from 32 HC (17 adult and 15 pediatric HC) was analyzed for the measurement of soluble biomarkers and stool specimens from 23 HC (age-matched and corrected for the specific household) were studied for the fecal microbiota characterization. Blood and stool specimens from patients and controls and colonic tissue samples from patients were obtained upon informed consent, according to protocols 18-I-0041 (NCT03394053 in https://clinicaltrials.gov) and 18-I-0128 (NCT03610802), approved by the National Institutes of Health Institutional Review Board, and to protocol 00035468, approved by the University of South Florida Institutional Review Board.

Rag1^w/w mice were generated by gene editing as previously described (Ott de Bruin et al., 2016). Animal work was conducted in accordance with the U.S. Public Health Service Policy on Humane Care and Use of Laboratory Animals, with protocols approved by the National Institute of Allergy and Infectious Diseases Animal Care and Use Committee (protocol LCIM 6E). Mice were group-housed (four to five mice of the same sex per cage) in a controlled environment with unrestricted access to water and a standard chow diet. Moreover, for specific sets of experiments (as detailed in the specific figure legends), we isolated Rag1^w/w mice from wild-type mice immediately after weaning, avoiding possible microbial contamination among the two groups due to cage-sharing and murine coprophagy.

Regarding the treatment strategies, ANMV antibiotic cocktail, vedolizumab, and BMT were tested. Mice received the ANMV antibiotic cocktail consisting of ampicillin 1 g/liter, neomycin 1 g/liter, metronidazole 1 g/liter, and vancomycin 0.5 g/liter in drinking water ad libitum. The antibiotic solution was changed three times a week. ANMV was given with three different approaches: (1) for 4 wk (8–12 wk of age), (2) starting at 3 wk of age until sacrifice at 12 wk of age, or (3) at pregnant mothers, to expose the mice to antibiotics from fetal life to sacrifice at 12 wk of age. Water intake was monitored throughout the antibiotic administration. For the mice receiving ANMV for 4 wk, following completion of the antibiotic administration phase, regular drinking water was provided ad libitum for the remainder of the experiment.

10-wk-old Rag1^w/w mice received 200 μg of vedolizumab (InVivoMAb anti-mouse LPAM-1) or of isotype control (InVivoMAb rat IgG2a isotype control) every 2 days for 2 wk and then every 5 days until sacrifice after 45 days from treatment initiation (Lindebo Holm et al., 2012; Rosser et al., 2014; Sheridan et al., 2014).

For BMT experiments, 7-wk-old Rag1^w/w mice received TBI (8 Gy) 1 day before transplantation. On the day of transplantation, total BM cells were obtained from 6- to 10-wk-old donor CD45.1 C57BL/6 wild-type mice and administered intravenously to Rag1^w/w mice at a dose of 2 × 10⁷ cells. Of note, antibiotic TMS water was initiated after TBI and interrupted after 8 wk. BMT-transplanted mice were sacrificed after 20 wk from BMT (27 wk of age).

Mouse colonic tissue samples were formalin-fixed and paraffin-embedded. Sections (1.5 µm) were used for routine H&E staining to check for basic histopathological changes. Sections were dewaxed and rehydrated, endogenous peroxidase activity was blocked by 0.1% H₂O₂ for 15 min, and nonspecific background was reduced with Rodent Block (Biocare Medical) for 30 min before microwave antigen-retrieval treatment (EDTA buffer, pH 8.0). Moreover, sections were incubated for 1 h at room temperature with the primary rabbit polyclonal antibody anti-CD3 (1:100; Thermo Fisher Scientific) and then incubated for 30 min with MACH 1 Universal HRP Polymer Kit (Biocare Medical). In the end, reactions were developed in Biocare’s Betazoid DAB and nuclei counterstained with hematoxylin.

Digital images were acquired by an Olympus XC50 camera mounted on a BX51 microscope (Olympus), with CellF Imaging software (Soft Imaging System GmbH). A previously reported gut histologic score (colitis score) (Rigoni et al., 2016) was adopted by a pathologist (E. Fontana) for evaluating blindly different severity degrees of inflammation: grade of inflammation (scoring from 0 to 4, where 0 is normal condition and 4 is a severe grade of inflammation), the structural changes of the glands (scoring from 0 to 3, where 0 is normal condition and 3 is severe structural changes), and the goblet cell alterations (scoring from 0 to 3, where 0 is normal condition and 3 is severe alterations).

All slides were baked for 60 min prior to staining. The manual RNAscope 2.5 High Definition—BROWN assay kit (Cat No. 322300; Advanced Cell Diagnostics, Inc. [ACD]) was used according to the manufacturer’s instructions to perform in situ hybridization for Hs-CXCL9 Probe (Cat No. 440161), Hs-IFNG Probe (Cat No. 310501), Mm-Cxcl9 Probe (Cat No. 489341), and Mm-Ifng Probe (Cat No. 311391) on formalin-fixed paraffin-embedded (FFPE) tissue sections (Wang et al., 2012).

Slides were deparaffinized in xylene (2 × 5 min), dehydrated in 100% EtOH (2 × 1 min), air-dried at room temperature (RT), and incubated in RNAscope Hydrogen Peroxide (Cat No. 322330; ACD) at RT for 10 min to quench endogenous peroxidases. Slides were washed in deionized (DI) water twice before being submerged in 700 ml of fresh boiling RNAscope 1X Target Retrieval Reagents solution (Cat No. 322000; ACD) for 15 min, and then washed twice in DI water, once in 100% EtOH, and air-dried at RT. A hydrophobic barrier was drawn around the tissue, and RNAscope Protease Plus (Cat No. 322330; ACD) was applied for 20 min in a HybEZ Oven at 40°C (ACD). Slides were then washed twice in DI water and incubated with the appropriate probe for 2 h in the HybEZ Oven. RNAscope signal amplification reagents (Cat No. 322310; ACD) AMP 1 (30 min), AMP 2 (15 min), AMP 3 (30 min), AMP 4 (15 min), AMP 5 (75 min), and AMP 6 (15 min) were applied and incubated in HybEZ Oven. Before adding each AMP reagent, the slides were washed twice with RNAscope 1X Wash Buffer (Cat No. 310091; ACD). After removing AMP 6, slides were washed twice in wash buffer and RNAscope DAB detection reagents (Cat No. 322310; ACD) were applied and incubated for 10 min in HybEZ Oven. Sections were washed in tap water, counterstained with Harris–Mayer’s hematoxylin, washed again in tap water, placed in 0.02% ammonia water for 10 s, and washed a final time with tap water. Sections were then dehydrated in graded alcohols, treated with xylene (2 × 5 min), and coverslipped.

All slides were baked prior to staining. CD3 immunohistochemical staining was performed on an automated immunostainer BenchMark Ultra (Roche). After baking, Ultra CC1 (Cat No. 950-224; Roche) was applied to the slide for a 64-min antigen retrieval. Predilute CD3 (Clone 2GV6, Cat No. 790-4341; Ventana) was incubated for 32 min and detected with ultraView Universal DAB Detection Kit (Cat No. 760-500; Roche). Hematoxylin was used to counterstain. The slides were dehydrated in graded alcohols, treated with xylene (2 × 5 min), and coverslipped. Images were taken with an Olympus Bx41 microscope, objective UPLanFI 10×/0.30, 20×/0.50 ∞/0.17, 40×/0.75∞/0.17, and 100×/1.30 oil with an adaptor U-TV0.5xC using an Olympus DP27 camera with an Olympus U-VTO.63xC adaptor, using CellSens, XV Image Processing, imported into Adobe Photoshop CC 2024.

Cytokine and chemokine biomarker analysis was performed on sera obtained from wild-type and Rag1^w/w mice. Aliquots were stored in a −85°C freezer prior to analysis. Cytokines (IFN-γ, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-9, IL-10, IL-12p70, IL-15, IL-17A/F, IL-27p28/IL-30, IL-33, IP-10, KC/GRO, MCP-1, MIP-1α, MIP-2, TNF-α) were measured using V-PLEX Mouse Cytokine 19-Plex Kit (Meso Scale Discovery) and analyzed on a MESO QuickPlex SQ 120 reader (Meso Scale Discovery) according to the manufacturer’s specifications. Standard curves were analyzed using nonlinear curve fitting, and unknowns were calculated based on the derived equation. Duplicate determinations yielded coefficients of variation that were normally <10%. Defined low, medium, and high control samples were run on each plate to validate the responses and to assess the interplate variation. Cytokine values from the control samples had <5% plate-to-plate variation and averaged 100.7% (range 83.1–147.4) of their expected values.

Analysis of soluble biomarker plasma levels was performed in RAG patients and in healthy donors using V-PLEX Human Cytokine 30-Plex Kit (Meso Scale Discovery) and a MESO QuickPlex SQ 120 reader (Meso Scale Discovery) or using a customized, magnetic bead-based, multiplex assay (R&D Systems), depending on the nature of the analyte, as previously described (Sacco et al., 2022).

The mouse and human CXCL9 DuoSet ELISA kits (R&D) were used to determine the concentration of CXCL9 in sera from healthy donors and RAG patients and wild-type and Rag1^w/w mice according to the manufacturer’s instructions.

Mice were euthanized with CO₂, and the colon was collected and placed into cold complete medium (RPMI 1640 supplemented with 2 mM L-glutamine, 1 mM sodium pyruvate, 1 mM nonessential amino acids, 20 mM HEPES, 50 mM β-mercaptoethanol, 100 U/ml penicillin, and 100 mg/ml streptomycin). For colonic LP preparation, the mesenteric adipose tissue was removed. Tissues were opened, washed in cold PBS to remove feces, cut into 1- to 2-cm segments, and treated with complete medium containing 5 mM EDTA and 0.145 mg/ml dithiothreitol for 20 min at 37°C with constant stirring. Tissues were shaken vigorously for 1 min and further digested with 10 ml of medium containing 500 mg/ml DNase I (Sigma-Aldrich) and 100 mg/ml Liberase TL (Roche) with continuous stirring at 37°C. Digested tissues were passed through 70-μm cell strainers. Leukocytes were enriched by resuspension in 4 ml of 37.5% Percoll and centrifuged at 400 g for 5 min. Cells were then washed with PBS before downstream analysis.

To assess cytokine production potential, single-cell suspensions were restimulated in complete medium containing 10% fetal bovine serum, 50 ng/ml phorbol myristate acetate (Sigma-Aldrich), 5 mg/ml ionomycin (Sigma-Aldrich), and a 1:1,000 dilution of GolgiPlug (BD Biosciences) for 2.5 h at 37°C.

The following fluorochrome-labeled antibodies were used for flow cytometry studies: PE-CF594 anti-B220, BUV737 anti-CD11b, PE-CF594 anti-NK1.1, BV421 anti-SiglecF, PE-CF594 anti-TCRβ, BUV737 anti-TCRβ, PE-CF594 anti-TCRγδ (all from BD Horizon); PerCP-Cy5.5 anti-CD103, PerCP-Cy5.5 anti-CD11c, APC-eF780 anti-CD45, PE-Cy7 anti-CD45, FITC anti-FoxP3, eF660 anti-GATA3, AF700 anti-MHC-II, PE anti-RORγT, PerCP-Cy5.5 anti-TNF-α (all from Invitrogen); BV510 anti-CD4, Pacific Blue anti-CD4, BV510 anti-CD45, APC-Cy7 anti-CD45.1, BV510 anti-CD45.2, APC anti-CD45.2, PE anti-CD62L, APC anti-CD64, PerCP-Cy5.5 anti-CD8α, AF700 anti-CD90.2, PerCP-Cy5.5 anti-GATA3, BV605 anti-IFN-γ, PE-Cy7 anti-IL-17A, APC anti-LPAM-1, BV605 anti-Ly6C, BV421 anti-T-BET (all from BioLegend); AF700 anti-CD3ε, APC-Cy7 anti-CD44, PE-Cy7 anti-Ly6G (all from BD Pharmingen); FITC anti-CD24 (eBioscience); PE anti-CCR2 (R&D); and BV650 anti-CD8β (BD OptiBuild). For intracellular cytokine and transcription factor staining, after surface staining, cells were fixed and permeabilized using the Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific) for at least 1 h at 4°C and stained with fluorophore-conjugated antibodies for at least 1 h at 4°C. All staining was performed in the presence of purified anti-mouse CD16/32 and purified rat gamma globulin.

Total RNA was extracted from 1 cm of colonic tissue. The colonic tissue was homogenized in PureZOL reagent (Bio-Rad Laboratories) using TissueLyser II (Qiagen). RNA was extracted using the RNeasy Lipid Tissue kit (Qiagen), and cDNA was synthesized using the High-Capacity cDNA Reverse Transcription kit (Thermo Fisher Scientific). Quantitative RT-PCR was performed using an RT-PCR system (ViiA 7; Thermo Fisher Scientific) with TaqMan probes. The relative amounts of mRNAs were calculated as 2^−ΔCT and expressed as arbitrary units. Each sample was analyzed in duplicate, and the relative level of expression was determined by normalization to β-actin (Actb) ribosomal RNA.

LP CD4⁺ T cells (2.5 × 10⁴) were isolated from each sample using a FACSAria II (BD) cell sorter. All samples were labeled with TotalSeqC hashtag antibodies (BioLegend), pooled, and encapsulated into droplets using Chromium Single Cell Controller (10X Genomics), and libraries were prepared using Chromium Single Cell 5′ Reagent Kits v2 (Dual Index) with Mouse TCR Amplification Kit and 5′ Feature Barcode Kit (10X Genomics). The Controller was loaded with 2.5 × 10⁴ cells per lane. mRNA, TCR, and hashtag oligonucleotide (HTO) libraries were prepared following the 10X Genomics user guide. Libraries were sequenced on NextSeq 550, with 10% of the lane occupied by the HTO library, 10% by the TCR library, and 80% by the mRNA library.

Six 10X Chromium mRNA libraries were sequenced in two runs with an average yield of ∼128 million and a total yield of ∼766 million total reads, whereas the average and total number of read pairs sequenced in the six TCR (VDJ) libraries were ∼19 million and ∼113 million, respectively. The overall sequencing quality was high; at least 96% of bases (for both mRNA and VDJ reads) in the barcode regions had Q30 (99.99% inferred base call accuracy) or above, whereas at least 90% of bases in the mRNA and VDJ reads had Q30 or above. Similarly, >95% of the bases in unique molecular identifier (UMI) had Q30 or above for both mRNA and VDJ reads.

Furthermore, the median gene count per cell range was 620–1,978 (17,283–19,595 genes detected per library) along with 53–64% of the reads mapped confidently to the transcriptome, whereas the mean read count per cell was ∼15,000 with an average value of 92% of the reads detected in a total number of 77,422 cells across the six mRNA libraries. The percentage of cells with productive V-J spanning (TRA-TRB) pairs ranged between 65% and 80%, and the mean read pair count per cell was ∼3,600 with an average value of 62.4% of the reads detected in a total number of 37,347 cells in the six VDJ libraries.

The initial processing of the expression data involved generating fastq files and count matrices using Cell Ranger (Zheng et al., 2017) v6.0 (10X Genomics) that was run with the mm10-2020-A transcriptome reference, whereas the TCR data were processed with Cell Ranger v6.0 and the vdj_GRCm38_alts_ensembl-5.0.0 reference. HTO counts were generated using CITE-seq-Count v1.4.3 (https://github.com/Hoohm/CITE-seq-Count) with a mapping rate of 96% to HTO barcodes across the six HTO libraries (∼78 million total reads).

Downstream analysis integrating the mRNA, TCR, and HTO data modalities was carried out using the R packages Seurat v4.1.0 (Hao et al., 2021) (mRNA and HTO) and scRepertoire v1.11.0 (TCR) (Borcherding et al., 2021), respectively. Cells that met all of the following quality control–based criteria were retained for downstream analysis: (1) cells with a minimum gene count of 200 that was set using the min.features parameter of the CreateSeuratObject in Seurat, (2) cells identified as singlets by the R package scDblFinder (v1.8.0 with default settings) that operates on the RNA assay, (3) cells with a number of median absolute deviations (nmads) < 4 for mitochondrial content and nmads of 3 and 2.5 defining the range of the low and high gene (nFeature_RNA) and UMI (nCount_RNA) counts, respectively, as identified by the isOutlier function of the R package scuttle (v1.4.0), and (4) cells identified as singlets by the HTODemux function (positive.quantile set to “0.95”) of Seurat that operates on the HTO assay. The expression data were normalized using the NormalizeData function of Seurat, where the normalization.method parameter was set to “LogNormalize” and “CLR” for RNA and HTO assays, respectively. The integration of the data from individual sequencing libraries was carried out by the FindIntegrationAnchors and IntegrateData (dims parameter set to “1:20” for both) functions of Seurat.

Pairwise differential expression analysis was performed on the log-normalized RNA expression data with the FindMarkers function of Seurat (parameter settings of min.pct: 0.25, test.use: “MAST,” min.cells.group:50, and min.diff.pct = 0.2) utilizing Model-based Analysis of Single-cell Transcriptomics (MAST [Finak et al., 2015]), whereas the FindAllMarkers (parameter settings of min.pct = 0.25, test.use = MAST, min.cells.group = 30, min.diff.pct = 0.2) function was used to identify cluster-specific markers that were utilized for cell-type annotation. The FindClusters function of Seurat was used to cluster the cells with the default Louvain clustering setting and a resolution value of 0.3. Enrichment analysis was performed using the R package WebGestaltR (v0.4.5) (Liao et al., 2019).

TCR repertoire sequencing data were analyzed with the scRepertoire package v1.11.0 (Borcherding et al., 2021). Starting with the filtered Cell Ranger contig annotation output, combineTCR and combineExpression functions were used for combining the TCR data from each sequencing library and for integrating the combined TCR data and scRNA-seq datasets, respectively. We utilized scRepertoire along with the amino acid sequence of the paired CDR3 regions of the α and β (TRA-TRB) chains to define each clonotype and to quantify its abundance, whereas pairwise clonotype overlap between different cell populations was based on TRA-only clonotype definitions. Clonotypes were classified into five groups based on their abundance (or size): single (represented by one cell), small (2–5 cells), medium (6–20 cells), large (21–100 cells), and hyperexpanded (101–500 cells). Pairwise clonotype overlap between cell populations was quantified by the Jaccard index, which is the ratio between the number of common (intersection) set of unique clonotype sequences at the amino acid level and the number of combined (union) set of unique clonotype sequences. For hydrophobic index analyses, the conserved TCR Vβ germline-encoded cysteine in CDR3β was defined as position 1 of CDR3β. The hydrophobic index equals the percentage of cells in which positions 6 and 7 (P6-P7) of CDR3β correspond to any of the 175 amino acid doublets identified as promoting self-reactivity when present at P6–P7 of CDR3β (Stadinski et al., 2016). The figures associated with scRNA-seq and TCR analyses were generated using the R packages ggpubr (v0.4.0.999), ggplot2 (v3.4.2), scCustomize (v0.7.0), Seurat (v4.1.0), stringr (v1.5.1), tidyverse (v2.0.0), and clusterProfiler (v4.2.2).

Murine and human fecal samples were processed separately at each stage of DNA extraction, library preparation, and next-generation sequencing. DNA was extracted from fecal samples in two stages. The first stage consisted of adding 50 mg of fecal sample and MBL lysis buffer (Qiagen PowerMicrobiome DNA/RNA EP Kit) to Lysis Matrix E tubes (MP Biomedicals). Fecal samples were lysed on a Precellys 24 Tissue Homogenizer (Bertin Technologies) and centrifuged, and the resultant supernatant was transferred to a deep 96-well plate. The second stage consisted of DNA isolation from the above supernatant using Qiagen MagAttract PowerMicrobiome DNA/RNA EP Kit on an Eppendorf automated liquid handling system as detailed by the manufacturer.

The composition of the bacterial microbiome was assessed using a dual-index amplification and sequencing approach of the V4 region of the 16S ribosomal RNA (16S rRNA) gene on the Illumina MiSeq Platform. This method used the 16S rRNA 515F and 806R primers with individual sample-specific indexes and Illumina sequencing adapters appended as described in Kozich et al., (2013). The V4 was amplified using 5 μM of F/R primers, 1X Phusion High-Fidelity DNA Polymerase (New England Biolabs), and 100 ng of DNA as starting material. PCR conditions for amplification were as follows: initial template denaturation at 98°C for 60 s; 25 cycles of denaturation at 98°C for 10 s, primer annealing 55°C for 30 s, and template extension at 72°C 60 s; and a final template extension at 72°C for 5 min. Final PCR products were isolated using AMPure XP beads (Beckman-Coulter) at a 1:1 ratio.

Final 16S V4 libraries were quantified using KAPA qPCR Library Quantification Kit (Kapa Biosystems) and pooled at an equimolar concentration. Pools were normalized to 8 pM, spiked-in with 15% phiX control library to add sequence diversity to the pool, and sequenced on the Illumina MiSeq instrument utilizing 600-cycle Paired-End (250 × 250) Reagent Plate with the addition of 16S V4-specific sequencing primers as detailed in Kozich et al. (2013).

Raw amplicon sequences were assessed for quality with FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and MultiQC. Additionally, error correction (denoising) and amplicon sequence variant (ASV) calling were done with the DADA2 R package (Callahan et al., 2016) and SILVA taxonomic database v138.1 (Quast et al., 2013). Microbial community analysis was performed in R and additional packages: vegan (diversity) (https://CRAN.R-project.org/package=vegan), MMuPhin v 1.12.1 (batch correction) (https://doi.org/10.18129/B9.bioc.MMUPHin, Ma, 2025), ggplot2 (visualizations) (Wickham, 2016), MaASLin2 (differential abundance) (Mallick et al., 2021), and LmerTest (alpha diversity comparisons) (Kuznetsova et al., 2017). PICRUSt2 v2.4.1 (Douglas et al., 2020) was used for functional inference of the communities based on their taxonomic content in the Nephele application (https://nephele.niaid.nih.gov/pipeline_details/picrust2/).

To identify putative oral taxa in the gut microbiota, taxonomic classification of the ASVs was done with DADA2’s rdp function and the Mouse Oral Microbiome Database v5.1 for the mouse samples (Joseph et al., 2021). For the human microbiome samples, SourceTracker2 was used to estimate gut, oral, and skin source fraction of microbial communities between pRD patients and controls. Gut, oral, and skin source communities were downloaded from AGP (https://www.ebi.ac.uk/ena/browser/view/PRJEB11419) with sink rarefication depth set to 1,000. Additionally, the ASVs were classified using the Human Oral Microbiome Database v15.23 (Al-Hebshi et al., 2015).

The raw human microbiome dataset was represented by 1,232 ASVs and 396 species with a sequence depth of samples varying between 9.5K reads and 90.7K reads. ASVs were filtered at relative abundance of 0.01% and prevalence in at least three samples of the dataset. This retained 276 high-quality, well-represented ASVs and 163 species across all samples of the human dataset. The batch effect due to multiple sequencing runs was successfully reduced (from significance of 7%) to a nonsignificant contribution to community differences (at 5%; MMUPHin).

For the time course study, the raw mouse microbiome dataset was represented by 612 ASVs with a sequence depth ranging between 38,100 and 91,200 reads. After filtering for prevalence of at least 0.1% in five samples, 413 ASVs remained. For the BMT study, there were 426 ASVs, and after filtering, 337 remained.

IgA-seq on murine stools was performed as described previously (Vujkovic-Cvijin et al., 2022). Briefly, stool samples were placed in buffer (1% wt/vol bovine serum albumin in phosphate-buffered saline, used for all subsequent washing and staining steps), were physically disrupted using sterile pipette tips, and were homogenized by repeated pipetting using wide-bore pipette tips. Stool suspensions were spun at 50 g for 1 min to pellet nonbacterial food particles. The supernatant was collected and passed through a 40-μm filter, and washed by centrifugation at 8,000 rcf for 3 min (same speed and time for subsequent washes). OD₆₀₀ readings were taken of each suspension for normalization of input samples. Samples were stained with SYTO62 (1:500 dilution; Invitrogen) and anti-IgA (PE, clone: 11-44-2, dilution 1:40; eBioscience). All stains were performed for 15 min at 4°C. Stains were simultaneously performed using isotype control antibodies conjugated to PE, which yielded minimal to no staining for all experiments. Samples were then subjected to magnetic column–based separation using anti-PE microbeads (#130-048-801; Miltenyi) at a concentration of 1:20, and subsequent steps were followed as per the manufacturer protocol, and the IgA eluate was collected and spun by centrifugation to remove excess buffer prior to freezing. The IgA⁺ fraction was then subjected to additional column enrichment. The column used for the first enrichment was washed by passing ethanol through the column with a plunger followed by PBS, and the IgA⁺ sample was then applied and the manufacturer protocol was again performed. The final IgA+ fraction was centrifuged, and after removing excess buffer, it was frozen for downstream DNA extraction and 16S rRNA sequencing.

IgA-seq was performed on human stool in a similar fashion as described above and previously (Vujkovic-Cvijin et al., 2022), with the exception of using anti-human IgA PE (#130-093-128, 1:40 dilution; Miltenyi) for staining.

For murine IgA-seq, samples were subjected to rarefaction at 40,000 reads per sample, and for human IgA-seq, rarefaction at 30,000 reads per sample was performed. Dada2 was performed to generate ASV tables, and taxonomy was assigned using the RDP 18 database and the dada2 naïve Bayes classifier. For genus- and family-level analyses, ASVs were binned at the genus and family levels, respectively. Taxa with prevalence in <10% of samples were filtered out. IgA scores for each taxon were calculated by log10-transforming read counts and subtracting abundance of the IgA-enriched fraction from that of the IgA-depleted fraction.

For comparing human IgA scores between pRD and controls, linear mixed-effects models were used to control for age, sex, and country of residence with sex and country of residence considered random effects. For mouse IgA score comparisons between mutant and wild-type mice, Mann–Whitney U tests were performed.

For all liquid chromatography–mass spectrometry (LCMS) methods, LCMS grade solvents were used. All samples were immersed in 0.4 ml of ice-cold methanol. To each sample, 0.4 ml of water and 0.4 ml of chloroform were added. Samples were shaken for 20 min under refrigeration and centrifuged at 16,000 g for 20 min 400 µl of each of the top (aqueous) layer was collected. A subaliquot of the aqueous layer was taken for O-benzylhydroxylamine (O-BHA) derivatization of carboxylic acids and SCFA analysis. The remaining aqueous layer was diluted as necessary in 50% methanol in water for LCMS analysis of central polar metabolites.

Samples were derivatized with O-BHA according to previously established protocols (Jaochico et al., 2019; Zeng and Cao, 2018). Reaction buffer was prepared fresh consisting of 1 M pyridine and 0.5 M hydrochloric acid in water. A 35 µl aliquot of the aqueous extract was taken and to the sample was added 10 µl of 1 M O-BHA in reaction buffer and 10 µl of 1 M 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide in reaction buffer. Samples were shaken at room temperature for 2 h. Each sample was quenched with 50 µl of 0.1% formic acid for 10 min. Derivatized carboxylic acid compounds were extracted with the addition of 400 µl ethyl acetate. Samples were centrifuged at 16k g for 10 min at 21°C to induce layering, and 200 μl of the upper (organic) layer was collected. The extract was dried under vacuum, and each sample was resuspended in 200 µl of water for LCMS injection.

Tributylamine and all synthetic molecular references were purchased from MilliporeSigma. LCMS grade water, methanol, isopropanol, and acetic acid were purchased through Thermo Fisher Scientific.

Aqueous metabolites were analyzed using a combination of two analytical methods with opposing ionization polarities (Groveman et al., 2023; McCloskey et al., 2015). Both methodologies utilized an LD40 XR UHPLC (Shimadzu Co.) system for separation. Negative mode metabolites were detected using a 5,500 QTrap, and positive mode metabolites were detected with a 6,500+ QTrap mass spectrometer (AB Sciex Pte. Ltd.). Negative mode samples were separated on a Waters Atlantis T3 column (100 Å, 3 µm, 3 × 100 mm) and eluted using a binary gradient from 5 mM tributylamine, 5 mM acetic acid in 2% isopropanol, 5% methanol, 93% water (vol/vol) to 100% isopropanol over 5 min. Two distinct multiple reaction monitoring (MRM) pairs in negative mode were used for each metabolite. Positive mode method samples were separated across a Phenomenex Kinetex F5 column (100 Å, 2.6 µm, 100 × 2.1 mm) and eluted with a gradient from 0.1% formic acid in water to 0.1% formic acid in acetonitrile over 5 min.

Derivatized SCFA samples were analyzed using an LD40 XR UHPLC (Shimadzu Co.) system for separation and a 6,500+ QTrap mass spectrometer (AB Sciex Pte. Ltd.) for detection. Samples were separated with a Waters Atlantis dC18 column (100 Å, 3 µm, 3 × 100 mm) using a 6-min gradient from 5 to 80 B with buffer A consisting of 0.1% formic acid in water and B consisting of 0.1% formic acid in methanol. SCFAs and central metabolic carboxylic acids were detected using positive mode MRMs from previously established methods, and identity was confirmed by comparison with derivatized standards (Jaochico et al., 2019; Zeng and Cao, 2018).

All signals were integrated using Sciex OS 3.1 (AB Sciex Pte. Ltd.). Signals with >50% missing values for a specific tissue set were discarded, and remaining missing values were replaced with the lowest registered signal value. Where appropriate, signals with a quality control coefficient of variance >30% were discarded. Metabolites or lipids with multiple MRMs were quantified with the higher signal-to-noise MRM. Filtered datasets of the negative mode aqueous metabolites and the bulk lipidomics were total sum-normalized after initial filtering. The SCFA dataset and the positive mode aqueous metabolomics dataset were scaled and combined with the negative mode aqueous metabolite dataset using common signal for citrate and serine, respectively. A Benjamini–Hochberg method for correction for multiple comparisons was imposed where indicated.

Fig. S1 shows mean recombination activity of mutant RAG alleles and colonic inflammation in pRD. Fig. S2 shows characterization of colonic and systemic inflammation in Rag1^w/w (W/W) mice. Fig. S3 shows scRNA-seq of LP CD4⁺ T cells in W/W and wild-type mice. Fig. S4 shows composition and diversity of fecal microbiota in W/W and wild-type mice. Fig. S5 shows effects of various treatment modalities on intestinal and systemic inflammation, and on TCR repertoire composition.

The mRNA, VDJ (TCR), and HTO sequencing data underlying Figs. 3, 5, 6, S3, and S5 have been deposited on GEO under the accession number GSE277151, whereas the analysis code and scripts are available on https://github.com/cihangenome/immunological-microbial-signatures-RAG-deficiency. The raw microbiome sequence and IgA-seq data underlying Figs. 4, 6, 7, and S4 have been deposited in NCBI’s Sequencing Read Archive database under the Bioproject number PRJNA1159340 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1159340/).

The authors thank Sean P. Conlan and Clayton Deming from the Microbial Genomics Section, Translational and Functional Genomics Branch, National Human Genome Research Institute, National Institutes of Health, Bethesda, MD, USA, for their support in the microbiota characterization of the mouse model.

This work was supported by the Division of Intramural Research of the National Institute of Allergy and Infectious Diseases (NIAID), National Institutes of Health (NIH) (grant AI001222 to L.D. Notarangelo). This project has been funded in part with Federal funds from the NIAID, NIH, Department of Health and Human Services under Bioinformatics and Computational Biosciences Branch Support Services Contract HHSN316201300006W/75N93022F00001 to Guidehouse Digital. R. Castagnoli developed part of this project while being a participant in the U.S. National Institutes of Health (NIH) Graduate Partnership Program. R. Castagnoli has been awarded the Fellowship “Progressi in Biologia e Medicina” from Fondazione Ghislieri, Collegio Ghislieri, Pavia, Italy, to support the investigation on microbiota–host interactions in inborn errors of immunity. I. Vujkovic-Cvijin was supported by NHLBI, NIH (grant DP1HL174182), NIDDK, NIH (grant U01DK062413), the Crohn’s & Colitis Foundation Career Development Award #831262, and the F. Widjaja Inflammatory Bowel Disease Institute of Cedars-Sinai Medical Center.

Author contributions: R. Castagnoli: conceptualization, data curation, formal analysis, investigation, methodology, project administration, validation, visualization, and writing—original draft, review, and editing. F. Pala: conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, and writing—original draft, review, and editing. P. Subramanian: data curation, formal analysis, methodology, software, validation, visualization, and writing—original draft, review, and editing. C. Oguz: data curation, formal analysis, software, and writing—original draft, review, and editing. B. Schwarz: data curation, formal analysis, investigation, methodology, resources, and visualization. A.I. Lim: investigation. A.S. Burns: investigation, resources, validation, and writing—review and editing. E. Fontana: investigation. M. Bosticardo: investigation and writing—review and editing. C. Corsino: investigation and project administration. A. Angelova: formal analysis. O.M. Delmonte: resources. H. Kenney: investigation, project administration, and resources. D. Riley: investigation and writing—review and editing. G. Smith: investigation, methodology, resources, and writing—review and editing. L. Ott de Bruin: writing—review and editing. V. Oikonomou: investigation. L. Dos Santos Dias: investigation and writing—review and editing. D. Fink: investigation. E. Bohrnsen: conceptualization, data curation, formal analysis, and writing—review and editing. C.D. Kimzey: formal analysis and investigation. G.L. Marseglia: resources. G. Alva-Lozada: resources. J.R.E. Bergerson: resources and writing—review and editing. A. Brett: resources. K.W. Brigatti: investigation, project administration, and resources. D. Dimitrova: resources. C.M. Dutmer: investigation, resources, and writing—review and editing. A.F. Freeman: resources, and writing—review and editing. H. Ale: resources. S.M. Holland: funding acquisition, investigation, and resources. F. Licciardi: investigation. S. Pasic: resources. L.E. Poskitt: data curation. D.E. Potts: data curation, investigation, and writing—review and editing. J.F. Dasso: resources and writing—review and editing. S.O. Sharapova: resources. K.A. Strauss: conceptualization, data curation, project administration, resources, supervision, and writing—review and editing. B.R. Ward: resources. M. Yilmaz: resources. D.B. Kuhns: formal analysis, investigation, and resources. M.S. Lionakis: funding acquisition, investigation, supervision, and writing—review and editing. S.R. Daley: formal analysis and software. H.H. Kong: conceptualization, investigation, resources, and writing—review and editing. J.A. Segre: methodology and writing—review and editing. A. Villa: conceptualization. S. Pittaluga: resources and visualization. J.E. Walter: data curation, funding acquisition, resources, and writing—original draft, review, and editing. I. Vujkovic-Cvijin: formal analysis, investigation, methodology, software, validation, visualization, and writing—review and editing. Y. Belkaid: conceptualization, investigation, resources, and supervision. L.D. Notarangelo: conceptualization, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, and writing—original draft, review, and editing.