Selection of cross-reactive T cells by commensal and food-derived yeasts drives cytotoxic TH1 cell responses in Crohn’s disease

Aberrant CD4+ T cell reactivity against intestinal microorganisms is considered to drive mucosal inflammation in inflammatory bowel diseases. The disease-relevant microbial species and the corresponding microorganism-specific, pathogenic T cell phenotypes remain largely unknown. In the present study, we identified common gut commensal and food-derived yeasts, as direct activators of altered CD4+ T cell reactions in patients with Crohn’s disease (CD). Yeast-responsive CD4+ T cells in CD display a cytotoxic T helper cell (TH1 cell) phenotype and show selective expansion of T cell clones that are highly cross-reactive to several commensal, as well as food-derived, fungal species. This indicates cross-reactive T cell selection by repeated encounter with conserved fungal antigens in the context of chronic intestinal disease. Our results highlighted a role of yeasts as drivers of aberrant CD4+ T cell reactivity in patients with CD and suggest that both gut-resident fungal commensals and daily dietary intake of yeasts might contribute to chronic activation of inflammatory CD4+ T cell responses in patients with CD.

Aberrant CD4 + T cell reactivity against intestinal microorganisms is considered to drive mucosal inflammation in inflammatory bowel diseases. The disease-relevant microbial species and the corresponding microorganism-specific, pathogenic T cell phenotypes remain largely unknown. In the present study, we identified common gut commensal and food-derived yeasts, as direct activators of altered CD4 + T cell reactions in patients with Crohn's disease (CD). Yeast-responsive CD4 + T cells in CD display a cytotoxic T helper cell (T H 1 cell) phenotype and show selective expansion of T cell clones that are highly cross-reactive to several commensal, as well as food-derived, fungal species. This indicates cross-reactive T cell selection by repeated encounter with conserved fungal antigens in the context of chronic intestinal disease. Our results highlighted a role of yeasts as drivers of aberrant CD4 + T cell reactivity in patients with CD and suggest that both gut-resident fungal commensals and daily dietary intake of yeasts might contribute to chronic activation of inflammatory CD4 + T cell responses in patients with CD.
Inflammatory bowel diseases (IBDs) are chronic inflammatory disorders of the gastrointestinal tract, the main forms of which are CD and ulcerative colitis (UC). Control of the immune process through conventional or targeted therapies is poor in these chronic diseases, which results in lifelong morbidity, strongly reduced quality of life and high medical costs. Aberrant immune reactions, particularly CD4 + T cell responses, against members of the intestinal microbiota are considered a causal or driving factor for an IBD. However, the high diversity of the and inflamed tissue of patients with CD, but not those with UC. These yeast-reactive CD4 + T cells display cytotoxic T H 1 cell effector functions, whereas C. albicans-specific T H 17 cell responses remain largely unchanged. Cytotoxic T H 1 cells are clonally expanded and highly cross-reactive to several commensal and food-derived yeast species, indicating their selection by chronic encounter with conserved antigens that are present in different yeasts. Our data identify commensal and dietary yeasts as drivers of aberrant CD4 + T cell reactivity in patients with CD and suggest that repeated encounter with conserved antigens present in different microbial species may lead to the expansion and chronic activation of cross-reactive, yeast-responsive, CD4 + T cells. This could reveal a general mechanism of how selection of cross-reactive T cells may allow adaptive immunity to cope with the enormous diversity of microbial antigens which, if not properly regulated, could come at the expense of fostering chronicity and contribution to therapy resistance in IBD due to persistent antigen activation.

CD4 + T cell responses against yeasts are increased in CD
To identify microbial species driving pathogenic T cell responses in IBDs, we analyzed the CD4 + T cell reaction against common bacterial and fungal species of the human intestinal microbiome in patients with IBD and healthy donors (see Extended Data Tables 1 and 2 for demographic data). Peripheral blood mononuclear cells (PBMCs) were stimulated ex vivo for 7 h with whole bacterial and fungal lysates. The differentially stimulated cells were labeled with CD4 antibody-based fluorescent barcodes, allowing multiplexed analysis of T cell reactivity against different species (Extended Data Fig. 1a). Microorganism-reactive CD4 + T cells were detected based on magnetic enrichment of CD154 (CD40L)-expressing cells (antigen-reactive T cell enrichment (ARTE)) 15,28 (Extended Data Fig. 1a,b). Specificity of CD154 + induction by microbial lysates was confirmed by blocking antigen presentation with a human leukocyte antigen (HLA)-DR antibody (Extended Data Fig. 2a-c) and by restimulation of expanded CD154 + T cells with specific or unrelated antigen lysates (Extended Data Fig. 2d).
Circulating CD45RA − memory T cells (T mem cells) reactive against different microbial species were detected with variable frequencies (range 0.0001-0.76%) (Fig. 1a,b). Although C. albicans, the most abundant intestinal fungal commensal 20,29 (Fig. 1c), elicited the strongest reactivity in healthy individuals, we observed only a slight increase in reactive T mem cells in patients with CD. In sharp contrast, reactivity against other less abundant Candida spp., as well as against the food-associated yeast Saccharomyces cerevisiae (Fig. 1c), was dramatically elevated in patients with CD, but not those with UC ( Fig. 1a,b). No significant differences were observed against the phylogenetically distant yeast Malassezia restricta or the analyzed bacterial species (Fig. 1a,b). Together this indicates an altered antifungal immune response to gut mycobiota in patients with CD. However, we also microbiome has largely prevented identification of disease-driving, microbial T cell targets and the corresponding microorganism-specific, pathogenic T cell phenotypes in IBDs. Furthermore, owing to the complexity of IBDs, pathogenic CD4 + T cell phenotypes and disease-relevant antigens may also vary for different patient subgroups.
T H 17 cells are regarded as central orchestrators of homeostatic immune responses against microorganisms. Dysregulated T H 17 cell responses, in turn, are thought to contribute to intestinal inflammation and increased levels of interleukin (IL)-17A-producing cells can be found in the intestinal mucosa of patients with IBD 1,2 . However, neutralization of IL-17 has been clinically ineffective in CD and even caused exacerbation in some patients 3,4 , suggesting that IL-17A could play a protective rather than a pathological role in CD or in a subgroup of these patients. In this regard, T H 17 cell plasticity and the development of pathogenic T H 17 cell subsets that coexpress inflammatory cytokines, such as interferon (IFN)-γ or granulocyte-macrophage colony-stimulating factor (GM-CSF), might be important determinants for a protective versus a pathogenic role for T H 17 cells 5 . In addition, T H 1 cells expressing the signature cytokine IFN-γ are consistently found to be increased in the intestine of patients with CD, but not those with UC 6 . Thus, deciphering the disease-relevant microbial antigens and unique features of pathogenic, microorganismspecific, CD4 + T cell responses could give rise to the development of targeted therapies.
Most studies to date have addressed alterations in the intestinal bacterial microbiota in IBDs and few studies have analyzed bacteriaspecific CD4 + T cell responses [7][8][9][10][11][12][13][14] . Although bacteria are more abundant organisms in the gastrointestinal tract, fungi are an integral, but largely neglected, part of the human intestinal microbiota. Only recently has the important role of individual fungal species on human health and immunopathology come into focus [15][16][17][18][19][20][21][22] . In particular, the intestinal commensal fungus Candida albicans has been identified as a major modulator of the human immune system by inducing homeostatic T H 17 cell responses 15,23 , as well as intestinal immunoglobulin (Ig)A 17,22 and systemic IgG antibodies 16 . It is still unclear whether C. albicans-specific T H 17 cells contribute to intestinal inflammation in patients with IBD or are recruited to the site of inflammation to foster healing. Most interestingly, a high prevalence of anti-Saccharomyces cerevisiae antibodies (ASCAs) has been consistently reported in the serum of patients with CD. Although S. cerevisiae mannans were usually used as a binding antigen in these assays, further studies have demonstrated that ASCAs also bind to mannans from other fungi, including C. albicans 24 . In addition to C. albicans, the human intestine also harbors a plethora of other commensal, as well as food-derived, fungi [25][26][27] . Whether these fungal species can also actively modulate human immune responses has not yet been addressed.
In the present study, we show that CD4 + T cell responses against commensal and food-derived yeasts are strongly increased in blood e, Frequencies of fungus-reactive CD4 + T cells in patients with CD, according to their ASCA status. f, Total CD4 + T mem cells isolated from biopsies of inflamed and uninflamed intestinal tissue of patients with CD who are ASCA + (n = 5). Cells were seeded at 200 cells per well in multiple wells, and expanded and restimulated with fungal lysates in the presence of autologous APCs. Mean frequencies of reactive CD154 + cells for all wells per patient are shown. g, Dot plot examples for restimulation of expanded CD4 + T mem cells from an inflamed biopsy. The percentage of CD154 + TNF + cells within CD4 + is indicated. Each symbol in a, d, e and f represents one individual donor and horizontal lines in a and e indicate the mean. Box-and-whisker plots are shown in c with center lines representing the median, box limits 25% (lower) and 75% (upper) quartiles and whiskers the minimum/maximum. Truncated violin plots with quartiles and range are shown in f. Statistical differences were obtained using the Kruskal-Wallis test with Dunn's post hoc test in a with only significant differences shown; the two-tailed Mann-Whitney U-test was used in e. Article https://doi.org/10.1038/s41591-023-02556-5 noticed a high heterogeneity in the frequencies of antifungal T cells between different patients with CD (Fig. 1a). ASCAs can be detected in the serum of 50-60% of patients with CD and are used as a biomarker of disease severity 30 . When stratifying our cohort of patients with CD according to their ASCA immunoglobulin IgG/IgA status (Fig. 1d)    in ASCA + , but not ASCA − , patients with CD (Fig. 1e). In contrast, the strong T mem cell response to C. albicans was independent of the ASCA status.
These results indicate that, in addition to augmented T mem cell responses to commensal C. albicans in patients with CD 15 , a cohort of ASCA + patients develops unexpected T cell reactivity to less represented fungal commensals and food-derived yeast species that so far have not been described as major inducers of T cell reactivity in humans.

Yeast-reactive CD4 + T cells are enriched in inflamed mucosa
We next analyzed whether yeast-responsive T cells are also present at the site of intestinal inflammation. As the cell numbers obtained from intestinal biopsies are too low for direct analysis by ARTE, we expanded total CD4 + T cells from inflamed and noninflamed gut biopsies polyclonally and restimulated them with various fungal lysates. Yeast-reactive T cells were selectively present in inflamed gut tissue from patients who were ASCA + , but not in noninflamed tissue from the same individuals (Fig. 1f,g), or in tissue from patients with CD     Fig. 2e). Notably, this applied also to other Saccharomyces spp., such as Saccharomyces pastorianus or the probiotic yeast Saccharomyces boulardii, which is sometimes used as a treatment supplement for IBD 31 , whereas reactivity to C. albicans was low (Fig. 1f). We further examined whether the increased T cell reactivity against yeasts might be influenced by disease state or duration. There were no significant correlations of yeast-reactive T cell frequencies with disease duration, medication or activity, as determined by the Harvey-Bradshaw index and the CD activity index (Extended Data Fig. 3).
In summary, these data reveal an unexpected increase of CD4 + T cell reactivity against intestinal commensal and food-derived yeasts in blood and inflamed tissue in patients with CD who are ASCA + . Although C. albicans dominates the antifungal T cell response in healthy individuals and C. albicans-reactive T cells are still highly present in patients with CD, the increased T cell reactivity in patients with CD who are ASCA + was mainly directed against non-albicans Candida spp., as well as food-associated Saccharomyces spp.

Yeast-reactive CD4 + T cells show increased IFN-γ production
Intestinal inflammation in IBD has been linked to aberrant T H 17 cell responses and the development of pathogenic T H 17 cells which coproduce inflammatory cytokines such as IFN-γ 1,2 . We recently showed that, within a number of species found in the human mycobiome, C. albicans is the major driver of fungus-reactive T H 17 cells in humans 15 . However, we detected no significant differences of fungus-reactive IL-17A production in patients with IBD compared with healthy donors (Fig. 2a,b). In contrast, CD4 + T cells reactive against the different yeast species, including C. albicans, showed a massively increased IFN-γ production in patients with CD ( Fig. 2a,b), which was again strongly enhanced in individuals who are ASCA + compared with those who are ASCA − (Fig. 2c). Just as for the frequencies (Fig. 1), the increased IFN-γ production was much more pronounced for non-albicans Candida spp. and especially for S. cerevisiae. We observed only low coexpression of IFN-γ and IL-17A (Fig. 2a), indicating selective T H 1 cell expansion rather than T H 1 cell-like modulation of pre-existing T H 17 cells. However, small populations (≤5%) of IL-17A + IFN-γ + coproducing cells were also increased against Candida spp. and S. cerevisiae (Extended Data Fig. 4a,b). Further characterization of yeast-reactive T cells revealed enhanced production of several inflammatory cytokines in response to S. cerevisiae and Candida tropicalis, including IL-2 and GM-CSF, whereas IL-10 production was significantly reduced in patients with CD who are ASCA + ( Fig. 2d and Extended Data Fig. 4c,d). Again, we observed no differences of M. restricta-reactive T cells (Extended Data Fig. 4d).
Collectively, these data show strongly increased inflammatory T H 1 cell responses against Candida and Saccharomyces yeasts in patients with CD who are ASCA + and only a minor development of IL-17A + IFN-γ + cells. In contrast, the proportion of IL-17A producers within C. albicans-specific T cells was conserved and remained largely unaffected by the disease.

Yeast-reactive T H 1 cells have a cytotoxic signature
To obtain deeper insights into the cellular composition of yeast-reactive T cells and their functional capacities, we performed single-cell RNA sequencing (scRNA-seq) of ex vivo FACS-purified, C. albicans-, C. tropicalis-and S. cerevisiae-reactive CD154 + T mem cells from three patients with CD who are ASCA + and three healthy donors (Fig. 3a). In total, we sequenced 15,012 C. albicans-, 21,623 C. tropicalis-and 23,495 S. cerevisiae-reactive CD4 + T cells. Within these cells, we identified eight clusters with a distinct transcriptional profile based on Leiden unsupervised clustering (Fig. 3b,c). Among these clusters, we detected again a dominant T H 1 cell population (marker genes IFNG and TBX21) that was additionally characterized by expressing high levels of cytotoxic marker genes (for example, PRF1, GZMB and SLAMF7) (Fig. 3c). In fact, cells within this cluster expressed a number of genes previously described for cytotoxic CD4 + T cells that arise in viral infections or cancer 32 (Extended Data Fig. 5a,b). Furthermore, these cells expressed high levels of HOPX, encoding a transcription factor that has been associated with chronic T cell activation 33,34 . The T H 1 cell-cytotoxic cluster was strongly increased in patients with CD compared with healthy donors (Fig. 3b,d). We also noticed the increase of an effector cluster in C. tropicalis-and S. cerevisiae-reactive cells from patients with CD expressing low levels of T H 17 cell signature genes (IL17A and IL22), but also of T H 1 cell marker genes IFNG and TBX21, as well as CSF2 and IL2 (Fig. 3c,d). Discrimination of healthy and patient-derived cells in this cluster revealed that CD-derived cells did not upregulate T H 17 cell genes, but rather several other inflammatory genes defining this cluster (that is, IFNG, TBX21, CSF2 and IL2) (Extended Data Fig. 5c). This is in line with the low proportion of IL-17A protein-expressing cells (Fig. 2b) and was further confirmed by direct hashtag antibody labeling of IL-17A protein-secreting T cells, which did not increase in patients with CD (Extended Data Fig. 5d,e). The other identified gene clusters either remained unchanged or were reduced in patients with CD versus controls or in general comprised only a small proportion (≤5%) of cells (Fig. 3d). 5,34 . Based on intracellular cytokine staining, we observed only a minor proportion of cells coexpressing IFN-γ and IL-17A cytokines (Fig. 2a). To further analyze the developmental origin of the yeast-reactive cytotoxic T H 1 cells (T H 1-CTLs), we next performed a pseudotime analysis, which reveals pseudo-temporal ordering of cells according to gene expression similarity. Using the Monocle algorithm 35 , we identified two major trajectories, starting from central memory cells (T CM cells) to either the effector/T H 17 cell cluster or the T H 1-CTL lineage (Fig. 3e). This supports the conclusion that the T H 1-CTLs do not primarily convert from T H 17 cell precursors. It is interesting that expression of IFNG and TBX21 preceded the expression of cytotoxic marker genes along the T H 1-CTL branch, pointing to a progressive acquisition of cytotoxic effector functions in T H 1 cells over time (Fig. 3f).

Development of T H 17 cells into pathogenic IL-17A + IFN-γ + coproducing cells or even a complete conversion of T H 17 cells into T H 1 cells has been suggested to occur in chronic inflammation
Together, our data reveal that, in CD, but not under a steady state, exposure to yeast antigens results in strong expansion of yeast-reactive T H 1 cells with a cytotoxic signature.

T H 1-CTLs have killing ability for IECs
We confirmed the protein expression of cytotoxic markers by yeast-reactive CD4 + T H 1 cells in patients with CD by flow cytometry  Fig. 4a,b). Again, T H 1-CTLs were present in patients with CD who were ASCA + , but not in those who were ASCA − (Fig. 4c and Extended Data Fig. 5f,g). To further investigate the killing capacity of yeast-reactive T H 1-CTLs, we set up an in vitro cytotoxicity assay (Fig. 4d). CD154 + T mem cells reactive to S. cerevisiae or M. restricta, used as a control antigen, were ex vivo isolated from healthy individuals and patients with CD Type-I IFN signature  who were ASCA + and cocultured with allogeneic primary intestinal epithelial cells (IECs) in different ratios. T cells and IECs were cross-linked by the addition of Staphylococcus enterotoxin B (SEB) and the killing of the IECs was measured by real-time cell analysis. In line with their cytotoxic signature, S. cerevisiae-reactive CD4 + T cells from patients with CD who were ASCA + , but not from healthy donors, lysed IECs in a dose-dependent manner. No cytotoxicity was observed when M. restricta-reactive CD4 + T cells were used (Fig. 4e,f). These data confirm that yeast-responsive T H 1 cells of patients with CD who are ASCA + have cytotoxic effector functions that result in potent killing ability of IECs in vitro. IECs have been reported to function as nonconventional antigen-presenting cells (APCs), presenting antigens for intestinal microorganisms 36 . It is therefore tempting to speculate that yeast-reactive T H 1-CTLs might promote killing of IECs in vivo and thus could be involved in intestinal barrier deterioration in patients with CD, which needs further investigation.

Yeast-reactive CD4 + T cells are altered in IBD-FDRs
So far, it remains unclear whether the development of yeast-responsive T H 1-CTLs is a consequence of the disease or could eventually contribute to disease development and/or exacerbation. To address potential alterations of yeast-responsive T cells already in a predisease state, we analyzed a cohort of first-degree relatives of patients with IBD (IBD-FDRs), who have increased risk of developing IBD themselves 37,38 . We detected neither increased frequency nor increased IL-17A production of C. tropicalis-and S. cerevisiae-reactive T cells in IBD-FDRs compared with non-FDR controls (Extended Data Fig. 6a,b). However, C. tropicalis-and S. cerevisiae-reactive T cells from IBD-FDRs showed significantly increased IFN-γ production (Extended Data Fig. 6c,d), although less pronounced than for patients with CD (Fig. 2), whereas Escherichia coli-reactive cells, used as a control, were not altered (Extended Data Fig. 6c). In contrast to IFN-γ, cytotoxic markers were not significantly increased compared with non-FDR controls (Extended Data Fig. 6e). However, individual IBD-FDRs showed clearly elevated    levels of CD319 and PRF1 on C. tropicalis-and/or S. cerevisiae-reactive T cells, whereas GZMB and CCL4 remained unchanged (Extended Data Fig. 6f). ASCA antibodies have been reported as a predictive biomarker that can be detected several years before the clinical diagnosis of CD 39,40 . Within our limited number of IBD-FDRs, we identified individuals with increased serum levels of ASCA IgG and IgA (Extended Data Fig. 6g). As for patients with CD, increased expression of cytotoxic markers was associated with elevated ASCA antibody titers in IBD-FDRs (Extended Data Fig. 6h).
In summary, these data indicate that a skewing toward a T H 1 cell phenotype of yeast-responsive T cells is already detectable in IBD-FDRs and thus before the clinical onset of IBD. They further suggest that the additional acquisition of cytotoxic effector functions may evolve progressively, maybe as a result of a more intense or sustained interaction with yeast antigens, due to increased intestinal permeability, which has been described in IBD-FDRs 41 .

Clonally expanded T H 1-CTLs cross-react to multiple yeast species
Our data so far suggest that chronic antigen stimulation may foster T H 1-CTL development in patients with CD who are ASCA + . To better understand how chronic exposure to a large variety of commensal and dietary fungi may affect the clonal composition of yeast-reactive T cells, we integrated T cell receptor (TCR) sequencing data from our scRNA-seq dataset. Indeed, T H 1-CTLs from patients with CD showed a massive expansion of individual clonotypes (Fig. 5a), further evidenced by a higher Gini's coefficient, as a measure of the evenness of a population (Fig. 5b). Strikingly, many of the S. cerevisiae-stimulated TCRs of patients with CD were shared with C. tropicalis-and C. albicans-stimulated cells, whereas shared TCRs for the different fungi were less abundant in healthy donors (Fig. 5c,d). Indeed, most of the strongly expanded clonotypes were shared between the TCR repertoires reactive against two or all three fungi (Fig. 5d). TCR cross-reactivity was most pronounced within the T H 1-CTL cluster, where >70% of the S. cerevisiae-reactive TCR-β sequences of patients with CD were also found in the C. tropicalis-and C. albicans-stimulated cells (Fig. 5e). This suggests that persistent or alternating exposure to conserved antigens present in different yeast species may promote the selective expansion of cross-reactive clonotypes. In particular, these cross-reactive clonotypes, but not clonotypes with a single species specificity, showed high expression of cytotoxic marker genes, as well as HLA-DR and HOPX, encoding for recent activation and chronic markers, respectively (Fig. 5f).
To further confirm the T cell cross-reactivity for individual TCRs, we re-expressed the most expanded cross-reactive T H 1-CTL TCRs from the single-cell dataset in primary T cells using clustered regularly interspaced palindromic repeats (CRISPR)-Cas9 42,43 (Extended Data Fig. 7a,b). The expanded TCR-transgenic T cells were re-challenged with a panel of different yeasts in the presence of autologous antigen-presenting cells (APCs). As shown in Fig. 5g-i, we observed high cross-reactivity of the transgenic T cells to several yeast species and the reactivity of individual TCRs further reflected the cross-reactivity pattern initially observed by TCR-seq (Fig. 5i). High cross-reactivity to several yeast species was also confirmed for ex vivo sorted and expanded, S. cerevisiae-reactive, cytotoxic, CD154 + CD319 + single-cell clones (Fig. 5j). Finally, to validate the selective enrichment of cross-reactive cells in patients who were ASCA + , we expanded bulk C. albicans-, C. tropicalis-and S. cerevisiae-reactive CD154 + T mem cells from patients with CD. Yeast-reactive T cells derived from patients with CD who are ASCA + , but not ASCA − , highly cross-reacted to various yeast species (Extended Data Fig. 7c,d), consistent with the presence of these cells in patients who are ASCA + . Cytotoxic CD4 + T cells have so far been mainly described in the context of persistent viral infections 32 . However, none of the TCR-transgenic T cells reacted to common viral antigens and we also observed no differences between patients with CD who are cytomegalovirus (CMV) positive and those who are CMV negative in the frequencies of yeast-responsive T cells (Extended Data Fig. 7e-g).
Taken together, these data show that CD4 + T H 1-CTLs that cross-recognize various yeast species are strongly expanded in patients with CD who are ASCA + . The repetitive encounter with conserved antigens, present in different yeast species, may thus lead to the selective expansion of the cross-reactive clones, which progress to cytotoxic cells owing to chronic stimulation in patients with CD.

CD4 + T H 1-CTLs respond to food-derived yeast antigens
The finding of a substantial cross-recognition of various yeast species by the CD4 + T H 1-CTLs prompted us to investigate potential further antigenic triggers of these cells. As several yeast species are widespread organisms and frequently consumed with the diet, we hypothesized that food-derived yeast species could also promote the activation of CD4 + T H 1-CTLs from patients with CD. Cheese contains live microorganisms, including various Candida, Saccharomyces, Penicillium and several other fungal species 44,45 . To analyze whether food-derived fungal species could activate CD4 + T H 1-CTLs in patients with CD, we stimulated PBMCs with heat-inactivated lysates of common cheese types and analyzed the reactive T cells by ARTE. Patients with CD who are ASCA + , but not healthy donors, showed clearly increased reactivity, especially to the mold-ripened cheeses camembert and gorgonzola (Fig. 6a,b). In line with our previous results, these cheese-activated T cells from patients with CD produced increased amounts of IFN-γ (Fig. 6b) and expressed elevated levels of cytotoxic markers (Fig. 6c). To confirm that the T cell reactivity observed with the cheese lysates was the result of T cell cross-reactivity against yeasts, we expanded gorgonzola-activated CD154 + T mem cells and restimulated them with different microbial species. Indeed, expanded T cell lines from patients with CD who are ASCA + exhibited up to 100% cross-reactivity to several yeast species, but not bacteria or milk proteins (Fig. 6d,e), indicating that the observed reactivity against the cheese is probably directed  against the yeast species present in it. This was further confirmed by reactivity of the TCR-transgenic cells against the gorgonzola lysate (Fig. 6f,g), as well as against the yeast Kluyveromyces lactis (formerly Saccharomyces lactis), whereas reactivity to phylogenetic distant Penicillium or Geotrichum mold species, which are also frequently contained in cheese 44,45 , was low.  In summary, our data reveal that, in addition to gut-resident commensals, food-derived yeasts can also activate the cytotoxic T H 1-CTLs in patients with CD. The selective expansion of cross-reactive CD4 + T cells by different members of the mycobiota, as identified in the present study, may contribute to chronicity and treatment resistance of CD, when persistently activated by different commensal and food-derived fungi.

Discussion
An inappropriate immune response against intestinal microorganisms has been suggested as a major contributor to the pathophysiology of CD. Despite extensive data supporting the involvement of T cells in IBD pathogenesis, the relevant intestinal antigens and the effector functions of the corresponding antigen-specific T cells, remain largely unknown. In the present study, we provide direct evidence that common intestinal fungal commensals, as well as food-derived yeasts, contribute to aberrant inflammatory CD4 + T cell responses in patients with CD.
We observed massively increased CD4 + T cell responses against different yeast species in patients with CD. In addition to C. albicans, which has been identified as a strong modulator of human adaptive immunity [15][16][17]22,46 , the altered T cell response against non-albicans Candida and food-associated Saccharomyces spp. adds an additional layer to the inflammatory immune reaction in CD. Development of pathogenic T H 17 cells has been discussed as contributing to IBD pathology and we recently showed that C. albicans is the major driver of fungus-induced homeostatic T H 17 cell responses in humans 15 . In the present study, we observed that the proportion of T H 17 cells against C. albicans was surprisingly unaffected in patients with CD, but that cytotoxic T H 1 cells develop that also strongly react against non-albicans Candida spp., as well as different Saccharomyces yeasts. Which fungal species is the initial trigger of the T H 1 response and which fungi merely drive the selection of cross-reactive TCR remains unresolved. Recently, Debaryomyces hansenii, a yeast commonly contained in dairy products, has been described as being specifically enriched in inflamed wounds of patients with CD 18 , suggesting that food-derived fungi may also adapt to the altered environment in the IBD gut. In this way, both intestinal fungal commensals and food-derived species could be involved in the initial priming of the T H 1 cell response observed here. Once established, these T H 1 cells may be propagated by multiple fungal species, leading to the evolution of a highly selected cross-reactive TCR repertoire.
The limited impact of IBD on the T H 17 cell reaction against C. albicans suggests that the strong homeostatic T H 17 cell response present in most individuals 15 is relatively inert to perturbations. Why T H 17 cell responses against C. albicans are less affected by IBD remains unclear. It is possible that fungus-reactive T H 1 and T H 17 cells recognize different antigens that have different availability in CD than in healthy individuals. Indeed, we detected high TCR cross-reactivity within the yeast-responsive T H 1-CTLs, compared with T H 17 cells. This indicates that T H 1-CTLs, but not T H 17 cells, are directed against conserved epitopes that are commonly present in many yeast species, thereby also increasing antigen abundance. Alternatively, a conversion of T H 17 into T H 1 cells may occur on continuous antigen stimulation in chronic inflammation 5,34 . However, yeast-reactive T H 17 and T H 1 cell cytokines were mainly produced by different cell populations and only small populations of IL-17A + FN-γ + cells develop against non-albicans Candida and Saccharomyces spp. Altogether this suggests that T H 17 cells and T H -CTLs represent two independent populations driven most probably by separated groups of antigens. In future studies, it will be important to define specific fungal proteins and peptides that are recognized by T H 17 cells and cross-reactive T H -CTLs, which may allow the development of targeted, antigen-specific, therapeutic modulations. The cross-recognized antigens could comprise precisely the same or similar epitopes present in different yeast species or even unrelated sequences forming structurally similar epitopes, which moreover could also vary for different TCRs.
Overall, our data suggest a model in which, after initial T H 1 cell priming, repeated exposure to conserved antigens present in multiple fungal species leads to the selection of a highly cross-reactive TCR repertoire, accompanied by the acquisition of cytotoxic effector functions, probably as a result of chronic T cell activation. This is also supported by the observation that increased IFN-γ production and certain cytotoxic markers were elevated on yeast-responsive T cells in individual IBD-FDRs, but not yet the entire cytotoxic phenotype. Furthermore, the most expanded cross-reactive clonotypes from patients with CD showed the highest gene expression of cytotoxic markers, as well as the chronic activation marker gene HOPX 33,34 . As healthy individuals also harbor commensal intestinal yeasts and frequently ingest S. cerevisiae in the diet, the observed altered T cell reactivity in patients with CD and the incipient changes in individual IBD-FDRs may reflect a more intense or sustained interaction with yeast antigens, owing to epithelial barrier dysfunction. T cell-mediated cytotoxicity has been described as contributing to IEC death 47,48 and exacerbated T H 1 immunity in mice promoted epithelial barrier defects and mucosal fungal infection 49 . It is therefore tempting to speculate that yeast-responsive T H -CTLs may be directly involved in disease initiation and/or exacerbation through disruption of the intestinal barrier. In this regard, cross-reactive, yeast-responsive TCRs could be prime targets for the development of specific therapeutic interventions by antigen-specific regulatory T cell (T reg cell) therapy or T cell-depleting strategies.
Yeast-reactive T cells were selectively altered in patients with CD who are ASCA + , but not in those who are ASCA − or in patients with UC. It is thus possible that yeast-reactive T H -CTLs or their T H 1 progenitors may be involved in providing B cell help for ASCA formation. The presence of ASCAs in CD was described more than 30 years ago 50,51 and ASCAs are used as a biomarker for disease severity 30 , as well as a prognostic marker in individuals at risk for developing IBD 40 . Despite this strong link, their direct pathogenic relevance has been questioned and the yeast-responsive T cells have so far not been analyzed in detail in patients with CD. S. cerevisiae, also known as bakers' yeast, is ingested daily via nutritional intake and therefore generally considered to have a low virulence in humans. Instead, C. albicans, due to its higher immunogenicity, has been speculated to be the inducing species, as a result of binding of ASCAs to C. albicans mannans 24 . Given the data presented in the present study, it will be important to further define the fungal species and mechanisms involved in ASCA formation and in particular the contribution of T H 1-CTLs.
Our data may also result in concepts for selecting patients with CD for particular therapies. For example, therapeutic inhibition of Janus kinases targets, among other cytokines, the release of IFN-γ and might show more effectivity in patients with CD who are ASCA + . Further studies will examine the role of nutritional interventions with selective removal of yeast antigens, which-if effective-would be an attractive treatment strategy supporting the efficacy of other anti-inflammatory therapies or remission maintenance. Antifungal therapy, as well as a yeast-free diet, will be analyzed in future clinical trials in patients with CD who are ASCA + .
Our study also has some limitations. Although we have provided strong correlative evidence for the involvement of yeast-reactive CD4 + T cells in CD, further investigations are necessary to determine their direct contribution to disease pathophysiology. Our initial attempts in this direction provide some encouraging results and evidence for the ability of these cells to directly kill intestinal epithelium in vitro. However, further studies need to clarify whether therapeutic targeting of intestinal yeasts or yeast-responsive T H 1-CTLs is clinically effective. Although these highly cross-reactive T cells in patients with CD react to multiple yeast species, future studies are needed to determine whether a single or multiple species serve as the initial trigger for the altered Article https://doi.org/10.1038/s41591-023-02556-5 yeast-reactive T cell response. In the present study, we analyzed only a limited number of bacterial and fungal species and potential differences induced by strain variation were not addressed. Furthermore, the proteins or peptides driving the cross-reactive, yeast-responsive population remain to be determined. The discovery that a specific subgroup of patients with CD develops altered CD4 + T cell responses and ASCAs against yeasts, whereas other patients with IBD do not, underscores the presence of fundamentally distinct host-microbe interactions within this subgroup of patients. This emphasizes the need for in-depth investigations to shed light on the underlying mechanisms and to discern the unique characteristics of this subgroup of patients with CD.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41591-023-02556-5.

Intestinal biopsies
Biopsies from inflamed and noninflamed intestinal tissue were collected from patients with CD who underwent colonoscopy (ethics committee CAU Kiel, under the file: B 231/89-1/13). All patients gave written informed consent before participation.

Small IECs
Primary small IECs were purchased from PELOBiotech. Cells were cultivated in Epithelial Cell Medium (Cell Biologics) in T25 flasks pretreated with gelatin-based coating solution (Cell Biologics). During expansion, medium was replenished and cells were split as needed.

Quantification of ASCAs
Serum ASCAs were measured by ELISA according to the manufacturer's instructions (Euroimmun). In short, ASCA IgA/IgG ELISA kits include three calibration sera, a positive and a negative control. Sera were diluted 1:101 in probe buffer and incubated at room temperature on precoated microtiter plates for 30 min. After washing, a peroxidase-conjugated, anti-human IgA or IgG was added for 30 min and the plates were washed again. Substrate solution was added and the color reaction was stopped after 15 min. The optical density was determined at 450 nm, with a reference wavelength of 620 nm. ASCA status was calculated after developing a standard curve using the three calibration sera. ASCA quantification was done in duplicate and mean values were used. Sera were considered positive if their activity was >20 relative units ml −1 . Both single-positive IgA or IgG samples and double-positive samples were classified as ASCA + donors.

ITS-sequencing
ITS (internal transcribed spacer) amplicons were generated with 35 cycles using Invitrogen AccuPrime PCR reagents. These amplicons were used in a second PCR reaction, applying Illumina Nextera XT v.2 (Illumina) barcoded primers to uniquely index each sample as previously described 20 . All libraries were processed for quality control using the DNA 1000 Bioanalyzer (Agilent) and Qubit (Life Technologies) to validate and quantify library construction before preparing a paired-end flow cell. The samples were randomly divided among flow cells to minimize sequencing bias. Clonal bridge amplification (Illumina) was performed using a cBot (Illumina); 2× 250-base pair (bp) sequencing-by-synthesis was performed using the Illumina MiSeq platform. Article https://doi.org/10.1038/s41591-023-02556-5 pools were resuspended according to manufacturers' instructions and cells were stimulated at a concentration of 0.5 µg per peptide per ml. Cows' milk protein extract was purchased from Greer Laboratories and used at a concentration of 100 µg ml −1 .

Antigen-reactive T cell enrichment
PBMCs were freshly isolated from EDTA blood on the day of blood donation by density gradient centrifugation (Biocoll, Biochrom). ARTE was performed as previously described 15,28,[52][53][54] , with slight modifications. In brief, 1-2 × 10 7 PBMCs were plated in RPMI-1640 medium (Gibco), supplemented with 5% (v/v) human AB-serum (Sigma-Aldrich) in 12-well cell-culture plates and stimulated for 7 h with microbial lysates in the presence of 1 µg ml −1 of CD40 and 1 µg ml −1 of CD28 pure antibody (both Miltenyi Biotec). Then, 1 µg ml −1 of Brefeldin A (Sigma-Aldrich) was added for the last 2 h. In some experiments 100 µg ml −1 of anti-HLA-DR pure antibody (Miltenyi Biotec; clone AC122 or BioXcell; clone LT43) was added during stimulation. To multiplex several specificities, the differentially stimulated cells were labeled with different concentrations of two CD4 antibody clones (CD4-BV421, BioLegend, clone OKT4, titer 1:20 and 1:200; CD4-APC-Vio770, Miltenyi Biotec, clone MT-466, titers 1:50 and 1:500). For lower concentrations, the respective unconjugated CD4 pure antibody was added at a concentration of 1 µg ml −1 to block intermixing of the barcode label. Barcoded populations were pooled and labeled with CD154-biotin followed by anti-biotin MicroBeads (CD154 MicroBead Kit, Miltenyi Biotec) and magnetically enriched by two sequential MS Columns (Miltenyi Biotec). Surface staining was performed on the first column, followed by fixation and intracellular staining on the second column. Frequencies of antigen-specific T cells were determined based on the cell count of CD154 + T cells after enrichment, normalized to the total number of CD4 + T cells applied on the column. For each stimulation, CD154 + background cells enriched from the nonstimulated control were subtracted.

Cytotoxicity assay
Cytotoxicity of yeast-reactive CD4 + T cells against adherent small IECs (PELOBiotech) was measured in duplicate by a Real Time Cell Analyzer (RTCA, X-Celligence, ACEA) 55 . By using RTCA, the impedance of the cells is monitored via electronic sensors located on the bottom of a 96-well micro-E-plate. Impedance of the cells reflects changes in cellular parameters such as cell proliferation, morphological changes (for example, spreading, adherence) and cell death, and is expressed as an arbitrary unit called the cell index (CI).
A total of 7,500 adherent IECs cells per well were added to a 96-well micro-E-plate to monitor the impedance of the cells every 3 min for 60 h. As the initial adherence in different wells can differ slightly, the CI was normalized to 1 after having reached the linear growth phase. After 24 h, freshly purified S. cerevisiae-and M. restricta-reactive CD154 + CD69 + T mem cells were added at different effector-to-target Nature Medicine Article https://doi.org/10.1038/s41591-023-02556-5 ratios (1:1, 1:2, 1:4). T cells and IECs were cross-linked by the addition of 1 µg ml −1 of SEB. T cell-induced lysis of IECs is indicated by the loss of impedance, that is, a decrease in the normalized CI. As a positive control, IECs were treated with 1% Triton X-100 (final concentration). For analysis with RTCA software (v.2.0.0.1301, ACEA), the mean of Triton X-100 samples was calculated and defined as 100% lysis after the addition of effector cells. The percentage of lysis for each sample was calculated in relation to a control sample without effector cells.
Orthotopic TCR replacement in primary human CD4 + T cells DNA templates (Supplementary Table 1) were designed in silico based on the previously published method 42 and were synthesized by Twist Biosciences in pTwist Amp vectors. Briefly, the constructs comprise the full length of the α-and β-chains of the inserting TCRs flanked by left and right homology arms and contain self-cleaving peptides (P2A and T2A) to provide separation, as well as a poly(A) tail (bGHpA). The β-chain consists of the human variable region and the murine constant region, which is used as a tracking marker.
TCR replacement was performed as previously described 42 . In brief, CD4 + T cells were isolated by negative magnetic selection (CD4 + T Cell Isolation Kit; Miltenyi Biotec). Then, 4 × 10 7 CD4 + cells were plated on 6-well cell-culture plates in expansion medium, containing TexMACS medium, supplemented with 5% (v/v) human AB-serum, 200 U ml −1 of IL-2 and 100 IU ml −1 of penicillin, 100 µg ml −1 of streptomycin and 0.25 µg ml −1 of amphotericin B. CD4 + T cells were subsequently activated for 2 d using the T Cell Activation/Expansion Kit (Miltenyi Biotec). Activated CD4 + T cells were harvested and the activation beads were removed using a MACSiMAG Separator (Miltenyi Biotec). For each nucleofection, 1 × 10 6 cells were used and mixed with 20 µM ribonucleoprotein (RNP) mixtures (Integrated DNA Technologies (IDT)) and a homology-directed repair template (1 µg) (Twist Biosciences). The RNPs (all components from IDT) were generated in two sequential steps: first, the guide RNAs (gRNAs) were prepared by mixing equal volumes of trans-activating CRISPR (tracr)RNA (80 µM stock) with hTRAC CRISPR (cr)RNA (80 µM stock) (crRNA sequences for gRNAs were 5′-GGAGAATGACGAGTGGACCC-3′ for TRBC32, targeting both TRBC1 and TRBC2, and 5′-AGAGTCTCTCAGCTGGTACA-3′ for TRAC) and heating at 95 °C for 5 min, followed by cooling to room temperature (RT). In the second step, gRNAs were assembled with Cas9 nuclease by mixing equal volumes of Cas9 nuclease (6 µM) and gRNA (40 µM) and incubated at RT for 15 min. An electroporation enhancer was added (400 µM stock) to the RNP mixtures. The P3 Primary Cell 4D-Nucleofector X Kit S (Lonza) and a 4D-Nucleofector X unit (Lonza) with the EH100 program were used. Electroporated cells were immediately transferred to 96-well cell-culture plates in expansion medium and were incubated for 5 d at 37 °C and 5% CO 2 . After 5 d, TCR-transgenic CD4 + T cells were sorted with FACS on a FACS Aria Fusion (BD Bioscience) based on their mouse TCR-β expression and were further expanded for 2-3 weeks in the presence of 2 × 10 5 irradiated allogeneic feeder cells until subsequent functional assays were performed. For FACS, cells were stained with fluorochrome-conjugated antibodies to CD4-Vioblue, CD8-PerCP, CD14-PerCP, CD20-PerCP (all Miltenyi Biotec) and mouse TCR-β chain-APC, as well as human TCR-α/β chain-FITC (both BioLegend). After sorting, knocked-in cells were expanded for 2 weeks.

ScRNA-seq assay (10× Genomics)
For scRNA-seq and TCR-seq, yeast-reactive IL-17A producers, IFN-γ producers and double-negative CD154 + T mem cells were sorted with FACS and labeled with barcoded antibodies (CITE-seq). In brief, PBMCs were stimulated with C. albicans, C. tropicalis and S. cerevisiae for 6 h in the presence of 1 µg ml −1 of CD40 and 1 µg ml −1 of CD28 pure antibody. Cells were labeled with IFN-γ and IL-17A catch reagent (cytokine secretion assay, both Miltenyi Biotec) and incubated for an additional 45 min in RMPI-1640 medium supplemented with 2% (v/v) human AB-serum under continuous rotation. Cells were further labeled with CD154-PE, IL-17A-APC and IFN-γ-FITC, followed by anti-PE MicroBeads, and magnetically enriched by MS Columns (all from Miltenyi Biotec).
Cells were sorted into precoated, low-bind collection tubes, containing sterile, filtered 1× PBS supplemented with 0.5% (v/v) human AB-serum. After sorting, cells were re-added to MS Columns to avoid cell loss during washing steps and stained on the column with 1 µl of hashtag antibodies (TotalSeq-C0251 anti-human hashtag 1 (LNH-94, 1:50), TotalSeq-C0252 anti-human hashtag 2 (LNH-94, 1:50) and TotalSeq-C0253 anti-human hashtag 3 (LNH-94, 1:50) in the presence of Human TruStain FcX in titer 1:20) (all BioLegend) in 50 µl of staining volume. Columns were washed twice and cells were eluted with 500 µl of sterile filtered 1× PBS supplemented with 0.5% (v/v) human AB-serum. Directly before loading, IL-17A + , IFN-γ and CD154 + populations for each antigen were combined, centrifuged, and loaded on a Chromium Next GEM Chip K (10× Genomics) according to the manufacturer's instructions for processing with the Chromium Next GEM Single Cell 5′ Kit v.2. Between 5,000 and 12,000 cells were loaded for each reaction. TCR single-cell libraries were subsequently prepared from the same cells with the Chromium Single Cell V(D)J Enrichment Kit, Human T Cell (10× Genomics). Libraries were sequenced on an Illumina NovaSeq 6000 machine with 2× 100 bp for gene expression and TCR libraries, aiming for 50,000 reads per cell and 5,000 reads per cell, respectively.

Single-cell transcriptome data analysis
The 10× sequencing data were processed using the 10× Genomics Cell Ranger v.6.0.0 software pipeline as previously described 56 . Cellranger mkfastq was used to demultiplex cellular barcodes and produce fastq files and cellranger count was used to map reads to the human genome (GRch38) with STAR aligner 57 and obtain a matrix of transcript counts by cells.
The data analysis was performed with the scanpy Python package v.1.9.0, using Python v.3.7 (ref. 58). As part of the quality control, cells were subjected to the following criteria: (1) >500 expressed genes, (2) >1,000 total unique molecular identifier counts, (3) no doublets based on hashtag antibodies and (4) <5% of mitochondrial RNA. Out of the initial 63,732 cells, 60,130 satisfied these criteria and were used in the subsequent analysis.
The data normalization step included library size normalization to 10,000 counts per cell and log 2 (transformation) of the expression data. Highly variable genes were identified and used to reduce the dimensionality of the data to 30 principal components (PCs) with the principal component analysis algorithm. The PC values were adjusted using the harmony algorithm 59 as implemented in scanpy.external. pp.harmony_integrate function to correct the batch effect introduced by differences in sample preparation. Unsupervised clustering of the cells was performed with the Leiden clustering algorithm (scanpy. tl.leiden) and cell subtypes were assigned to the clusters based on the differentially expressed marker genes.

ScTCR repertoire data analysis
Cellranger vdj pipeline was used for V(D)J genes calling from scTCR-seq data. Scirpy Python package v.0.12.0 was used 60 to integrate TCR data with the transcriptomic data, and TCR characteristics were inferred for each cell and defined the clonotypes. Cells were considered to belong to the same clonotype if they had the same TCR-β amino-acid sequence. As a quantitative measure of the clonality of TCR repertoires we calculated Gini's coefficient 61 . Gini's coefficient ranges from 0 to 1, with 0 corresponding to equal abundance of all clonotypes and values close to 1 representing total inequality (that is, high prevalence of one sequence). For Gini's coefficient analysis, samples were normalized by random selection of 500 cells from each individual per fungal antigen to remove the impact of different sample sizes (number of cells per sample). TCR-sharing networks were created and visualized using the Cytoscape software v.3.9.1 (ref. 62). Clonotypes found in one cell only were excluded for network analyses.

Trajectory inference
Pseudotime trajectory analysis was performed using the Monocle v.2 algorithm 35 . Samples were normalized by random selection of 500 cells from each individual per fungal antigen to remove the impact of different sample sizes (number of cells per sample) and the analysis restricted to the gene set of 36 markers (Extended Data Table 3).

Statistical analysis
Statistical parameters including the exact value of n, the definition of center, dispersion and precision measure, and statistical significance are reported in the figures and figure legends. Statistical tests were performed with GraphPad Prism software v.9.4.1. Statistical tests were selected based on appropriate assumptions with respect to data distribution and variance characteristics; P < 0.05 was considered statistically significant.

Software
Flow-cytometry data were analyzed using FlowJo v. 10

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
Raw, paired-end, scRNA-seq fastq files from both gene expression and TCR libraries, as well as processed gene expression and TCR data as feature-barcode matrix files and clonotype tables, respectively, have been deposited at the Gene Expression Omnibus (accession no GSE227638) and GitHub under https://github.com/agbacher/fungi. All other data are available in pseudonymized form within 4 weeks upon request to the corresponding author.