HIF-1α⁺ CD4⁺ T cells coordinate a tissue resident immune cell network in the lung

Summary:

A deeper understanding of how tissue localized immune cells arise and function is critical for developing mucosal vaccines. Currently, there are no murine models that specifically target tissue T cells while leaving their lymphoid counterparts untouched. Here we leverage the observation that during influenza infection, HIF-1α regulatory activity is higher in the lung compared to lymph node CD4⁺ T cells. Inducible deletion of Hif1a in CD4⁺ T cells, at the onset of its activity in the lung, reduces the tissue resident T cell compartment with minimal impact on peripheral immunity. HIF-1α-active CD4⁺ T cells occupy the border of tertiary lymphoid structures, where they coordinate an IL-21-dependent network of spatially co-localized immune cells including macrophages, NK cells and IgA⁺ B cells. A similar HIF-1α-dependent network is engaged in a lung adenocarcinoma model, highlighting a broader role for HIF-1α⁺ CD4⁺ T cells in integrating protective immunity during infection and cancer.

INTRODUCTION

T cell responses are initiated in draining lymphoid organs, where T cells recognize cognate antigen presented by antigen presenting cells. Following clonal expansion, T cells migrate to non-lymphoid tissues, where they differentiate into a heterogeneous array of effectors. This directional sequence of events, in which adaptive immunity originates in the lymph node and is executed in the tissue, is referred to as an “inside-out” immune response¹. Ongoing inflammation in the tissue can trigger the formation of tertiary lymphoid structures (TLS) composed of T cells, B cells, dendritic cells, macrophages and stromal cells². TLS are elicited in response to infection, autoimmunity and cancer and may play host protective or pathologic roles depending on the immunologic context. TLS have a well known role in supporting mucosal antibody production and have also been identified as sites for de novo T cell recruitment and priming, capable of independently organizing an “inside-out” immune response within the tissue³. However, the signals regulating T cell diversification within the TLS microenvironment remain poorly understood.

T cell diversification is governed by a variety of inputs including antigen receptor engagement, cytokines, hypoxia and nutrient availability, all of which are dynamically regulated in time and space. The distribution of these signals in and outside of the TLS likely plays a central role in directing tissue T cell fate. For example, influenza infection elicits the development of heterogeneous lung resident CD4⁺ cells including TLS-localized T resident helper (TRH) cells, which express the surface marker folate receptor 4 (FR4), and FR4⁻ Th1 cells which are distributed throughout the tissue^4–6. TRH cells, similar to T follicular helper (Tfh) cells in lymphoid organs, require intrinsic Bcl6 expression, support local IgG antibody responses and retain the capacity to generate diverse secondary effectors^4,7,8. TRH cells can be further divided into two distinct subsets regulated by the transcription factors Bcl6 and HIF-1α, respectively⁸. While the role of Bcl6 in T cells has been extensively documented, the role of HIF-1α is more controversial⁹.

Genetic disruption of HIF-1α in T cells leads to pleiotropic effects on Th17, Treg and Th1 cell differentiation, with outcomes that are strongly influenced by microenvironmental cues^10–15. In lymphoid organs, HIF-1α can either restrain or promote TFH cell development^16–19. In the tissue, HIF-1α stabilization promotes a residency gene signature and anti-tumor effector function in CD8⁺ T cells but paradoxically impairs the accumulation of lung resident CD4⁺ T cells, leading to enhanced disease susceptibility in a mouse model of tuberculosis infection^20–23. One possible explanation for these discrepancies is that the genetic models used in prior studies ablated or stabilized HIF-1α prior to T cell priming. This prevents an assessment of whether HIF-1α has distinct effects at later time points or within specific tissue niches. HIF-1α can be induced by exposure to antigen, TGFβ and hypoxia, all of which are implicated in the formation of both tissue resident T cells and TLS^21,24–28. However, the spatial relationship between HIF-1α–activating signals and TLS, as well as their influence on the diversity and function of tissue-resident CD4⁺ T cells, remain unclear.

To address this knowledge gap, we used spatial transcriptomics to examine the lymphocytes localized within and outside TLS during influenza infection. Strong T cell - B cell interaction and Tfh gene signatures were concentrated in the TLS core, while gene signatures for T cell receptor signaling and hypoxia extended into the TLS periphery, where a subset of HIF-1α-active CD4 T cells was also detected. Inducible deletion of Hif1a, allowing temporal control of gene ablation after priming, revealed its central role in stabilizing the lung-resident CD4⁺ T cell compartment with negligible effect on peripheral T cell responses. HIF-1α⁺ CD4⁺ T cells are responsible for coordinating a network of diverse immune responses including replenishment of alveolar macrophages, accumulation of lung resident NK cells and maintenance of influenza specific IgA antibody titers in the respiratory tract. These seemingly disparate responses converge upon a requirement for IL-21 provided by HIF-1α⁺ CD4⁺ T cells.

RESULTS

Spatially structured immune cell interactions in lung TLS

To investigate the cellular architecture of TLS in an infection context, we performed spatial transcriptomic profiling (Xenium, 10x Genomics) on murine lung tissue 14 days after influenza infection. TLS structures were clearly identifiable as high density regions of the B cell-specific Ig-beta gene Cd79b that were consistent with H&E staining (Figure 1A). Accordingly, annotated B cells populated the TLS core while peripheral regions were enriched for transcripts from dendritic cells, macrophages, T cells and plasma cells, forming a layer around the B cell-dominated core (Figure 1B and C). To link immune cell positioning to known lymphoid organization cues, we assessed chemokine and receptor expression as a function of distance from the edge of TLS (Figure 1D and S1A). The TLS core areas were enriched for transcripts associated with B cell follicles in SLO including Cxcl13 and its receptor Cxcr5, as well as Cxcr4 and Ccr6 which mark B cell memory precursors^29,30. Increased expression of Ccr7, Ccl19, Cxcr3, Cxcl9 and Cxcl10 at the boundary of the B cell core suggests this region may be a site for the recruitment and retention of CD4⁺ T cells and resident memory B cells^31–33. Chemokine transcripts found primarily outside of the B cell core included Ccr2, Ccl21a/b, and Cxcr6 with its ligand Cxcl16^34,35. Spatially distinct expression of PD-1 ligands was also observed: Cd274 (encoding PD-L1) was more prominent in macrophages and B cells outside TLS, while B cells inside TLS expressed more Pdcd1lg2 (encoding PD-L2) (Figure 1E, S1A and S1B). This parallels findings in lymph nodes where PD-L1 from bystander B cells restricts Tfh cell access to follicles unless overridden by ICOS signaling³⁶. The peak expression of receptors and their ligands was not always colocalized, suggesting dynamic migration processes or the predominance of other signals (ex. Cxcr4 - Cxcl12 and Ccr7 - Ccl21a/b).

Figure 1. Spatially structured immune cell interactions in lung TLS.

(A-G) Xenium spatial transcriptomic data, mouse lung day 14 post influenza, one section (n = 1 mouse).

(A) H&E (top) and Xenium spatial transcriptomic transcriptome analysis (bottom) of mouse lung at 14 days post influenza. Bottom colored by Cd79b high TLS structure (dark) with surrounding periphery (faded).

(B) Proportions of annotated cell types within each tissue location.

(C) Proportions of tissue location within selected annotated cell types.

(D-F) Centered scaled expression of genes or calculated MSigDB gene signature scores averaged in 8 μm bins by distance from the TLS borders (n = 1 mouse). Rows are sorted using the maximum of a rolling average across 5 bins with a 1 bin step. The line at zero marks the TLS edge; negative values are inside TLS. Maximum and minimum values are limited to −2 or 2 respectively.

(D) Chemokine-related genes on all cells.

(E) Selected genes and gene set scores on B cells.

(F) Selected genes and gene set scores on T cells.

(G) Gene sets enrichment on all cells contrasting TLS core with TLS periphery.

Statistical analyses: gene set enrichment analysis using limma camera, FDR < 0.05.

Building on these observations, we examined B and T cell gene expression patterns in the core and peripheral regions of the TLS. Core B cells resembled germinal center (GC) B cells expressing genes such as Cd40, Il21r, Bcl6, and Cxcr5 (Figure 1E). B cells deepest inside the largest TLS expressed light zone genes (Ciita, Pdcd1lg2, S1pr1, Cd86, and Myc), while those near the edge expressed dark zone associated genes (Aicda, Mki67, Cxcr4, and Il4ra), with Mki67 expression further corroborated by microscopy for Ki67, suggesting an anatomically inverted organization compared to GCs in SLO (Figure 1E and S1C)^30,31. Outside the TLS core, B cells were enriched for memory-associated (Cxcr3, Ccr6, Pdcd1, Nr4a3) and plasma cell-associated (Prdm1, Cxcr4, Jchain, Igha) genes, with immunofluorescence staining confirming the location of IgA⁺ cells^37,38 (Figure 1E, S1D and S1E). T cells in the TLS core exhibited a Tfh phenotype (Bcl6, Sh2d1a, Icos, Tcf7, Pdcd1, and Il21) while T cells outside the core showed higher expression of Cxcr3, Cxcr6, Tbx21, Selplg, Id2 and Il7r, consistent with type 1 immunity^4,39,40 (Figure 1F and S1A). Gene set enrichment analysis corroborated these findings with elevated signatures for GC B cells (BCR signaling, cell cycle and Myc) and GC Tfh cells at the TLS core. Similar to what has been described at the T cell-B cell border in SLO, the outer TLS was also enriched for a non-GC Tfh signature, suggesting a functional division between Tfh cells located in the core compared to those in the periphery. Signatures for hypoxia and TGFβ signaling, both known to support HIF-1α in T cells, were also present at the TLS periphery, where Hypoxyprobe staining was also highest (Figure 1F, 1G and S1F)^{18,19,21,41–43}. This pattern aligns with our previous finding that two TRH subsets - one enriched for Bcl6 activity and the other for HIF-1α activity — emerge in the lung during influenza⁴. Increased hypoxia and TGFβ associated signaling at the TLS periphery suggest it may represent a distinct microenvironment that drives HIF-1α activity to uniquely shape T cell differentiation.

Lung CD4⁺ T cells increase HIF-1α activity during influenza infection

To understand the dynamics of Hif1a expression during influenza infection, we re-analyzed single cell RNA sequencing (scRNA-seq) data of influenza (NP) specific CD4⁺ T cells isolated from the lung and draining lymph nodes at days 9, 14 and 30⁴. Transcriptional regulatory activity of HIF-1α was higher in lung CD4⁺ T cells compared to those in the lymph node, increasing over time and peaking at day 30 (Figure 2A). To study this in more detail, we infected Egln3-YFP mice, which report the transcriptional activation of HIF-1α⁴¹.

Figure 2. Lung CD4⁺ T cells increase HIF-1α activity during influenza infection.

(A) Centered, scaled SCENIC AUC of scRNA-seq data from NP⁺ CD4⁺ T cells in the lung and lymph node (LN) (n = 3 mice).

(B) Flow cytometry analysis showing frequency of lung NP-specific Egln3-YFP⁺ FR4⁺ (TRH) CD4⁺ T cells from Egln3-YFP mice at 9, 14, and 30 days post-PR8 infection (mean ± SD, n = 3 mice).

(C) Dynamics of Egln3-YFP⁺ NP⁺ CD4⁺ T cells across lung resident (ivCD45⁻), circulating, LN, and blood compartments (mean ± SD, n = 4 mice).

(D) Confocal microscopy of TLS showing Egln3-YFP⁺ and Egln3-YFP⁻ CD4⁺ T cells and distribution of distances (μm) of Egln3-YFP⁺ and Egln3-YFP⁻ CD4⁺ T cells from TLS borders in 50 TLS across mice using confocal images. Analysis 30 days after infection. Bars, 100μm (mean ± SD, n = 3 mice).

(E) Frequency and numbers of NP-specific Egln3-YFP⁺ TRH cells in Egln3-YFP mice at day 20 post-PR8 infection, treated (intraperitoneal) i.p. with isotype control (ctrl) or LTβR-Fc on days 9 and 14 (mean ± SD, n = 5 mice).

(F) Numbers of ivCD45⁻ NP-specific CD4⁺ T cells in Egln3-YFP mice at day 20 post-PR8 infection, treated i.p. with ctrl or LTβR-Fc on days 9 and 14 (mean ± SD, n = 5 mice).

Statistical significance determined by ANOVA (B) and t-tests (E). Adjusted p-values: *p<0.05, **p<0.005, ***p<0.0005, ****p<0.0001, ns: not significant. Data representative of two independent experiments.

Consistent with the sequencing data, Egln3-YFP was progressively upregulated in both lung and lymph node CD4⁺ T cells with generally higher expression in lung over lymph nodes (Figures 2B, S2A and S2B). In the lung, Egln3-YFP expression was restricted to FR4 positive CD4⁺ T cells, marking the TRH compartment (Figure 2B). Sorted Egln3-YFP⁺ TRH cells had enhanced gene expression of Hif1a compared to both Egln3-YFP⁻ TRH cells Th1 cells, validating the use of YFP as a readout for HIF-1α activity (Figure S2C). Although the frequency of lung Egln3-YFP⁺ CD4⁺ T cells increased until day 30, it sharply decreased at later time points, indicating a transient HIF-1α-active subset with Tfh-like features (Figure 2C). In the draining lymph node, Egln3-YFP was primarily expressed by CXCR5⁺ Tfh cells (Figure S2A) and, in contrast to its transient expression in the lung, was maintained until at least 100 days after infection (Figure 2C). No appreciable Egln3-YFP was detected in CD4⁺ T cells from the blood or in lung “circulating” T cells that stained positively for intravenous (i.v.) anti-CD45, suggesting that lung CD4⁺ T cells independently upregulate HIF-1α activity in response to local cues (Figure 2C).

Next, we used histology to assess where Egln3-YFP⁺ CD4⁺ T cells were localized in the lung at days 14 and 30 after infection. Unlike Bcl6-high CD4⁺ T cells, which are centrally contained within TLS⁴, Egln3-YFP⁺ CD4⁺ T cells were more abundant at the TLS perimeter at both days 14 and 30 (Figure 2D and S2D). This distribution contrasts with the spatial transcriptomics data at day 14 where Hif1a transcripts were enriched within the TLS core (Figure 1F). To reconcile these findings, we assessed HIF-1α protein expression at days 14 and 30. On day 14, HIF-1α⁺ CD4⁺ T cells were broadly distributed through the TLS, consistent with transcriptomic enrichment in the core. By day 30, however, HIF-1α⁺ CD4⁺ T cells shifted toward the TLS periphery (Figure S2E), matching the distribution of Egln3-YFP CD4⁺ T cells (Figure 2D). One explanation for the discrepancy between HIF-1α protein and Egln3-YFP reporter expression is antagonism by Bcl6. Prior work shows that Bcl6 represses some aspects of HIF-1α dependent gene expression, including Egln3⁴⁴. To test this, we inducibly deleted Bcl6 in CD4⁺ T cells beginning at day 14 after infection, a time point when TRH cells are already present in the lung^4,5. Loss of Bcl6 led to an expansion of TRH cells and increased HIF-1α expression (Figure S2F–S2H). Together with our previous observation that Bcl6 deletion leads to the accumulation of CD4⁺ T cells outside the TLS core⁴, these results suggest a dynamic interplay between Bcl6 and HIF-1α that shapes CD4⁺ T cell localization in and around the TLS. To determine whether Egln3-YFP⁺ CD4⁺ T cells in the lung depend on TLS, we treated influenza infected mice with LTβR-Fc to prevent TLS formation⁴². LTβR-Fc treatment led to a clear reduction of Egln3-YFP⁺ CD4⁺ T cells in the lung while total numbers of influenza specific T cells were less affected (Figures 2E and F, S2I and S2J). Together these data reveal a transient subset of HIF-1α-active TRH cells whose generation and/or maintenance depend on local TLS.

HIF-1α⁺ CD4⁺ T cells produce IL-21 and support T cell residency in the lung

To specifically address the role of HIF-1α in lung CD4⁺ T cells, we generated Hif1a^flox/floxCD4^Cre-ERT2 (Hif1a^iKO) mice, enabling tamoxifen-inducible Hif1a deletion in CD4⁺ T cells. Heterozygous control (Hif1a^flox/+CD4^Cre-ERT2, Hif1a^ctrl) and Hif1a^iKO mice were infected with influenza and treated with tamoxifen starting at day 15 after infection (Figure 3A), allowing us to preserve HIF-1α activity during initial priming of CD4⁺ T cells in lymph nodes. Although TLS size and gross morphology were not affected after five days of tamoxifen treatment (Figure S3A), the total number of influenza specific lung resident CD4⁺ T cells was decreased by approximately 10-fold in Hif1a^iKO mice, while neither circulating CD4⁺ T cells in the lung nor Tfh cells in the draining lymph node were affected (Figures 3B, S3B and S3C). Th1 cells were significantly decreased in Hif1a^iKO mice while the absolute number of TRH cells was less affected (Figure S3D). Consistent with limited effects on TRH cells, Hif1a deletion had no impact on influenza-specific IgG titers in the bronchoalveolar lavage fluid (BALF) or serum (Figure S3E).

Figure 3. HIF-1α⁺ CD4⁺ T cells produce IL-21 and support T cell residency in the lung.

(A) Experimental design for (B-F).

(B) Flow cytometry analysis and quantification of resident (ivCD45⁻) NP-specific CD4⁺ T cells (mean ± SD, n = 5 mice).

(C) Analysis and quantification of resident NP-specific CD4⁺ T cell subsets in Hif1a^ctrl and Hif1a^iKO mice (mean ± SD, n = 5 mice).

(D) Il21 expression measured by TaqMan qPCR in sorted lung Th1 (Egln3-YFP⁻FR4⁻), Egln3⁻ TRH (Egln3-YFP⁻FR4⁺), and Egln3⁺ TRH (Egln3-YFP⁺ FR4⁺) cells (mean ± SD, each dot represents 2 pooled mice).

(E) Sorted NP-specific TRH cells from Hif1a^ctrl and Hif1a^iKO mice (mean ± SD, each dot represents 2 pooled mice).

(F) Quantification of NP-specific Th1, CXCR6⁺ and CXCR6⁻ TRH cells in Hif1a^ctrl and Hif1a^iKO mice following intratracheal IL-21 treatment (days 14–20) (mean ± SD, n = 4 mice).

Statistical significance determined by t-tests (A), one-way ANOVA (D, F) or two-way ANOVA (C); adjusted p-values: *p<0.05, **p<0.005, ***p<0.0005, ****p<0.0001, ns: not significant. Data representative of two independent experiments.

Similar to TFH cells in lymphoid organs, TRH cells maintain differentiation plasticity while Th1 cells are more terminally differentiated⁴⁵. We therefore considered the possibility that HIF-1α, most highly expressed by TRH cells, promotes their transition to Th1 cells. In support of this idea, visualization of surface markers FR4 and CXCR6, expressed on TRH and Th1 cells, respectively, revealed a subset of CXCR6⁺ TRH that was significantly reduced following Hif1a deletion (Figure 3C). Although HIF-1α activity was highest within the CXCR6⁻ TRH subset, its loss primarily affected CXCR6⁺ TRH and Th1 populations (Figures S3F). In control mice, CXCR6⁻ TRH cells had higher Ki67 expression, more BrdU incorporation and higher levels of Bcl-2 compared to CXCR6⁺ TRH and Th1 cells (Figure S3G and SH). These findings suggest that CXCR6⁻ TRH cells are intrinsically better equipped to proliferate and survive which might explain their numerical preservation when the total tissue resident compartment contracts in Hif1a^iKO mice.

We next addressed which HIF-1α derived signals might lead to induction of CXCR6 on lung TRH cells. At day 30 after influenza infection, approximately 50% of TRH cells actively transcribe IL-21. In previous work, Hif1a-deficient CD4⁺ T cells cultured under Th polarizing conditions produce less IL-21 than their WT counterparts^4,13. Similar to HIF-1α activity, IL-21 expression was detected in both CXCR6⁻ and CXCR6⁺ TRH cells (Figure S3I). To investigate a linkage between HIF-1α activity and IL-21 production, we assessed Il21 transcripts in sorted Egln3-YFP⁺ TRH cells and in remaining TRH cells isolated from Hif1a^iKO mice. Il21 was most highly expressed by Egln3-YFP⁺ TRH cells, but decreased following Hif1a deletion (Figures 3D and 3E). IL-21 protein was also reduced in the lung homogenate of Hif1a^iKO mice (Figure S3J). Hif1a deletion also led to decreased expression of PD-1 on lung NP⁺ CD4⁺ T cells, in agreement with one study showing that exposure of CD4⁺ T cells to IL-21 increases PD1 expression and another report showing that Il21-deficient TFH cells express lower levels of Pdcd1 transcript (Figure S3K)^43,46.

To explore pathways that might connect HIF-1α to IL-21 production, we next examined regulators linked to IL-21-producing T helper subsets in the lung: T-bet⁺ TFH cells and Blimp-1⁺ CD4⁺ T cells enriched for Th1 gene signatures^47,48. CXCR6⁺ TRH cells expressed intermediate levels of T-bet compared with CXCR6⁻ TRH and Th1 cells (Figure S3L). In addition, Hif1a^iKO CD4⁺ T cells cultured in vitro under IL-21 inducing conditions expressed less Blimp-1 and IL-21 relative to controls, with diminished IL-21 confined to the Blimp-1⁺ subset (Figure S3M). Finally, to understand if local production of IL-21 contributes to the differentiation of lung CXCR6⁺ T cells, Hif1a^iKO mice were treated with intratracheal IL-21 concurrent with tamoxifen gavage. Exogenous IL-21 was sufficient to restore both PD1 expression and numbers of CXCR6⁺ TRH and Th1 cells in Hif1a^iKO mice but had minimal impact on lung or LN CD4⁺ T cells from Hif1a^ctrl mice (Figures 3F and S3N–S3P). Together these data raise the possibility that HIF-1α dependent IL-21 secretion by lung TRH cells supports CXCR6 expression and seeding of tissue-resident T cells outside of TLS structures.

HIF-1α⁺ CD4⁺ T cells orchestrate lung immunity during influenza

The sole ligand for CXCR6 is the chemokine CXCL16, constitutively expressed by lung epithelial cells and macrophages^49,50. Our secondary analyses of lung macrophage scRNA-seq data revealed increased expression of both Cxcl16 and Il21r two weeks after influenza infection, coinciding with the onset of HIF-1α activity in the lung (Figure S4A and 2B)⁵¹. In the spatial data, Cxcl16 transcripts were abundantly expressed in the TLS periphery, where both Cxcl16⁺ and Il21r⁺ macrophages outnumbered their non-expressing counterparts (Figures S4B and S4C). Although Il21r transcripts primarily coincided with Cd79b expressed by B cells in the TLS core, they were also detected in this peripheral area and could be assigned to a variety of cell types including T cells, monocytes and macrophages (Figure S1B). While Izumor1r⁺Cxcr6⁺ T cells were more scarcely detected, they were concentrated in outer TLS, highlighting a potential role for these cells in supporting lung macrophage recovery after infection (Figure S4D).

Given the putative co-localization of HIF-1α-active CD4⁺ T cells and CXCL16⁺ myeloid cells, we assessed the impact of Hif1a deletion in CD4⁺ T cells on lung macrophages which are known to be transiently depleted and replenished during influenza infection^51–53. In Hif1a^iKO mice, alveolar macrophages (AMs) were diminished and had decreased expression of MHC-II, CD86 and LFA-1 compared to Hif1a^ctrl mice, deficiencies which could be restored by exogenous IL-21 (Figures 4A–C and S4E). To understand if IL-21 is intrinsically required by lung macrophages, we generated irradiation chimeras transplanted with a mixture of wildtype and IL-21R^−/− bone marrow (Figure S4F). At baseline, reconstitution of the BALF was biased toward IL-21R^−/− immune cells, including both AMs and lymphocytes, while the lung was slightly better reconstituted by wildtype lymphocytes (Figure S4F). However, 14 days after influenza infection, IL-21R^−/− AMs in the BALF and monocyte derived macrophages in the lung were at a clear competitive disadvantage compared to wildtype cells (Figure 4D, S4G and S4H). Consistent with our secondary analysis⁵¹, monocyte-derived macrophages also displayed the highest IL-21R expression in both lung digests and BALF at days 14 and 30 post-influenza infection (Figure S4I). Together, these data indicate that macrophage intrinsic IL-21 sensing is required for full replenishment of the lung macrophage compartment following infection. Together these data indicate that macrophage intrinsic IL-21 sensing is required for full recovery of the lung macrophage compartment following infection.

Figure 4. HIF-1α⁺ CD4⁺ T cells orchestrate the lung immune response during influenza.

(A) Representative flow cytometry plot and number of AMs from the lungs of Hif1a^ctrl and Hif1a^iKO mice (mean ± SD, n = 4 mice).

(B) Numbers of AMs of Hif1a^ctrl and Hif1a^iKO mice following intratracheal IL-21 treatment (days 14–30) (mean ± SD, n = 4 mice).

(C) MHCII, CD86, and LFA1 expression (gMFI) of lung AMs of Hif1a^ctrl and Hif1a^iKO mice following intratracheal IL-21 treatment (days 14–30) (mean ± SD, n = 4 mice).

(D) Ratio of AMs in BALF and Lung monocyte-derived macrophages (moMACs) of chimeric mice (WT/IL-21R^−/−) (mean ± SD, n = 3-4 mice).

(E) Number of lung circulating and resident NK cells in Hif1a^ctrl and Hif1a^iKO mice following PR8 infection (mean ± SD, n = 6 mice).

(F) Lung resident NK cell numbers in Hif1a^ctrl and Hif1a^iKO mice following intratracheal IL-21 administration (days 14–30) (mean ± SD, n = 4 mice).

(G) Numbers of lung CD38⁺CXCR3⁺CCR6⁺ and GL7⁺CD38⁻ B cells from Hif1a^ctrl and Hif1a^iKO mice at 30 days post-PR8 infection (mean ± SD, n = 6 mice).

(H) ELISA for PR8-specific IgA from BALF and serum (mean ± SD, n = 5-6 mice).

(I) Numbers of lung CD38⁺CXCR3⁺CCR6⁺ and GL7⁺CD38⁻ B cells in Hif1a^ctrl and Hif1a^iKO mice after intratracheal IL-21 treatment (days 14–30) (mean ± SD, n = 6 mice).

Statistical significance determined by t-tests (A, D, F, I), one-way ANOVA (B, C, E, H), two-way ANOVA (F, G, H) Adjusted p-values: *p<0.05, **p<0.005, ***p<0.0005, ****p<0.0001, ns: not significant. Data representative of two independent experiments.

We next considered if HIF-1α dependent IL-21 might contribute to the maintenance of additional IL-21-sensitive cells in the lung. Consistent with prior studies showing that IL-21 is required to sustain tissue resident CD8⁺ T cell responses^5,54, CD8⁺ T cell numbers in the lung were decreased in Hif1a^iKO compared to Hif1a^ctrl mice with no changes detected in the draining lymph node (Figure S4J). Pseudobulk analysis indicated that NK cells isolated from influenza infected Hif1a^iKO lung at day 30 had decreased expression of Tcf7, Gramd3, and Rora, all shown to be important for the establishment of tissue resident, memory-like cells. Meanwhile, NK cells from control mice were enriched for a signature for ILC1 cells, prominent early producers of IFNγ following viral infection^55–60 (Figure S4K). To assess the lung resident NK cell compartment by flow cytometry, mice were injected intravenously with anti-CD45 shortly before lung harvest. Although the majority of lung NK cells from both Hif1a^ctrl and Hif1a^iKO mice were in the vasculature (Figure S4L), only tissue resident NK cells were decreased in Hif1a^iKO mice (Figure 4E). In contrast, NK cell numbers in the draining lymph node, spleen and liver were comparable across the strains, indicating that the effects of Hif1a deletion are restricted to the lung (Figure S4M). A time course analysis in influenza infected wild type mice showed that resident NK cell numbers were elevated at day 9, stably maintained until at least 100 days after infection, and comprised of both conventional (Eomes⁺) and ILC1 (Eomes⁻) NK cells (Figure S4N). Tissue resident NK cells expressed CXCR6, suggesting possible colocalization with HIF-1α active T cells, and low levels of KLRG1, a marker for terminal NK cell differentiation (Figure S4O). Expression of these markers was unaffected in Hif1a^iKO mice, suggesting that Hif1a deletion affects their recruitment and/or maintenance as opposed to their function (Figure S4P). Administration of exogenous IL-21 to Hif1a^iKO mice was sufficient to restore the lung resident NK cell compartment (Figure 4F). These data reveal that influenza infection leads to the stable accumulation of lung resident NK cells which depend on IL-21 produced by HIF-1α⁺ CD4⁺ T cells.

IL-21 regulates B cell fate including GC, memory and plasma cell responses, and promotes IgA production in mucosal tissues. Compared to Hif1a^ctrl mice, Hif1a^iKO mice had lower numbers of GL7⁺CD38⁻ GC B cells and CCR6⁺CXCR3⁺ resident memory B cells including IgM⁺, isotype switched (swIg) and IgA⁺ subsets (Figure 4G and S4Q). B cell subsets in the draining LN were unaffected, indicating that inducible Hif1a deletion primarily affects mucosal rather than peripheral immune responses (Figure S4R).

Although spatial transcriptomics revealed Il21 enrichment in the TLS core at day 14 (Figure 1F), IL-21 protein localized mainly to the periphery at both days 14 and 30, likely reflecting rapid consumption by GC B cells in the core (Figure S4S). Using IL-21-responsive gene signatures as a proxy for cytokine activity, we found enrichment of IL-21 signaling in B and T cells in the TLS core at both time points (Figure S4T). Among non-lymphocytes, the IL-21 response extended into the periphery at day 14 and by day 30 was found even farther away from the TLS.

To link HIF-1α and IL-21 dynamics with B cell and humoral responses, we compared inducible Hif1a deletion initiated at different time points after infection. Deletion at day 15, when HIF-1α protein is core-enriched, reduced both GC B cell numbers and IgA titers; deletion at day 25, when HIF-1α shifts to the TLS periphery, primarily affected IgA titers (Figure S4U). Consistent with this, IgA was peripherally localized both by spatial transcriptomics (Igha) and immunofluorescence microscopy (Figure S1C and D). As observed for macrophages and NK cells, intratracheal IL-21 was sufficient to restore these B cell responses in Hif1a^iKO mice, highlighting its capacity to rectify broad deficiencies in the lung humoral compartment (Figure 4I).

HIF-1α⁺ CD4⁺ T cells coordinate mucosal immunity to infectious and tumor challenge

Our data indicate that Hif1a deletion in CD4⁺ T cells has wide ranging effects on the influenza induced immune cell network, including lung macrophages, NK cells, and B cells. The coordinated activity of these same cell types is seen following heterotypic influenza challenge^31,61. To probe this further, Hif1a^ctrl and Hif1a^iKO mice were infected with influenza-X31 followed by tamoxifen administration to delete Hif1a, and secondary infection with influenza PR8 (Figure 5A). Two days after the challenge, Hif1a^iKO mice had decreased oxygen saturation in the blood, a measure of pulmonary function, and increased infectious virus titers (Figures 5B and 5C). Exogenous IL-21 provided during the primary infection restored viral control following challenge, although recovery of lung function was modestly delayed compared to Hif1a^ctrl mice (Figure 5B and 5C). Both IFN-γ expression by lung NK cells and influenza specific IgA titers in the BALF were reduced, underscoring the importance of HIF-1α activity in orchestrating early protective immunity in the mucosa (Figures S5A and S5B). To understand if HIF-1α might coordinate a similar network of responses in a distinct context, we utilized the KPAR tumor model, a lung adenocarcinoma that induces adaptive immune responses that contribute to tumor control (Figure 5D)⁶². Transplantation of KPAR cells into Egln3-YFP reporter mice induced the accumulation of lung resident HIF-1α active CD4⁺ T cells, validating the use of this model (Figure S5C). Transplantation of KPAR cells into Hif1a^ctrl and Hif1a^iKO mice followed by tamoxifen treatment led to significantly decreased survival in Hif1a^iKO mice (Figure 5E). Decreased survival was accompanied by diminished numbers of FR4⁺CXCR6⁺ CD4⁺ T cells, ILC1s, IgA⁺ B cells and AMs in the lung (Figures 5F–I and S5D). Hif1a^iKO mice also harbored more MHC-II^lo myeloid cells which have been associated with compromised immunity in the tumor microenvironment (Figure 5J)⁶³. Survival in Hif1a^iKO mice, but not in control mice, could be prolonged by the exogenous administration of IL-21 (Figure 5E).

Figure 5. HIF-1α⁺ CD4⁺ T cells coordinate mucosal immunity to infectious challenge and tumor.

(A) Experimental design for the analyses shown in (B-C).

(B) Measurement of oxygen saturation (SpO₂) after challenge (mean ± SD, n = 5 mice).

(C) Influenza A (PR8) RT-qPCR for viral titers of Hif1a^ctrl and Hif1a^iKO mice after challenge (mean ± SD, n = 5 mice).

(D) Experimental layout for the data in (E–J).

(E) Survival curve of Hif1a^ctrl and Hif1a^iKO mice following intratracheal IL-21 treatment (days 14–30) (mean ± SD, n = 4 mice).

(F) Frequency and number of FR4⁺CXCR6⁺ CD4⁺ T cells (mean ± SD, n = 5 mice).

(G) Numbers of lung ILC1 cells (mean ± SD, n = 5 mice).

(H) Numbers of IgA⁺ resident B cells (mean ± SD, n = 5 mice).

(I) Representative plot and numbers of AMs (mean ± SD, n = 4 mice).

(J) Frequency of MHC-II negative of lung myeloid cells from Hif1a^ctrl and Hif1a^iKO mice (mean ± SD, n = 4 mice).

Statistical (mean ± SD) significance determined by t-test Aggregate comparison AUC (B), Mantel-Cox test (E), one-way ANOVA (C) and t-tests (F-J). Adjusted p-values: **p<0.005, ***p<0.0005, ns: not significant. Data representative of two independent experiments.

DISCUSSION

The tissue microenvironment is composed of a variety of resident immune cells working in concert to maintain tissue homeostasis. During infection, T cells in the tissue and TLS may re-encounter antigen, respond to inflammatory cytokines, engage with new cell types, and experience fluctuations in nutrient and oxygen availability. T cell plasticity - a cornerstone of the immune response - allows T cells to sense and adapt to these dynamic, tissue derived signals. This plasticity is facilitated by the flexible expression of transcription factors that fine tune T cell function. However, genetic models to conditionally delete transcription factors in tissue T cells are lacking. In this study, inducible deletion of Hif1a in CD4⁺ T cells at the height of its activity in the lung restricts the effect to the tissue, revealing HIF-1α’s role as a nexus of tissue T cell plasticity. Prior studies link HIF-1α to CD8⁺ Trm cell differentiation^21,22, but these models were either in vitro or relied on constitutive stabilization of HIF-1α in all T cells. By using a temporally controlled, CD4-specific deletion during influenza infection, our study demonstrates a unique role for HIF-1α in regulating tissue-resident CD4⁺ T cell responses. This contrasts with findings in an M. tuberculosis infection model where constitutive stabilization impairs CD4⁺ T cell accumulation in the lung, highlighting that HIF-1α can have distinct effects depending on T cell subset and tissue context⁶⁴.

During influenza, HIF-1α is most highly expressed by lung TRH cells, which share phenotypic and functional characteristics with lymphoid Tfh cells, and which exhibit the flexibility to differentiate into both TRH and Th1 cells following adoptive transfer and challenge^4,5. We demonstrate that HIF-1α⁺ CD4⁺ T cells contribute to TRH plasticity via their production of IL-21, which leads to CXCR6 upregulation and seeding of tissue resident cells at the TLS perimeter. An open question is whether IL-21 produced by lung TRH cells acts in an autocrine or paracrine manner to increase CXCR6 expression. IL-21 can be produced by lymphoid Tfh cells can act in trans, indicating that both scenarios are possible⁶⁵.

Future efforts should focus on elucidating the signals that drive Hif1a expression in lung CD4⁺ T cells. Although our data suggest that HIF-1α⁺ CD4⁺ T cells arise locally from lung TLS, TLS themselves persist long after HIF-1α activity has waned, indicating that while TLS are necessary, they are not sufficient for the maintenance of HIF-1α⁺ CD4⁺ T cells⁶⁶. It remains to be shown whether these cells reduce their expression of HIF-1α over time, depart from the lung or die. HIF-1α in T cells can be elicited by a variety of inputs including hypoxia, T cell receptor signaling and TGFβ, all of which were detected at the TLS border by spatial transcriptomics^{18,21,22,67–69}. It is possible that the final differentiation outcome of HIF-1α activity and IL-21 sensing depends on additional, spatially compartmentalized signals that stabilize one fate over another^70,71. This idea is consistent with prior work and our own data showing that Bcl6 deletion enhances HIF-1α expression and accumulation CD4⁺ T cells outside of TLS⁴⁵. Although Bcl6 has not been shown to directly repress Hif1a transcription, it suppresses glycolytic and hypoxia-responsive pathways that promote HIF-1α stability. The increased HIF-1α expression observed following Bcl6 deletion is consistent with a loss of this indirect, metabolic restraint.

A parallel can also be drawn with B cells, where IL-21 regulates both Bcl6 and its antagonist Blimp-1, balancing high affinity selection in germinal centers with plasma cell differentiation^72–74. For tissue T cells, IL-21 and cognate antigen presentation by B cells in the TLS core might reinforce TRH cell identity, while interactions with CXCL16 expressing antigen presenting cells outside of TLS may favor Blimp-1, required for tissue residency and a predicted regulator of Th1 cell identity^75,76,4. Supporting this model, IL-21 produced by Th1-Tfh cells in the lung supports IgA responses in the respiratory tract^47,48. Together with prior evidence that HIF-1α can enhance Prdm1 expression in T cells, this raises the possibility that Blimp-1 not only programs residency but also acts downstream of HIF-1α to couple IL-21 production with mucosal effector functions^22,68. Macrophages and epithelial cells - both sources of CXCL16 - may contribute to this spatially organized regulation of tissue resident T cells^49,77,78. It is also worth noting that in the gut mucosa, deletion of Hif1a in T cells has pleiotropic effects across Th17, Tfh and regulatory T cell subsets, all of which are capable of interconverting^{79,12,80,81,10}. It is tempting to speculate that HIF-1α plays a broader role in modulating T cell plasticity. Given that Bcl6 can antagonize certain aspects of HIF-1α activity, co-expression of these transcription factors may mark a poised state, enabling rapid engagement of a HIF-1α-driven transcriptional program upon decreased expression of Bcl6⁴⁴. An additional layer of control provided by HIF-1α may account for the functional redundancy among Th subsets when an individual lineage determining transcription factor is deleted^82–84.

Our results underscore a role for HIF-1α⁺ CD4⁺ T cells in establishing a mucosal immune cell network including replenishment of the alveolar macrophage compartment, accumulation of lung resident NK cells, and maintenance of humoral immunity in the respiratory tract. These diverse responses converge on a shared requirement for IL-21 provided by HIF-1α⁺ CD4⁺ T cells. A similar network is seen with influenza specific memory B cells, persisting in alveolar as opposed to TLS niches, requiring both alveolar macrophages and NK cells to migrate to infected foci after challenge^31,61. Our data suggest that HIF-1α⁺ CD4⁺ T cells may coordinate an analogous IL-21 dependent immune cell network within the tumor microenvironment. These data are aligned with the correlation between the intratumoral expansion of IL-21⁺ CD4⁺ T cells and successful outcomes to immune checkpoint blockade^85,86.

In summary, HIF-1α activity is central to CD4⁺ T cell plasticity in the tissue and facilitates IL-21 dependent control over a multifaceted immune response in the lung. TLS and surrounding microenvironmental cues emerge as key sites for the differentiation of HIF-1α⁺ CD4⁺ T cells, although downstream effector functions are likely shaped by additional signals encountered outside the TLS core. While these findings highlight a tissue-specific role for HIF-1α, they do not exclude its involvement in early peripheral lymphoid responses, and may help resolve some seemingly contradictory effects Hif1a deletion on Th cells.

LIMITATIONS OF THE STUDY

While our data suggest that CD4⁺ T cell intrinsic HIF-1α activity is the main source of IL-21 in the influenza-infected lung, we cannot exclude contributions from other cell types. Mixed bone marrow chimera models will be needed to formally rule out alternative IL-21 sources. Although our data indicate that macrophage intrinsic IL-21R signaling is required, we could not test whether other mucosal localized IL-21 responsive populations (e.g. B or NK cells) also depend on intrinsic IL-21R sensing. Defining these cell type specific effects will require additional inducible knockouts. Finally, while Bcl6 restrains HIF-1α expression in TRH cells, the mechanistic basis remains unresolved. Our results suggest that HIF-1α promotes Blimp-1 dependent IL-21 production, but this relationship is correlative. Epistasis models (e.g. single and double inducible Prdm1 and Hif1a deletion models) will be required to test this hierarchy more directly.

RESOURCE AVAILABILITY

LEAD CONTACT

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Carolyn King (carolyn.king@unibas.ch).

MATERIALS AVAILABILITY

This study did not generate new unique reagents. Mouse lines generated in this study are available from the lead contact upon request.

DATA AND CODE AVAILABILITY

Single-cell RNA-seq data have been deposited at GEO: accession number and are publicly available as of the date of publication. Microscopy data reported in this paper will be shared by the lead contact upon request. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mice

C57BL/6Rj (CD45.2) were purchased from Janvier Labs. Egln3-YFP mice, B6-Elgn3<tm1YFPAgo>, were generated by A. Goldrath²². Bcl6^flox/flox mice⁸⁸ and Hif1a^flox/flox mice⁸⁹ were crossed with CD4^ERT2 × Rosa-26 mice⁹⁰. IL-21R^−/− mice, <B6.129-Il21rtm1Kopf/J>, were generated by M.Kopf⁸⁷. Il21-GFP mice were generated by S. Nutt⁷. T-bet-Zs-Green, Tg(Tbx21-ZsGreen)E3ZJfz, were generated by W. Paul⁹¹. Mice in each experiment were 6–8 weeks old, same-sex littermates, maintained and bred in the specific pathogen–free animal facility at the University of Basel. All animal experiments were performed in accordance with local and Swiss federal guidelines.

Influenza infection

Influenza A/PR8/34 (hereafter PR8) and X31 were produced by J. Sun, University of Virginia. PR8 was used at 500 PFU/TCID50 and X31 at 800 PFU/TCID50. Mice were anesthetized with vaporized isoflurane and intratracheally infected intratracheally with virus diluted in 50 μL of PBS.

Bone Marrow Chimera

To produce mixed chimeras, recipient B6-CD45.1 mice were irradiated with two doses of 450 cGy given 4 hours apart. After irradiation, recipients were reconstituted by intravenous injection of hematopoietic cells collected from femurs and tibiae of donor mice (50% B6-CD45.1 and 50% IL-21R^−/−), as described before⁹². Mice were used for experiments 6–8 weeks after irradiation.

KPAR tumor

KPAR cells were kindly provided by Julian Downward, Francis Crick Institute, London^62,93. 2.5×10⁵ KPAR cells (maintained at a maximum of 5 passages) were intravenously injected into the tail vein of each mouse.

METHOD DETAILS

In vivo treatments

Tamoxifen (Sigma-Aldrich) dissolved in corn oil (25 mg/ml) and administered by oral gavage (3.75 mg/mouse) at indicated time points. Anti-CD45 AF700 (Biolegend) was intravenously injected (3 μg/mouse) just prior to sacrifice and organ harvest. Recombinant mouse IL-21 protein (PeproTech) was intratracheally administered every other day (10 μg/treatment/mouse) for the indicated time points. LTβR-Fc was generated and kindly provided by Bart Lambrecht and Stijn Vanhee. Mice were intraperitoneally injected with 100ug/mouse for the indicated time points. Hypoxyprobe-1 (Pimonidazole by Hpi) was intravenously injected (60mg/kg) 90 minutes prior to organ harvest. In situ anti-pinomidazole and secondary antibody staining were performed according to the manufacturer instructions. IL-21 (BD Bioscience) staining for histology was made with lungs from animals previously i.v. injected with Brefeldin A (Invitrogen, 2.5 mg/kg) 16h before organ harvest.

In vitro assay

Naïve CD4⁺ T cells were isolated from mouse spleen and lymph nodes according to the manufacturer’s instructions (Stemcell Technologies). Cells were plated at a density of 2×10⁵ cells per well in flat-bottom plates pre-coated with anti-CD3 and anti-CD28 antibodies (BD Bioscience, 5 μg/mL each). CD4⁺ T cells were cultured for 72 hours in T cell activation medium (RPMI 1640 supplemented with 10% FBS, 2 mM L-glutamine, 100 U/mL penicillin, 100 μg/mL streptomycin, 10 mM HEPES, and 50 μM β-mercaptoethanol) containing TGF-β1 (5 ng/mL), IL-6 (20 ng/mL), and IL-23 (20 ng/mL). After activation, cells were re-stimulated for 5 hours with a protein transport inhibitor and cell stimulation cocktail (Invitrogen), following the manufacturer’s protocol. Finally, cells were collected and stained for flow cytometry analysis

Tissue preparation

Lungs were harvested and diced into gentleMACS C Tubes (Miltenyi Biotec) and washed down with 3 ml of media [RPMI, 10 mM aminoguanidine hydrochloride (Sigma-Aldrich), 10 mM Hepes, penicillin-streptomycin-glutamine (100X, Gibco), 2-mercaptoethanol (50 μM, Gibco)], prewarmed in a 37°C water bath. Digestion mix containing liberase TL (for lymphoid cells) or liberase TM (for myeloid cells) (33.3 μg/ml, Merk) and deoxyribonuclease I (58 μg/ml; Applichem) was added. Lungs were dissociated on a gentleMACS Dissociator (Miltenyi Biotec) and mashed through 70-μm MACS SmartStrainers as previously described (Miltenyi Biotec)⁴. Mouse mediastinal LNs and spleens were harvested and mashed through 100-μm Corning nylon cell strainers. For scRNA-seq experiments, mediastinal LNs were additionally digested with liberase TL. Mouse livers were perfused with PBS and filtered through 100-μm Corning nylon cell strainers. Lymphocytes from the liver were further purified on a 70%/40% Percoll gradient (Fisher Scientific).

Tetramer and flow cytometry staining

I-Ab NP311–325 (QVYSLIRPNENPAHK) allophycocyanin (APC-) conjugated was obtained from the National Institutes of Health (NIH) tetramer core facility. Single-cell suspensions were tetramer-stained, enriched, and counted. For all fluorochrome-conjugated antibody dilutions, FACS buffer (PBS, 2% FCS, and 0.1% sodium azide) containing Fc block (Clone 2.4G2 InVivoMAb anti-mouse CD16/CD32; BioXCell, #BE0307) was used. Live-dead stain (provided by Zombie Fixable Viability^™ Sampler Kit; BioLegend) was added to the antibody mix and stained at 4°C for 30 min. FoxP3 Transcription Factor Staining Buffer Set (eBioscience) was used for intracellular staining at room temperature (RT) 30–60 minutes. Flow cytometric analysis was performed on BD LSR Fortessa and Cytek Aurora. Data was analyzed using FlowJo X software (BD) and SpectraFlow (Cytek). Absolute cell numbers were based on AccuCheck Counting Beads (Invitrogen) manufacturer.

Confocal microscopy

Lungs were inflated with a 1:1 mixture of Tissue-Tek^® OCT. Compound (Sakura) and PBS containing 1% methanol-free PFA (Thermo Fisher Scientific) and fixed with 4% PFA for 24 hours. Lungs were washed in PBS, incubated in a 30% sucrose solution for 36 hours and embedded in OCT. For immunostaining of lung sections from Egln3-YFP mice, tissues were sectioned at 10–30 μm on a cryostat and incubated with 5% blocking buffer (Jackson ImmunoResearch) for 2 hours. Slides were incubated with indicated antibodies overnight at 4°C followed by washing with PBS, DAPI staining (2 μg/mL, Merck) and treatment with TrueView Autofluorescence Quenching Kit (Vector) to reduce autofluorescence. Sections were mounted using SlowFade Diamond Antifade Mountant (Thermo Fisher) to preserve Egln3-YFP fluorescence. Z-stack images (10μM to 30μM, step size of 0.9μM) were acquired on a Nikon ECLIPSE Ti2 inverted microscope, equipped with a X-light V3 spinning disk confocal module and a 40X Apo Plan lambda, NA=0.95 objective.

Image quantification

Multichannel fluorescence images were analyzed with QuPath (ver. 0.4.3). Cells were segmented based on nuclear DAPI staining using a pre-trained deep learning model cyto3 from cellpose, with parameters: pixel size: 0.33; object diameter: 30, cellExpansion: 2.0. Positivity of cells (Biolegend CD4 AF647 and Egln3-YFP) was determined based on the mean intensity measured within the cell (ROI). The B cell cluster was identified using a pixel classifier based on B220 staining. A training image was composed of a representative set of 6 images which was annotated by a user to identify positive and negative B220 (Thermo Fisher, eFluor-570) signals. ANN_MLP classifier was trained with the following parameters: resolution: 2.66μm; filters: Gaussian and Laplacian; scales: 1 and 4. Finally, the signed distance of each cell to the B cell cluster annotation was calculated using the detectionToAnnotationDistancesSigned function, and individual cell distance measurements were extracted in a csv file.

Real-time PCR

RNA was extracted from sorted NP⁺ Th1, Egln3⁻ TRH cells, and Egln3⁺ TRH cells using PicoPure^™ RNA Isolation Kit (Thermo Fisher). RNA was reverse transcribed into cDNA using iScript cDNA (Synthesis Kit, Bio-Rad). Taqman Fast Advanced probes (Applied Biosystems) for Il21 (#Mm00517640_m1), Hif1a (#Mm00468869_m1) and Rpl13a (#Mm05910660_g1). were used for this assay. Influenza A RNA titers were measured using the NZYTech kit, following the manufacturer’s instructions. For data presentation, relative expression was calculated and normalized to Rpl13a. Fold change values were calculated using the 2^−ΔΔCT method.

Blood oxygen saturation (SpO₂)

Blood SpO₂ was measured using a MouseOx^™ pulse oximeter (Starr Life Sciences). Anesthetized mice were placed on a heating pad to maintain body temperature and an oximeter paw sensor was secured to the hind limb using medical adhesive tape. During the recording, SpO₂ measurements were taken 3 times, each measurement lasting approximately 10 minutes.

Viral titers during influenza challenge

Mice were infected with the X31 influenza strain for 30 days followed by secondary infection with PR8 influenza. Viral titers in the BALF were measured 36 hours after challenge as previously described⁹⁴.

ELISA

The coating method for plates, antibody concentrations used, and the steps for detecting PR8-specific IgG and IgA were performed as previously described⁹⁵. Lung homogenate samples were pooled and concentrated using Amicon Ultra Centrifugal Filters (Merk) with a cut-off of 30kDa. IL-21 ELISA assays were performed according to the manufacturer’s instructions (Invitrogen).

Single cell RNA sequencing

Following influenza infection, control (CD4^Cre-ERT2ROSA^ki/+) and Hif1a^flox/floxCD4^Cre-ERT2 mice were treated with tamoxifen to induce gene deletion in CD4 T cells. Individual mice were hash tagged using TotalSeq-B anti-mouse hashtag antibodies (B0301-B0304, BioLegend), and CD45⁺ cells were sorted from lungs at d45 p.i. and submitted for library preparation. A total of 10,000 target cells per sample were loaded. Single-cell capture and cDNA library preparation was performed according to manufacturer’s instructions using 10x Genomics Chromium Next GEM Single Cell 3’ Kit v2 reagents using dual indexing. Sequencing was carried out on an Illumina Novaseq 6000 at the Genomics Facility Basel, ETH Zurich.

Sequencing reads were mapped to the mouse transcriptome (mm39) using STARsolo for both gene expression and antibody-bound hashtag oligos using cell barcode whitelist from cellranger-7.0.0. Unless otherwise noted, analysis steps were performed using Bioconductor packages: SingleCellExperiment, scater, scran, scuttle, bluster, DropletUtils, SingleR, edgeR, limma^96–103. Visualizations with ggplot¹⁰⁴. Demultiplexing hashtag oligos and doublet detection were done using DropletUtils. Cells were further filtered for library size, library complexity and mitochondrial content using adaptive thresholding provided by the R package scuttle’s perCellQCFilters function with default arguments. Normalization was performed using the R package scran’s deconvolution method. Technical noise within gene expression was modeled using scran, and biologically relevant highly variable genes (HVG) were used after separating the technical from biological variance. Principal component analysis (PCA) was run on the normalized data using the top 2000 most variable genes by biological variance, and the PCA was denoised to account for the modeled technical variation. Cells were clustered using the bluster R package, either hierarchically using Ward’s method or with the graph-based leiden method. Cell annotation was performed using the SingleR package with references from the celldex package followed by manual curation based on cluster-identifying genes enabled the identification of NK/ILC1 cells. Secondary analysis of published datasets was downloaded from Gene Expression Omnibus (GEO). Single-cell regulatory network inference and clustering (SCENIC) was performed on data from GEO dataset GSE146626 using version 0.9.7 and the published workflow on the pySCENIC github repository^105,106. GSEA between conditions was performed using cameras from the limma R package on standard gene set categories (MSigDB) as well as sets curated from relevant publications.

Spatial transcriptomics

Lungs were fixed with 4% methanol-free PFA for 24 hours at 4°C, washed with PBS and paraffinized using Tissue Processing Center (TPC 15 Trio, Medite) and embedded in paraffin using TES station machine (Tissue Embedding Station, Medite). FFPE blocks were kept at 4°C and 5μm sections and placement of sections were performed according to the protocol CG000578 (10X Genomics), handled by the histology facility of Basel. Xenium Prime 5K processing and acquisition (with cell segmentation) was performed by Functional Genomics Center Zurich, following the user guides GC000760, GC000749 and GC00582 using Xenium in situ 5K gene expression kit (10X Genomics).

Probability-based segmentation was performed using Proseg as an alternative to 10x Genomics’ multi-modal segmentation¹⁰⁷. Analysis steps such as finding HVGs, PCA, cell annotation and clustering were done with bioconductor R packages as in sc-RNAseq processing. Identification of TLS: normalized expression of Cd79b was summed for the 32 nearest neighbors of each cell. Every cell with a log10 neighbor expression higher than 1.05 was deemed a TLS cell (threshold determined with manual comparison to H&E image). The general idea was to find areas not only where an individual cell has high Cd79b expression, but also where it is surrounded by “enough” other cells that also have high Cd79b expression. TLS cells were then clustered hierarchically based on their X and Y coordinates to define discrete TLS structures across the section. Cells within 300 μm of TLS cells that were themselves not also TLS cells were considered to be in the TLS periphery, and the distance to nearest TLS edge was recorded for TLS and TLS periphery cells (with TLS and TLS periphery cells (with TLS and TLS periphery cells receiving negative and positive distance values respectively). Gene set scores calculated using the UCell R package. Gene positivity trends were modeled using a binomial general linearized model using the stats R package with the design: (num_cells_with_gene, num_cells_without_gene) ~ step + TLS_ID where cells with > 0 counts of a given gene in a given distance step are considered positive and step is an ordered factor of the 8μm distance bins. The TLS_ID of the largest TLS (n=4) was used as a covariate to account for differences between different structures/location in the tissue slice.

QUANTIFICATION AND STATISTICAL ANALYSIS

Prism software (GraphPad 10. 4.1) was used for all statistical analyses. All statistical tests used are two-tailed unpaired t-tests, one/two-way ANOVA, Aggregate comparison (AUC) and Mantel-Cox test, which are assigned to each individual figure panel legend. Statistical significances are indicated by *p<0.05, **p<0.005, ***p<0.0005, ****p<0.0001, ns, not significant (p > 0.05). In summary graphs, points indicate individual samples, and horizontal lines are mean unless otherwise noted. All error bars are standard deviation.

Supplementary Material

Supplemental Figures S1-s5

Document S1. Figures S1–S5.

KEY RESOURCES TABLE

REAGENT OR RESOURCE	SOURCE	IDENTIFIER
Antibodies
CD11b BV510 (Clone M1/70)	Biolegend	Cat#101245, RRID:AB_2561390
IA/IE AlexaFluor700 (Clone M5/114.15.2)	Biolegend	Cat#107621, RRID:AB_493726
CD86 PE-Dazzle594 (Clone GL-1)	Biolegend	Cat#105041,RRID:AB_2566408
F4/80 BV605 (Clone BM8)	Biolegend	Cat#123133, RRID:AB_2562305
CCR2 BV421 (Clone SA203G11)	Biolegend	Cat#150605, RRID:AB_2571913
CD64 BV711 (Clone X54-5/7.1)	Biolegend	Cat#139311, RRID:AB_2563846
CD11c BUV395 (Clone HL3)	BD Bioscience	Cat#564080, RRID:AB_2738580.
SiglecF PE-cy7 (Clone S17007L)	Biolegend	Cat#155527, RRID:AB_2890715
CD11a/LFA1 FITC (Clone M17/4)	Biolegend	Cat#101106, RRID:AB_312779
MerTK PE (Clone 2B10C42)	Biolegend	Cat#151505, RRID:AB_2617036
CD45 APC-Fire810 (Clone 30-F11)	Biolegend	Cat103173, RRID:AB_2860599
CD45 AlexaFluor700 (Clone 30-F11)	Biolegend	Cat#103127, RRID:AB_493714
CD3 APC (Clone145-2C11)	Biolegend	Cat#100235, RRID:AB_2561455
NKp46 APC (Clone 29A1.4)	Biolegend	Cat#37607, RRID:AB_10612749
CD19 APC (Clone 6D5)	Biolegend	Cat#at#115512, RRID:AB_313647
CD4 BUV496 (Clone GK1.5)	BD Bioscience	Cat#741050, RRID:AB_2813886
CD44 BUV395 (Clone IM7)	BD Bioscience	Cat#568507, RRID:AB_3677624
PD1 BV785 (Clone 2RF.1A12)	Biolegend	Cat#135225, RRID:AB_2563680.
PSGL1 BB700 (Clone 2PH1)	BD Biosciences	Cat#742183, RRID:AB_2871417.
CXCR6 BV421 (Clone K041E5)	Biolegend	Cat#356013, RRID:AB_2562514
CXCR5 BV421 (Clone L138D7)	Biolegend	Cat#145511, RRID:AB_2562127
CXCR4 BV605 (Clone L276F12)	Biolegend	Cat#Cat#146519, RRID:AB_2687359
CXCR3 BV650 (Clone CXCR3-173)	Biolegend	Cat#26531, RRID:AB_2563160
CD11c BV510 (Clone N418)	Biolegend	Cat#562949, RRID:AB_2732056
F4/80 BV510 (Clone BM8)	Biolegend	Cat#Cat#123135, RRID:AB_2562622
B220 BV510 (Clone RA3-6B2)	Biolegend	Cat#103247, RRID:AB_2561394
CD3 BV510 (Clone 17A2)	Biolegend	Cat#100233, RRID:AB_2561387
NK1.1 BV510 (Clone PK136)	Biolegend	Cat#108737, RRID:AB_2562216
Ly6G BV510 (Clone 1AB)	Biolegend	Cat#127633, RRID:AB_2562937
CD11b PE-cy5 (Clone M1/70)	Biolegend	Cat#101209, RRID:AB_314159
F4/80 PE-cy5 (Clone BM8)	Biolegend	Cat#123111, RRID:AB_893494
B220 PE-cy5 (Clone RA3-6B2)	Biolegend	Cat#103209, RRID:AB_312994
CD3 PE-Cy5 (Clone 17A2)	Biolegend	Cat#555276, RRID:AB_395700
NK1.1 PE-cy5 (Clone PK136)	Biolegend	Cat#108715, RRID:AB_493591
Ly6G PE-cy5 (Clone 1A8)	Biolegend	Cat#127671, RRID:AB_2904289
FR4 PE-cy7 (Clone 12A5)	Biolegend	Cat#125012, RRID:AB_1134199
NK1.1 BV605 (Clone PK136)	Biolegend	Cat#108739, RRID:AB_3662406.
EOMES PE (Clone W17001A)	Biolegend	Cat#157705, RRID:AB_2888891
CD38 BV421 (Clone 90/CD38)	BD Bioscience	Cat#659470, RRID:AB_2870499
IgM BV785 (Clone R6-60.2)	Biolegend	Cat#564028, RRID:AB_2738553
IgD BV510 (Clone 11-26C.2a)	Biolegend	Cat#405723, RRID:AB_2562742
IgA APC (Clone mA-6E1)	Thermo Fisher	Cat#17-4204-82, RRID:AB_2848294
IFNγ APC-cy7 (Clone XMG1.2)	Biolegend	Cat#505849, RRID:AB_2616697
CD4 AF674 (Clone RM4-5)	Biolegend	Cat#100533, RRID:AB_493372
B220 eFluor570 (Clone RA3-6B2)	Thermo Fisher	Cat#41-0452-80, RRID:AB_2573598
CD8a APC-cy7 (Clone 53-6.7)	Biolegend	Cat# 100714, RRID:AB_312753
CD103 Percp-cy5.5 (Clone 2E7)	Biolegend	Cat# 156907, RRID:AB_2876530
CD69 BV711 (Clone H1. 2F3)	Biolegend	Cat# 104537, RRID:AB_2566120
IL-21 PE (clone mhalx21)	Thermo Fisher	Cat#12-7213-80 RRID:AB_1834467
Ki67 AF488 (Clone Ki-67)	BioLegend	Cat# 350508, RRID:AB_10933085
Blimp-1 AF488 (Clone 5E7)	Thermo Fisher	Cat# 53-9850-82, RRID:AB_2848453
Bcl2 PE (Clone Bcl2/100)	BD Biosciences	Cat# 340576, RRID:AB_400061
HIF-1α (Clone E1V6A)	Cell Signaling Technology	Cat# 65036BC, RRID:AB_3073709
Goat anti-rabbit IgG AF488	Invitrogen	Cat# A-11008 RRID:AB_143165
TotalSeq B0302	Biolegend	Cat#155833, RRID:AB_2814068
TotalSeq B0303	Biolegend	Cat#155835, RRID:AB_2814069
TotalSeq B0304	Biolegend	Cat#55837, RRID:AB_2814070
LTβR-Fc	VIB, Drs. Bart Lambrecht and Vanhee Stijn	N/A
InVivoMAb rat IgG2a isotype control, anti-trinitrophenol	Bio X Cell	Cat#BE0251, RRID:AB_2687732
Bacterial and virus strains
PR8 Influenza A virus (strain A/Puerto Rico/8/1934 H1N1)	University of Virginia, Dr. Jie Sun	N/A
X-31 Influenza A virus (strain A/X-31 H3N2)	University of Virginia, Dr. Jie Sun	N/A
Biological samples
N/A	N/A	N/A
Chemicals, peptides, and recombinant proteins
DAPI	Merk	Cat#D9542-1MG
Tamoxifen	MCE	Cat#HY-13757A
Liberase TL	Merk	Cat#5401020001
Liberase TM	Merk	Cat#5401119001
DNAse-I	Merk	Cat#10104159001
Mouse IL-21 Recombinant Protein	PeproTech	Cat#210-21-1MG
AccuCheck Counting Beads	Invitrogen	Cat#PCB100
Brefeldin A	Invitrogen	Cat# B7450
eBioscience™ Protein Transport Inhibitor Cocktail (500X)	Invitrogen	Cat#00-4980-93
eBioscience™ Cell Stimulation Cocktail (500X)	Invitrogen	Cat# 00-4975-93
Recombinant Mouse TGF-β1	Biolegend	Cat#763104
Recombinant Mouse IL-6	Biolegend	Cat#575708
Recombinant Mouse IL-23	Biolegend	Cat#589008
Mouse naive CD4 T Isolation Kit	Stemcell Technologies	Cat#19765
Dynabeads mouse anti-CD3/CD28	Invitrogen	Cat# 11452D
DMOG	Merk	Cat#D3695-50MG
Critical commercial assays
TrueView Autofluorescence Quenching Kit	VectorLabs	Cat#SP-8400
Zombie Fixable Viability™ Sampler Kit	Biolegend	Cat#423117
TaqMan Fast Advanced Master Mix	Invitrogen	Cat#4444556
iScript cDNA Synthesis Kit	Bio Rad	Cat#1708891
PicoPure RNA Isolation Kit	Invitrogen	Cat#KIT0204
Chromium Next GEM Single Cell 3’ Kit v2	10x Genomics	Cat#1000269
Xenium in situ 5K gene expression kit	10x Genomics	Cat#1000724
Illumina Novaseq 6000	Illumina	Cat#1000724
Mouse IL-21 ELISA kit	Invitrogen	Cat#BMS6021
Amicon Ultra Centrifugal Filters	Merk	Cat#UFC9030
Hypoxyprobe-1 kit	Hpi	Cat#HP1-XXX
FITC BrdU Flow Kit	BD Biosciences	Cat#559619
Influenza A virus RT-qPCR Kit	NZY Tech	Cat#MD07312
Deposited data
sc-RNAseq	Gene Expression Omnibus	GSE291400
Xenium image-based spatial transcriptomics	Gene Expression Omnibus	GSE291539
Experimental models: Cell lines
KPAR 1.3v	Francis Crick Institute (UK), Dr. Julian Downward	N/A
Experimental models: Organisms/strains
Mouse: C57BL/6Rj	Jackson Laboratories	RRID:IMSR_JAX:000664
Mouse: B6 CD45.1	Jackson Laboratories	RRID:IMSR_JAX:002014
Mouse: B6-Elgn3-YFP	Likanen et al22	N/A
Mouse: HIF1αflox	Jackson Laboratories	RRID:IMSR_JAX:007561
Mouse: CD4CreERT2	Jackson Laboratories	RRID:IMSR_JAX:022356
Mouse: IL-21R−/−	Fruhlich et al.87	RRID:IMSR_JAX:019115
Mouse: IL-21-GFP	Luthej et al.7	N/A
Mouse: IL-21-VFP	Son et al.6	RRID:IMSR_JAX:030295
Mouse: T-bet Zs-Green	Jackson Laboratories	RRID:IMSR_TAC:8419
Mouse: Bcl6flox	Jackson Laboratories	RRID:IMSR_JAX:023727
Oligonucleotides
Primer: Taqman Hif1a probe	Thermo Fisher	Cat#Mm00468869_m1
Primer: Taqman Il21 probe	Thermo Fisher	Cat#Mm00517640_m1
Primer: Taqman Rpl13a probe	Thermo Fisher	Cat#Mm05910660_m1
Recombinant DNA
N/A	N/A	N/A
Software and algorithms
FlowJo	BD Bioscience	https://www.flowjo.com/solutions/flowjo
GraphPad Prism	GraphPad	https://www.graphpad.com/
QuPath	QuPath	https://qupath.github.io/
Stardist	QuPath/Stardist	https://stardist.net/
STARsolo	N/A	https://github.com/alexdobin/STAR/blob/master/docs/STARsolo.md
Bioconductor R packages (see below)	N/A	https://bioconductor.org/
pySCENIC	N/A	https://github.com/aertslab/pySCENIC
10x Genomics Xenium Explorer	10X Genomics	https://www.10xgenomics.com/support/software/xenium-explorer/latest
Gene sets: SHAFFER_IRF4_TARGETS_IN_ACTIVATED_B_LYMPHOCYTE REACTOME_DOWNSTREAM_SIGNALING_EVENTS_OF_B_CELL_RECEPTOR_BCR SIG_BCR_SIGNALING_PATHWAY WP_T_CELL_RECEPTOR_AND_COSTIMULATORY_SIGNALING GOBP_POSITIVE_REGULATION_OF_ACTIVATED_T_CELL_PROLIFERATION GOBP_POSITIVE_REGULATION_OF_T_CELL_RECEPTOR_SIGNALING_PATHWAY GOCC_T_CELL_RECEPTOR_COMPLEX GSE13738_RESTING_VS_TCR_ACTIVATED_CD4_TCELL_DN	Broad Institute	https://www.gsea-msigdb.org/gsea/msigdb
Other
N/A	N/A	N/A