Selective elimination of circulating effector CD8 T cells via LTβR blockade separates anti-CD137 efficacy from toxicity

Abstract

The major barrier to clinical translation of αCD137 immunotherapy is separating antitumor efficacy from hepatotoxicity driven by IFN-γ–producing CD8 T cells. We propose a strategy to limit toxicity by promoting contraction of excessively expanded CD8 T cells. We identify CD11c⁺KLRG1⁺ effector CD8 T cells (CD11c⁺T_E) as the primary source of IFN-γ, recirculating between blood and secondary lymphoid organs (SLOs), where they undergo apoptosis during contraction. We show that lymphotoxin β receptor (LTβR) signaling restrains this contraction. Mechanistically, lymphotoxin-expressing B cells activate LTβR in fibroblastic reticular cells (FRCs), suppressing apoptosis of CD11c⁺T_E cells in the spleen and sustaining their systemic circulation and liver infiltration. Pharmacological LTβR blockade abrogates hepatotoxicity by reducing the accumulation of IFN-γ–producing CD11c⁺T_E cells while preserving tumor-specific CD8 T cell responses. These findings identify LTβR as a key regulator of effector CD8 T cell persistence and support the use of LTβR antagonists to improve the safety of αCD137-based immunotherapy.

LTβR blockade curbs CD137-driven hepatotoxicity by promoting effector CD8 T cell contraction while preserving anti-tumor immunity.

INTRODUCTION

CD137 (also known as 4-1BB or tumor necrosis factor receptor superfamily member 9) is an inducible costimulatory receptor that enhances CD8 T cell responses upon engagement with its natural ligand CD137L (1–3). Clinically, increased frequencies of CD137⁺ T cells correlate with improved overall survival and favorable responses to programmed cell death protein 1 (PD-1) blockade in patients with cancer, highlighting CD137 as an attractive target for combined immunotherapy (4, 5). Systemic stimulation with agonistic αCD137 antibodies induces a robust CD8 T cell activation, promoting survival, reversing anergy and exhaustion, and driving proliferation and effector differentiation in a T cell receptor (TCR)–independent manner (6–10). Consistent with these effects, CD137 agonists, either alone or in combination with immune checkpoint inhibitors, enhance CD8 T cell–mediated antitumor immunity in numerous preclinical cancer models (11–13). However, despite these promising preclinical results, clinical translation of αCD137 immunotherapy has been limited by severe immune-related adverse events (irAEs). Early clinical trials of urelumab, a strong agonistic αCD137 antibody, were halted because of dose-limiting hepatotoxicity (14). Although subsequent dose reductions or the use of weaker agonists such as utomilumab improved safety, therapeutic efficacy at these lower doses has been limited, leading to suspension of clinical development for both agents (2, 15). Thus, achieving an effective balance between antitumor efficacy and immune-mediated toxicity remains a major challenge for αCD137-based immunotherapy (14). Current strategies to address this limitation include engineering next-generation CD137-targeting molecules (16–22) and modulating host immune responses to selectively limit pathogenic immunity while preserving antitumor function (23–26).

Mechanistic studies in animal models have demonstrated that αCD137 immunotherapy–induced hepatotoxicity is driven predominantly by the excessive expansion of CD8 T cells that lack specificity for tumor cells or hepatocytes. αCD137 agonistic antibodies induce robust, antigen-independent expansion of CD8 T cells (8, 27), leading to multifocal intrahepatic infiltration of interferon-γ (IFN-γ)–producing effector CD8 T cells in tumor-free mice (23, 28, 29). These infiltrating cells are polyclonal and largely lack liver antigen specificity (28, 30). IFN-γ produced by these CD8 T cells mediates hepatocyte injury (30), while inflammatory cytokines derived from Kupffer cells and monocyte-derived macrophages further amplify CD8 T cell activation, survival, and recruitment (24, 25, 31). Depletion of CD8 T cells or IFN-γ abrogates the hepatotoxicity (30). Recent clinical studies report clonal expansion and activation of IFN-γ–producing effector CD8 T cells even at low doses of urelumab (8 mg per patient) (32, 33), underscoring the translational relevance of these mechanisms. At the same time, CD8 T cells are essential for the antitumor efficacy of αCD137 immunotherapy, as depletion of total CD8 T cells abolishes therapeutic benefit (23, 34). Notably, CD137 stimulation promotes the survival of tumor-specific CD8 T cells rather than driving their massive expansion (35). Together, these observations highlight a central challenge in αCD137 immunotherapy: how to uncouple protective antitumor responses mediated by tumor-specific CD8 T cells from pathogenic responses driven by excessively expanded, bystander CD8 T cells.

Lymphotoxin β receptor (LTβR), a member of the tumor necrosis factor (TNF) receptor superfamily, is a potential target for mitigating irAEs. LTβR has emerged as a regulator of CD8 T cell–driven pathology across diverse inflammatory conditions, including viral infection, allograft rejection, graft-versus-host disease, and autoimmunity (36–40). Beyond immunopathogenic contexts, lymphotoxin (LT)-LTβR signaling promotes CD8 T cell responses during infection and antigen-dependent immune activation by inducing type I IFNs (41–44), highlighting its broad role in regulating adaptive immunity. In addition, LTβR serves as a central mediator of immune-stromal communication and is essential for lymph node (LN) development during embryogenesis and for maintenance of secondary lymphoid organs (SLOs) architecture and homeostasis in adulthood (45–50). These structural and homeostatic roles are primarily mediated through LTβR signaling in fibroblastic reticular cells (FRCs), which regulates chemokine production and immune cell positioning within SLOs (41, 51). Collectively, these observations indicate that LTβR signaling exerts pleiotropic effects on CD8 T cell immune responses. Given the central role of CD8 T cells in αCD137 immunotherapy–induced toxicity, we hypothesized that LTβR blockade could be leveraged to mitigate hepatotoxicity by attenuating effector CD8 T cell responses.

In this study, we used complementary genetic and pharmacological approaches to investigate how LTβR signaling regulates the αCD137 antibody–induced CD8 T cell response. We identify LTβR as a negative regulator of bystander CD8 T cell contraction within SLOs, thereby sustaining their systemic circulation and pathological accumulation in the liver. These findings reveal LTβR signaling as a target to dissociate αCD137 immunotherapy–induced hepatotoxicity from antitumor potency, thereby addressing a major barrier to the therapeutic translation of CD137 agonists.

RESULTS

CD11c⁺KLRG1⁺CD8 T cells are the major producers of IFN-γ in the liver following αCD137 immunotherapy

To investigate the mechanisms underlying αCD137 immunotherapy–induced hepatotoxicity, we established a mouse model in which mice received four consecutive intraperitoneal injections of an agonistic αCD137 antibody (Fig. 1A). This treatment elevated serum alanine aminotransferase (ALT) levels, which peaked between days 12 and 15 (Fig. 1B and fig. S1) and was accompanied by immune cell infiltrates localized primarily in perivascular regions (Fig. 1C and fig. S1B) and adjacent to apoptotic hepatocytes (fig. S1C). The αCD137 antibody induced the intrahepatic accumulation of diverse CD8 T cell subsets that express markers of effector differentiation, such as KLRG1 and CD11c (integrin alpha X) (24, 29, 31, 52). Flow cytometry analysis of perfused livers also revealed a marked accumulation of CD8 T cells along with other cell types following our treatment scheme (Fig. 1D and fig. S1D). Most intrahepatic CD8 T cells expressed KLRG1, a marker of effector T cells (T_E), which were further subdivided into CD11c⁺ and CD11c⁻ populations (Fig. 1E and F). CD11c⁺T_E cells exhibited greater heterogeneity, comprising both CD127⁻ short-lived effectors (T_SLE) and CD127⁺ double-positive effectors (T_DPE), while CD11c⁻T_E mainly consisted of T_SLE (fig. S1, E and F). We next examined the contribution of these infiltrating CD8 T cell subsets to IFN-γ production. CD11c⁺T_E and CD11c⁻T_E subsets were the major IFN-γ producers in the liver (Fig. 1, G to I). Given that αCD137 immunotherapy–induced hepatotoxicity is driven by CD8 T cells in an IFN-γ–dependent manner (23, 30), our findings suggest that T_E cells contribute to liver pathology.

Fig. 1. αCD137 immunotherapy induces intrahepatic accumulation of IFN-γ–producing CD11c⁺T_E cells.

(A to H and K) C57BL/6 mice were treated with an αCD137 agonistic antibody or an isotype control and analyzed on day 13 as shown. Data were pooled from three independent experiments. (A) Experimental schematic. (B) Serum ALT. (C) Hematoxylin and eosin (H&E) staining of liver sections. Scale bars, 100 μm. (D) Cell numbers of indicated populations in the liver. (E) Gating strategy for effector KLRG1⁺CD8 T cells (T_E). (F) Number of CD8 T cell subsets in the liver expressing indicated markers. (G) Frequency of indicated IFN-γ⁺ CD8 T cell subsets in the liver. (H) IFN-γ production by the indicated subsets of CD8 T cells from the liver. Solid black line, total CD8 T cells from a control mouse. (I) IFN-γ production by intrahepatic CD8 T cells without ex vivo stimulation from GREAT reporter mice (IFN-γ–eYFP) treated with αCD137 antibody as indicated in (A) (one experiment). Dashed black line, fluorescence minus one (FMO) control. (J) Volcano plot depicting differential gene expression between CD11c⁺ and CD11c⁻ T_SLE in the liver on day 8 (n = 4, one experiment). The top 10 up- and down-regulated genes are annotated. Flow cytometry gating strategy for T_SLE is shown in fig. S1E. (K) NKG2A and PD-1 expression on CD11c⁺T_E and CD11c⁻T_E in the liver. (L to P) C57BL/6 mice were treated with αNKG2A or αPD-1 inhibitory antibodies alone or in combination with αCD137 antibody as indicated in (L). Data were pooled from two to three independent experiments. (L) Experimental schematic. [(M) and (O)] Serum ALT. (N) Number of intrahepatic CD11c⁺T_E and CD11c⁻T_E cells following treatment with αNKG2A or (P) αPD-1 antibodies along with αCD137 agonist. P values are as indicated by Student’s t test (B), two-way analysis of variance (ANOVA) plus Sidak’s [(D), (F), and (G)] or Tukey’s multiple comparisons test [(M) to (P)]. C, control; Ab, antibody; NKT, natural killer T cells; NK, natural killer; Neu, neutrophils; KCs, Kupffer cells; WT, wild type; FC, fold change. The cartoon mouse was created in BioRender. Korchagina, A. (2026) https://BioRender.com/tssh5ms.

Further transcriptomic and protein analysis of liver-infiltrating T_E subsets revealed that the CD11c⁺ population expressed the inhibitory receptor NKG2A, the activating receptor NKG2D, granzyme A, and the adhesion molecule Ly6C (Fig. 1, J and K, and fig. S1, G and H). In contrast, well-known effector markers, such as CD44, granzyme B, T-bet, Eomesodermin (EOMES), and chemokine receptors CXCR3 and CX3CR1, were abundantly expressed in both CD11c⁺ and CD11c⁻ subsets (fig. S1H). In addition, we observed increased expression of various chemokines, including ligands for CXCR3 (Cxcl9 and Cxcl10), adhesion molecules (Vcam1 and Icam1), cytotoxic T cell cytokines (Ifng, Tnf, and Lta), and macrophage-associated cytokines (Il12 and Il27) in the liver of αCD137 antibody–treated mice, indicating T_E cell recruitment and Kupffer cell activation (fig. S1I). Expression of H2-T23 (Qa-1), a ligand for NKG2A, was also up-regulated (fig. S1I). Given the expression of the inhibitory receptor NKG2A on IFN-γ–producing CD11c⁺T_E cells in the liver, we hypothesized that NKG2A blockade may exacerbate αCD137 antibody–induced liver injury. Administration of an αNKG2A antagonistic antibody further increased the ALT elevation induced by αCD137 immunotherapy and led to enhanced accumulation of CD11c⁺T_E cells in the liver, whereas αNKG2A treatment alone had no effect (Fig. 1, L to N).

CD137 agonistic antibody treatment is known to increase PD-1 expression on CD8 T cells (34). Moreover, combining a PD-1 inhibitor with an αCD137 antibody exacerbates hepatotoxicity (34). Our data show that PD-1 was expressed on both CD11c⁺ and CD11c⁻T_E populations (Fig. 1K). To investigate how PD-1 blockade enhances αCD137 antibody–induced liver injury, we examined intrahepatic CD8 T cells following treatment with αCD137 and αPD-1 antibodies. Similar to NKG2A blockade, combining αPD-1 inhibition with αCD137 agonist, but not αPD-1 alone, led to ALT elevation and increased accumulation of CD11c⁺T_E in the liver (Fig. 1, O and P). Thus, these data suggest that NKG2A and PD-1 inhibitors exacerbate hepatotoxicity induced by αCD137 antibody, possibly by facilitating the accumulation of CD11c⁺T_E.

Circulating CD11c⁺T_E cells are eliminated in SLOs

Following αCD137 immunotherapy, CD11c⁺CD8 T cells were detected across multiple tissues, including the spleen, LNs, lungs, and tumors (52–55). To determine whether these extrahepatic CD11c⁺CD8 T cells exhibit an effector phenotype similar to their intrahepatic counterparts (fig. S2A), we assessed KLRG1 expression on these cells in the blood and nonliver-draining SLOs. As in the liver, most of CD11c⁺CD8 T cells coexpressed KLRG1 (Fig. 2, A and B), with further subdivision into T_SLE and T_DPE subsets based on CD127 expression (fig. S2B). In addition to the CD11c⁺ subset, CD11c⁻T_E cells were also detected in the blood and SLOs (Fig. 2, A and B). This accumulation of effector CD8 T cells was accompanied by marked enlargement of SLOs, resulting in both splenomegaly and lymphadenopathy (fig. S2, C and D). These findings suggest that T_E cells not only infiltrate the liver but also circulate in the blood and accumulate in SLOs. To investigate whether blood circulating T_E cells traffic to SLOs and the liver, we adoptively transferred CD8 T cells from αCD137 antibody–treated donor mice (CD45.1) into naïve recipients (CD45.2) and examined their distribution in peripheral tissues (Fig. 2C). Five days after intravenous injection, donor-derived CD11c⁺T_E and CD11c⁻T_E cells accumulated in the spleen, inguinal lymph nodes (iLNs), and liver (Fig. 2D), confirming their traffic from the blood to these tissues.

Fig. 2. αCD137 antibody–induced CD11c⁺T_E cells are eliminated in the spleen and LNs via apoptosis.

(A, B, and E) C57BL/6 mice were treated with αCD137 antibody or isotype control as indicated in Fig. 1A (two independent experiments). (A) CD11c and KLRG1 expression on CD8 T cells in the liver, blood, spleen, and iLNs. (B) Frequency of indicated CD8 T cell subsets in the blood, spleen, and iLNs. Blood data are representative, whereas spleen and iLN data are pooled from two independent experiments. (C and D) Naïve C57BL/6 mice (CD45.2) were intravenously injected with 1 × 10⁸ splenocytes from congenic mice (CD45.1) treated with αCD137 antibody as shown in (C) (one experiment). (C) Experimental schematic. (D) Accumulation of donor-derived CD45.1⁺T_E cells in the spleen, iLNs, liver, and blood. (E) Representative histogram of annexin V staining of CD11c⁺T_E and CD11c⁻T_E cells in the spleen, iLNs, liver, and blood. The total fraction of CD8 T cells from control mice was used as a reference (solid black line). (F to I) C57BL/6 mice were treated with αCD137 antibody and FTY720 or vehicle and analyzed on day 13. (one experiment). (F) Experimental schematic. (G) Frequency of CD11c⁺T_E and CD11c⁻T_E in the inguinal and mesenteric LNs. (H) Frequency of the indicated CD8 T cell subsets in the blood. (I) Number of CD11c⁺T_E and CD11c⁻T_E in the liver. P values are as indicated by two-way ANOVA plus Sidak’s multiple comparisons test. The gating strategy for CD8 T cell subsets is shown in Fig. 1E. mLNs, mesenteric lymph nodes. The cartoon mouse was created in BioRender. Korchagina, A. (2026) https://BioRender.com/tssh5ms.

We next examined changes in circulating T_E cells during and after αCD137 treatment. Both CD11c⁺ and CD11c⁻ subsets peaked in the blood around day 14, followed by a marked decline (fig. S2E). Consistent with this, CD11c⁺T_E and partially CD11c⁻T_E cells were substantially reduced in the liver, spleen, and iLNs at day 33 compared with day 14 (fig. S2, F to H). These data demonstrate a clear expansion phase of T_E cells, peaking around day 14, followed by a contraction phase characterized by a marked reduction in both subsets. To determine where CD11c⁺T_E cells undergo contraction, we assessed their apoptosis rate in these tissues. Strong annexin V staining was detected on live (nuclear unstained) CD11c⁺T_E in the spleen, iLNs, and liver, indicating early apoptosis of these cells in these tissues (Fig. 2E and fig. S2I). In contrast to the SLOs, annexin V staining of CD11c⁺T_E remained low in the peripheral blood, suggesting that CD11c⁺T_E apoptosis was induced by the tissue microenvironment.

Although the liver is a well-established site for the clearance of infiltrating T cells, the contribution of SLOs to the elimination of T_E cells remains largely unknown. To investigate whether the lymphoid tissue microenvironment promotes the elimination of effector CD8 T cells, we examined the survival of CD11c⁺T_E cells sequestered in the LNs after FTY720 treatment (Fig. 2F). In contrast to the increased accumulation typically observed for naïve T cells under egress blockade, we found a reduction in CD11c⁺T_E cells in the inguinal and mesenteric LNs following FTY720 treatment (Fig. 2G). These findings suggest that LN environments promote the elimination of CD11c⁺T_E cells. Moreover, these cells were absent from the circulation, including blood and spleen, and failed to accumulate in the liver (Fig. 2, H and I, and fig. S2J), indicating that blood traffic is necessary for the intrahepatic accumulation of CD11c⁺T_E cells.

Last, we tested whether NKG2A or PD-1 signaling contributes to the elimination of αCD137 antibody–induced CD11c⁺T_E within SLOs. NKG2A blockade increased the accumulation of CD11c⁺T_E in iLNs, whereas PD-1 inhibition promoted their accumulation in the spleen (fig. S2, K and L). Together, these findings indicate that SLOs promote contraction of αCD137 antibody–induced CD11c⁺T_E and that the NKG2A and PD-1 pathways regulate their persistence within lymphoid tissues.

LTβR signaling in FRCs promotes the circulation and intrahepatic accumulation of CD11c⁺T_E cells

Given that LTβR signaling regulates lymphoid tissue microarchitecture (49, 50) and promotes CD8 T cell–mediated pathology in various inflammatory conditions (36–40), we investigated its role in shaping the CD8 T cell response following αCD137 immunotherapy. To test this, we used mice with inducible LTβR deletion (iLTβR^Δ) (50), which exhibit normal LN development and fewer immune abnormalities, compared with complete LTβR knockout mice (49). Following αCD137 antibody treatment iLTβR^Δ mice showed a significant reduction in CD11c⁺T_E cells in the blood, liver, spleen, and iLNs (Fig. 3, A to E, and fig. S3A), including both T_SLE and T_DPE subsets (fig. S3B), compared with wild-type controls . In contrast, the overall CD8 T cell fraction, the CD11c⁻T_E subset, and the size of the spleen and LNs remained unchanged (Fig. 3, D and E, and fig. S3, A, C, and D), indicating that LTβR signaling deficiency predominantly reduced the CD11c⁺T_E population. Notably, both untreated iLTβR^Δ and wild-type mice exhibited very low numbers of CD11c⁺KLRG1⁺CD8 T cells across all examined organs (Fig. 3, B, D, and E, and fig. S3A) with no substantial differences between groups, suggesting that the observed phenotype is specifically induced by αCD137 antibody treatment.

Fig. 3. LTβR on FRCs promotes the accumulation of CD11c⁺T_E cells in the liver and SLOs.

(A to E) iLTβR^Δ and littermate LTβR^f/f mice were treated with αCD137 antibody or isotype control and analyzed on day 14, (A). Data were pooled from three independent experiments. (A) Experimental schematic. (B) Frequency of the CD11c⁺ fraction among CD8 T cells in the liver, spleen, and iLNs. (C) CD11c and KLRG1 expression in CD8 T cells from the liver, blood, spleen, and iLNs. [(D) and (E)] Frequency of CD11c⁺T_E and CD11c⁻T_E in the liver, blood, spleen, and iLNs. (F and G) Irradiated LTβR^−/− and LTβR^f/f control mice were reconstituted with 10⁷ bone marrow (BM) cells from LTβR^f/f donors, treated with αCD137 antibody, and analyzed on day 14 (one experiment). (F) Experimental schematic. (G) Frequency of CD11c⁺T_E and CD11c⁻T_E in the liver, blood, and spleen of LTβR^−/− and LTβR^f/f recipient mice. (H to J) CCL19^ΔLTβR and littermate LTβR^f/f mice were treated with αCD137 antibody and analyzed on day 13, as indicated in fig. S3G. Data are representative of two independent experiments. (H) Frequency of the CD11c⁺ fraction among CD8 T cells from the liver, blood, and spleen. (I) Frequency of CD11c⁺T_E and CD11c⁻T_E in the liver, blood, and spleen. (J) Number of CD11c⁺T_E and CD11c⁻T_E in the iLNs. P values are as indicated by two-way ANOVA plus Tukey’s [(B), (D), and (E)] or Sidak’s multiple comparisons test [(G) to (J)]. The cartoon mouse was created in BioRender. Korchagina, A. (2026) https://BioRender.com/tssh5ms.

We further sought to elucidate which LTβR-expressing cells are critical for the accumulation of CD11c⁺CD8 T cells following αCD137 immunotherapy. Previous studies have shown that LTβR is broadly expressed by stromal, epithelial, endothelial, and myeloid cells, but not by T and B lymphocytes or natural killer cells under homeostatic conditions (56–58). We therefore examined LTβR expression on splenic myeloid populations in the context of αCD137 immunotherapy. LTβR expression remained unchanged in neutrophils and classical type 1 dendritic cells (CD8α⁺CD11b⁻), whereas it was modestly up-regulated in macrophages and classical type 2 dendritic cells (CD8α⁻CD11b⁺) (fig. S3, E and F). As expected, LTβR expression was not detected on T or B cells.

To determine whether LTβR signaling in these myeloid subsets contributes to αCD137 immunotherapy–induced accumulation of CD11c⁺T_E cells, we used mice with conditional LTβR deletion in dendritic cells (CD11c^ΔLTβR) and in macrophages and neutrophils (LysM^ΔLTβR) (fig. S3G). We observed no significant changes in the total CD11c⁺ fraction in either CD11c^ΔLTβR or LysM^ΔLTβR mice (fig. S3, H and I). However, LysM^ΔLTβR mice exhibited a reduction in CD11c⁺T_DPE cells, a minor subset within the broader CD11c⁺T_E population (fig. S3J). These findings suggest that LTβR signaling in myeloid cells is largely dispensable for the accumulation of most CD11c⁺CD8 T cells following αCD137 immunotherapy.

To test whether LTβR signaling in nonhematopoietic cells contributes to the accumulation of CD11c⁺CD8 T cells, we used a bone marrow (BM) transfer approach. Lethally irradiated LTβR^−/− recipients were reconstituted with LTβR-sufficient hematopoietic cells (LTβR^f/f → LTβR^−/−) (Fig. 3F). These chimeric mice, which lack LTβR expression on radioresistant cells, including stromal cells, but retain LTβR expression in immune cells, exhibited reduced frequencies of circulating, intrahepatic, and splenic CD11c⁺T_E cells, along with a decrease in CD11c⁻ effector cells (Fig. 3G). These findings support a role for LTβR signaling in radioresistant cells in promoting T_E cell accumulation following αCD137 immunotherapy.

We next aimed to identify which LTβR-expressing radioresistant cells are required for CD11c⁺T_E accumulation. Given the established role of LTβR on FRCs in promoting CD8 T cell responses (41), we hypothesized that LTβR signaling in FRCs is necessary for αCD137 immunotherapy–induced accumulation of CD11c⁺T_E cells. To test this, we generated mice with selective inactivation of LTβR in FRCs by crossing LTβR^f/f mice (59) with CCL19-Cre transgenic mice (41). CCL19^ΔLTβR mice exhibited a 60% reduction in CD11c⁺T_E cells across multiple tissues, including the liver, blood, spleen, and iLNs (Fig. 3, H to J). These results indicate that LTβR signaling in FRCs promotes the circulation and accumulation of CD11c⁺T_E cells in the liver and SLOs following αCD137 immunotherapy.

B cell lymphotoxin promotes the circulation and intrahepatic accumulation of CD11c⁺T_E cells

LTβR interacts with two ligands: lymphotoxin (LT), whose functional form is a membrane-anchored heterotrimer (LTα1β2), and LIGHT (TNFSF14), which is present in both membrane-bound and soluble homotrimer forms (46, 60, 61). To determine which LTβR ligand contributes to the CD8 T cell response induced by αCD137 immunotherapy, we used mice with complete deletion of LTβ (LTβ^−/−) and LIGHT (LIGHT^−/−) (Fig. 4A). Similar to LTβR-deficient mice, LTβ^−/− mice exhibited a marked reduction in the CD11c⁺T_E subset in the liver and spleen, with a less pronounced effect on the CD11c⁻T_E subset (Fig. 4B). In contrast, LIGHT⁻/⁻ mice showed no reduction in CD11c⁺T_E cells (fig. S4A). These results indicate that LT, but not LIGHT, drives αCD137 antibody–induced accumulation of CD11c⁺T_E cells.

Fig. 4. B cell LTβ promotes αCD137 antibody–induced accumulation of CD11c⁺T_E cells.

Mice were treated with αCD137 antibody or an isotype control and analyzed on day 13Fig. 4A. (A) Experimental schematic. (B) Frequency of CD11c⁺T_E and CD11c⁻T_E cells in the liver and spleen from LTβ^−/− mice (pooled from two independent experiments). (C) Representative histogram of LTβ expression in the indicated cell populations in the liver and spleen from an LTβ^f/f mouse treated with αCD137 antibody. Solid black line, secondary anti-rabbit antibody control. (D to G) CD19^ΔLTβ and littermate LTβ^f/f mice were treated as described in (A). Data were pooled from two independent experiments. (D) Frequency of the CD11c⁺ fraction among CD8 T cells in the liver, spleen, and iLNs. (E) CD11c and KLRG1 expression on CD8 T cells in the liver. (F) Number of CD11c⁺T_E and CD11c⁻T_E cells in the liver. (G) Frequency of CD11c⁺T_E and CD11c⁻T_E in the liver, spleen, and iLNs. (H to J) CD19^ΔLTβ and littermate LTβ^f/f mice were injected subcutaneously (sc) with 5 × 10⁶ MC38 cells and treated with αCD137 antibody once the mean tumor volume reached 150 mm³ (one experiment). LTβ^f/f mice treated with αCD137 antibody were used as tumor-free controls. (H) Experimental schematic. (I) Number of CD11c⁺T_E and CD11c⁻T_E cells in the liver, nontumor-draining iLNs, spleen, and (J) tumor. P values are as indicated by two-way ANOVA plus Tukey’s [(B), (D), (F), (G), and (I)] or Sidak’s multiple comparisons test (J). AF 647, Alexa Fluor 647. The cartoon mouse was created in BioRender. Korchagina, A. (2026) https://BioRender.com/tssh5ms.

Next, we sought to identify the cellular source of LTβ. LTβ is primarily expressed by lymphocytes, including T and B cells, under homeostatic conditions (60, 62) and after CD137 stimulation (Fig. 4C and fig. S4, B to D). To determine the specific contributions of LTβ from T and B cells to the accumulation of CD11c⁺T_E cells following αCD137 immunotherapy, we used conditional knockout mice lacking LTβ in both CD8 and CD4 T cells (CD4^ΔLTβ) or in B cells (CD19^ΔLTβ). Notably, CD19^ΔLTβ mice exhibited a marked reduction in CD11c⁺T_E in the liver, spleen, and iLNs (Fig. 4, D to G). In contrast, CD4^ΔLTβ mice showed no notable differences (fig. S4, E and F), indicating that LTβ from B cells, but not T cells, is required for the accumulation of CD11c⁺T_E. These findings are consistent with previous reports demonstrating a dominant role for B cell–derived LTβ in regulating antiviral CD8 T cell responses and maintaining lymphoid tissue architecture (44, 62).

To further investigate the role of LTβ- LTβR signaling in a tumor-bearing setting, mice were implanted subcutaneously with MC38 colon adenocarcinoma cells and treated with an αCD137 antibody (Fig. 4H). Similar to results observed in nontumor-bearing mice, tumor-bearing CD19^ΔLTβ mice exhibited a reduction in CD11c⁺T_E cells across multiple tissues, including the liver, tumor-draining and nontumor-draining iLNs, spleen, tumor tissue, blood, and lung, compared to control mice, whereas the reduction in CD11c⁻T_E cells was less pronounced (Fig. 4, I and J, and fig. S4, G to J). Thus, tumor development itself did not alter T_E cell accumulation. Consistent with changes in total T_E subsets, IFN-γ–producing CD11c⁺T_E and CD11c⁻T_E cells were also reduced in CD19^ΔLTβ mice (fig. S4K). CD19^ΔLTβ mice showed similar tumor growth under αCD137 immunotherapy compared to LTβ^f/f controls (fig. S4, L and M). These results collectively suggest that B cell-LT promotes systemic accumulation of CD11c⁺T_E in response to αCD137 immunotherapy.

LTβR signaling reduces the contraction of circulating CD11c⁺T_E cells in the spleen

To further investigate how LTβ-LTβR signaling promotes the accumulation of CD11c⁺T_E cells, we first examined whether the reduction of these cells in LTβ- and LTβR-deficient mice results from an altered lymphoid microenvironment or from intrinsic defects in CD8 T cells. To address this, we transferred CD8 T cells from lymphotoxin-deficient donors into LTβR-sufficient T cell-deficient (TCRβδ^−/−) mice and treated the recipients with an αCD137 antibody (fig. S5A). CD8 T cells from LTβ^−/− mice generated a CD11c⁺ subset normally in LTβR-sufficient recipients (fig. S5B), indicating that the LTβR-deficient microenvironment, rather than intrinsic T cell defects, is primarily responsible for the impaired accumulation of CD11c⁺T_E cells.

Next, we investigated whether the reduced accumulation of CD11c⁺T_E cells in LTβ-deficient mice results from impaired proliferation. To test this, we used CD19^ΔLTβ mice and assessed proliferation during the active expansion phase (day 8) after αCD137 immunotherapy (Fig. 5A). At this time point, CD19^ΔLTβ mice already exhibited reduced CD11c⁺ cells in the liver and spleen, with a similar trend in the blood (Fig. 5B). However, Ki-67 expression in CD11c⁺T_E cells was similar across all analyzed tissues (Fig. 5C), suggesting that their proliferative capacity was not impaired. In contrast, apoptosis analysis revealed a trend to increased annexin V staining in CD11c⁺T_E cells from the spleen of CD19^ΔLTβ mice, but not from the liver or inguinal LNs, indicating enhanced cell death specifically in the spleen (Fig. 5D).

Fig. 5. LTβR signaling inhibits the contraction of blood-circulating CD11c⁺T_E cells.

(A to D) CD19^ΔLTβ and littermate LTβ^f/f mice were treated with αCD137 antibody and analyzed on day 8 as indicated (one experiment). (A) Experimental schematic. (B) Frequency of the CD11c⁺ fraction among CD8 T cells. (C) Frequency of Ki-67 expression among CD11c⁺T_E cells. (D) Median fluorescence intensity (MFI) of annexin V staining in CD11c⁺T_E cells from the blood, spleen, iLNs, and liver. (E to K) Naïve LTβR^−/− or LTβR^f/f mice (CD45.2) were intravenously transferred with 10⁸ pooled splenocytes from congenic mice (CD45.1) treated with αCD137 antibody (one experiment). (E) Experimental schematic. (F) Change in the frequency of the CD11c⁺ subset among CD45.1⁺CD8 T cells in the blood. The initial frequency of the CD11c⁺ subset in the transferred CD45.1⁺CD8 T cells is indicated on day 0. (G) Frequency of the CD11c⁺ fraction among CD45.1⁺CD8 T cells on day 5. (H) Frequency of CD8 T cell subsets expressing the indicated markers among CD45.1⁺ cells on day 5. The gating strategy is shown in Fig. 1E. (I) Annexin V staining in CD11c⁺ and CD11c⁻ subsets of CD45.1⁺CD8 T cells from the spleen and blood on day 5. Dashed black line, FMO based on indicated CD8 T cell subsets from an LTβR^f/f mouse. (J) Frequency and (K) median fluorescence intensity (MFI) of annexin V staining in CD11c⁺ and CD11c⁻ subsets of CD45.1⁺CD8 T cells in the spleen and blood on day 5. Frequency is calculated among CD45⁺ cells (B), CD11c⁺T_E (C), CD45.1⁺CD8 T cells [(F) and (G)], CD45.1⁺ cells (H), and CD11c⁺ and CD11c⁻ subsets of CD45.1⁺CD8 T cells (J). P values are as indicated by two-way ANOVA plus Sidak’s multiple comparisons test. The cartoon mouse was created in BioRender. Korchagina, A. (2026) https://BioRender.com/tssh5ms.

To test whether LTβ-LTβR signaling regulates the elimination of effector CD8 T cells, we examined the persistence of CD11c⁺T_E in LTβR-deficient environment by transferring congenically marked preactivated CD45.1⁺CD8 T cells (fig. S5C) into LTβR^−/− or LTβR^f/f recipient mice (CD45.2) (Fig. 5E). LTβR^−/− recipients exhibited a reduction in circulating donor-derived CD11c⁺cells and decreased accumulation of these cells in the liver and spleen (Fig. 5, F and G), mainly affecting the CD11c⁺T_E subset (Fig. 5H and fig. S5, D and E). In addition, LTβR^−/− recipients showed an increased annexin V staining of donor-derived CD11c⁺CD8 T cells in the spleen but not in the blood (Fig. 5, I to K, and fig. S5F). Consistently, Ki-67 expression by circulating CD11c⁺ cells was similar between LTβR^−/− and control LTβR^f/f mice (fig. S5, G to I). These results suggest that LTβR deficiency promotes the contraction of αCD137 antibody–induced CD11c⁺T_E by facilitating their apoptosis in the spleen microenvironment, thereby reducing their circulation and trafficking to the liver.

To determine whether LTβR activation reduces apoptosis of CD11c⁺T_E cells, we administered an agonistic αLTβR antibody (clone ACH6) together with an αCD137 antibody to LTβ⁻/⁻ mice (fig. S5J). Wild-type (LTβ^f/f) and LTβ^−/− mice treated with αCD137 antibody alone served as controls. LTβR stimulation reduced annexin V staining of splenic CD11c⁺T_E cells and increased their presence in the blood of LTβ^−/− mice (fig. S5, K and L), restoring apoptotic levels and cell circulation toward those observed in wild-type mice. However, accumulation of CD11c⁺T_E cells in the spleen and liver was not substantially increased (fig. S5M), consistent with the modest effect observed in the blood. Collectively, these results demonstrate that LTβR signaling limits apoptosis of CD11c⁺T_E cells within the splenic microenvironment.

To investigate splenic alterations associated with increased apoptosis of CD11c⁺T_E cells in LTβR-deficient mice, we examined changes in splenic microarchitecture following αCD137 immunotherapy. αCD137 treatment disrupted B cell follicles and caused the loss of CD169⁺ and SIGN-R1⁺ marginal zone macrophages in both wild-type and iLTβR^Δ mice (fig. S6A). Similar structural changes have been attributed to LTβR deficiency (50), suggesting that αCD137 treatment may partially disrupt LTβR signaling. In addition, iLTβR^Δ mice exhibited unexpected structural abnormalities, including marked infiltration of F4/80⁺ macrophages and increased staining with the ER-TR7 antibody (fig. S6, A to C), which recognizes type VI collagen (63), a marker predominantly associated with red pulp stromal cells in the spleen (64). These markers are typically associated with red pulp macrophages and stromal elements normally restricted to the red pulp. Together, these findings indicate a profound disruption of white pulp organization, with features suggestive of a shift toward a red pulp-like architecture in LTβR-deficient mice.

LTβR blockade abrogates hepatotoxicity induced by combined αCD137 and αPD-1 immunotherapy

Given that liver-infiltrating CD11c⁺T_E cells are a main source of IFN-γ, we investigated whether therapeutic blockade of LTβR could prevent hepatotoxicity induced by combined αCD137 and αPD-1 immunotherapy. To test this, C57BL/6 mice were administered with LTβR-Fc, a soluble LTβR antagonist, along with αCD137 agonistic and αPD-1 blocking antibodies and analyzed on day 12 (Fig. 6A). LTβR-Fc treatment reduced serum ALT and aspartate aminotransferase (AST), intrahepatic IFN-γ expression, and immune cell infiltration in the liver (Fig. 6, B to D).

Fig. 6. LTβR antagonist prevents αCD137/PD-1 antibody–induced hepatotoxicity by reducing the accumulation of IFN-γ⁺CD11c⁺T_E cells in the liver.

C57BL/6 mice were treated with αCD137 agonistic and αPD-1 antagonistic antibodies, with or without the LTβR inhibitor, and analyzed on day 13. Control mice received an isotype control. Data were pooled from three independent experiments. (A) Experimental schematic. (B) Serum ALT and AST levels. (C) Relative mRNA expression of Ifng in the liver. (D) H&E staining of liver sections showing perivascular and parenchymal infiltrates. Scale bars, 200 μm. (E) Frequency and cell number (log₁₀ transformed) of CD8 T cells in the liver. (F) Uniform Manifold Approximation and Projection (UMAP) analysis of intrahepatic CD8 T cells based on flow cytometry data (n = 4). (G) Histograms showing CD11c and KLRG1 expression in intrahepatic CD8 T cells. (H) Frequency of CD11c⁺ and CD11c⁻ fractions of CD8 T cells in the liver. (I) Flow cytometry plots of CD11c and KLRG1 expression in intrahepatic CD8 T cells. (J) Frequency of CD11c⁺T_E and CD11c⁻T_E in the liver. (K) Representative histogram of IFN-γ expression in CD8 T cells from the liver. Dashed black line, FMO control. (L) Frequency and number of IFN-γ–producing CD11c⁺ and CD11c⁻ subsets of CD8 T cells in the liver. (M) Frequency of the cumulative fraction of naïve and memory CD8 T cells (CD127⁺KLRG1⁻CD11c⁻) in the liver. P values are as indicated by one-way ANOVA [(B), (C), (E), and (M)] or two-way ANOVA plus Tukey’s multiple comparisons test [(H), (J), and (L)]. The cartoon mouse was created in BioRender. Korchagina, A. (2026) https://BioRender.com/tssh5ms.

Analysis of immune cell composition in the liver revealed a reduction in CD8 T cells (Fig. 6E), whereas the numbers of Kupffer cells, B cells, CD4 T cells, and neutrophils were not significantly affected (fig. S7A). LTβR-Fc treatment primarily reduced the total CD11c⁺ fraction and its major T_E subsets (Fig. 6, F to J, and fig. S7, B and C), whereas CD11c⁻ populations, including the T_SLE subset, remained unchanged (fig. S7D). These findings suggest that LTβR blockade predominantly targets CD11c⁺T_E cells, consistent with our observations in mice with inducible inactivation of LTβR (Fig. 3, A to E).

Furthermore, LTβR-Fc administration selectively reduced the accumulation of IFN-γ–expressing CD11c⁺CD8 T cells, whereas the abundance of IFN-γ⁺ CD11c⁻ cells was unaffected (Fig. 6, K and L). LTβR-Fc treatment increased the accumulation of IFN-γ⁻ CD8 T cell populations in the liver, including CD127⁺KLRG1⁻ naïve and memory subsets (Fig. 6M). Consistent with these cellular changes, expression of additional cytotoxic T cell–associated genes, including GzmB, GzmK, and Prf1, was reduced in the liver following LTβR-Fc treatment (fig. S7E).

In addition, LTβR blockade reduced the accumulation of CD11c⁺T_E cells in the spleen, iLNs, and blood (fig. S7, F and G), indicating a systemic effect on this population. Gene expression analysis in the spleen revealed partial preservation of Il7 and a trend toward reduced Cxcl13 expression following LTβR-Fc treatment. In contrast, no significant changes were observed in expression of chemokines (Cxcl9, Cxcl10, Ccl19, and Ccl21), inhibitory molecules (Pdl1, Pdl2, Fasl, Lgals9, Ido1, Nos2, and Arg1), and cytokines (S100a4, Csf1, Ifng, Tgfb, Il2, Il6, Il10, Il15, Il18, Il21, Ifnb, Ifna2, Ifna4, Il23a, Il27p28, Il12a, Il12b, and Tgfb) (fig. S7H). Collectively, these findings indicate that LTβR blockade mitigates αCD137/αPD-1 antibody–induced hepatotoxicity by preventing the circulation and subsequent accumulation of IFN-γ–producing CD11c⁺T_E cells in the liver.

LTβR blockade preserves the antitumor efficacy of combined αCD137 and αPD-1 immunotherapy

We investigated the impact of LTβR blockade on therapeutic efficacy using a subcutaneous murine MC38 colon adenocarcinoma model, previously validated for combined αCD137 and αPD-1 immunotherapy (Fig. 7A) (65). Consistent with findings in tumor-free mice, the LTβR antagonist reduced serum ALT levels and decreased circulating CD11c⁺T_E, including T_SLE and T_DPE subsets, compared with mice treated with αCD137 and αPD-1 combination therapy (Fig. 7, B and C, and fig. S8A). Notably, despite a significant reduction in effector CD8 T cells, LTβR blockade did not impair immunotherapy-induced tumor suppression during or after treatment (Fig. 7, D to G, and fig. S8B). Across two independent experiments, complete tumor rejection was observed in 6 of 15 mice per treatment group (fig. S8B). Moreover, follow-up tumor rechallenge demonstrated that LTβR blockade preserved durable antitumor memory responses in these mice (fig. S8, C and D). These findings indicate that LTβR blockade did not impair the antitumor efficacy of combined αCD137 and αPD-1 immunotherapy.

Fig. 7. LTβR blockade does not impair the antitumor efficacy of αCD137/PD-1 immunotherapy in MC38 tumor–bearing mice.

(A to G) C57BL/6 mice with subcutaneous MC38 tumors received treatment with αCD137 and αPD-1 antibodies with or without LTβR-Fc. The control group received an isotype control. The data are representative of two independent experiments. (A) Experimental schematic. (B) Serum ALT on day 9. (C) Frequency of CD11c⁺T_E and CD11c⁻T_E in the blood on day 13. [(D) and (E)] Mean ± SEM and individual tumor growth curves in male mice (n = 8). P values are indicated on days 18 and 21. Arrows mark injection days. A total of 5 × 10⁶ MC38 cells were implanted, and treatment was initiated when the mean tumor volume reached 120 mm³. [(F) and (G)] Mean ± SEM and individual tumor growth curves in female mice (n = 6 to 7). P values are indicated on days 25, 27, and 31. A total of 2 × 10⁶ MC38 cells were implanted, and treatment was initiated when the mean tumor volume reached 70 mm³. (H to P) MC38-OVA tumor–bearing C57BL/6 mice were treated with αCD137 antibody with or without LTβR-Fc.. The control group received an isotype control. (H) Experimental schematic. (I) Frequency of total and (J) tetramer⁺CD8 T cells. (K) Frequency of CD11c⁺T_E and CD11c⁻T_E. (L) CD11c expression in total (top) and tetramer⁺ CD8 T cells (bottom). (M) Frequency of the CD11c⁺ fraction among total and (N) tetramer⁺ CD8 T cells. (O) Frequency of total CD11c⁺CD8 T cells and (P) tetramer⁺CD11c⁺CD8 T cells. Frequency is calculated among live CD45⁺ cells [(O) and (P)], total CD8 T cells (M), or tetramer⁺ CD8 T cells (N). P values are indicated by one-way ANOVA [(B), (I), (J), and (M) to (P)] or two-way ANOVA plus Tukey’s multiple comparisons test [(C), (D), (F), and (K)]. The gating strategy is shown in fig. S7B. ns, not significant. The cartoon mouse was created in BioRender. Korchagina, A. (2026) https://BioRender.com/tssh5ms.

Because the antitumor effects of αCD137 immunotherapy are mediated by CD8 T cells (23, 34), we next sought to determine the effect of LTβR blockade on tumor-specific CD8 T cells following αCD137 immunotherapy. To address this, we used an ovalbumin (OVA)–expressing MC38 tumor model and analyzed OVA-specific CD8 T cells using the SIINFEKL-H2-K^b tetramer (Fig. 7H and fig. S8E). LTβR-Fc treatment did not reduce either tetramer⁺ or total CD8 T cells in the tumor compared to αCD137 antibody alone (Fig. 7, I and J) but specifically decreased CD11c⁺T_E cells and the total CD11c⁺ fraction (Fig. 7, K to O). Notably, the CD11c⁺ subset among tetramer⁺ CD8 T cells remained unaffected (Fig. 7, L, N, and P), indicating that LTβR blockade did not reduce tumor-specific CD8 T cells. These data further suggest that most of CD11c⁺CD8 T cells induced by αCD137 immunotherapy do not substantially contribute to antitumor immunity. Given that the αCD137 antibody induces antigen-independent expansion of CD8 T cells (8, 27), our findings indicate that most CD11c⁺CD8 T cells represent a bystander population that expands excessively and contributes to hepatotoxicity rather than tumor control. Therefore, LTβR blockade prevents hepatotoxicity induced by αCD137 immunotherapy by promoting the contraction of bystander CD11c⁺T_E cells while preserving tumor-specific CD8 T cell responses required for effective tumor suppression (fig. S9).

DISCUSSION

In this study, we show that blocking the LT-LTβR pathway can uncouple toxicity from the antitumor effects of αCD137 immunotherapy. This strategy promotes bystander CD8 T cell contraction without compromising tumor-specific responses. Our data reveal a previously unknown role for LTβR signaling in FRCs in controlling the elimination of circulating effector CD8 T cells in the spleen. These results highlight the critical role of SLOs in moderating CD8 T cell–mediated immunopathology and identify the LT-LTβR axis as a key regulator of this process.

Our data suggest that LTβR regulates the elimination of CD8 T cells, primarily the CD11c⁺ subset. Expression of CD11c (integrin αX) on CD8 T cells has been previously linked to effector differentiation, IFN-γ production, and responses to various immune challenges, including immunization, infections, and tumor development (52, 53, 66–69). CD11c up-regulation is thought to promote T cell cytotoxicity and migration (70, 71). Consistent with these observations, we found that CD11c⁺ CD8 T cells coexpress KLRG1 and exhibit high levels of effector-associated molecules, including granzyme B and the transcription factor T-bet. In humans, CD11c⁺CD8 T cells have been identified in immune disorders (72–74), suggesting a potential link between CD11c up-regulation and immunopathology. Our mouse data demonstrate that the CD11c⁺CD8 T cells are the major source of IFN-γ in the liver, and their accumulation correlates with exacerbated hepatotoxicity following treatment with αCD137, αPD-1, and αNKG2A antibodies. Given that αCD137 immunotherapy–induced hepatotoxicity is driven by CD8 T cells (23) in an IFN-γ–dependent manner (30), our findings suggest a key role for CD11c⁺T_E cells in liver pathology.

Although the liver is a well-established site for lymphocyte clearance (75), our data reveal a key role for SLOs in removing effector CD8 T cells after αCD137 immunotherapy. Both the spleen and LNs exhibited strong annexin V staining in CD11c⁺T_E cells, and pharmacological blockade of lymphocyte egress with FTY720 reduced rather than increased accumulation of these cells in LNs. These findings suggest that the SLO microenvironment not only traps dying effector cells but may also actively promote their apoptotic elimination. However, because S1PR1 signaling itself has been implicated in the regulation of T cell survival and apoptosis (76), a direct suppressive effect of FTY720 on CD11c⁺T_E cells cannot be excluded. Our observations of CD8 T cell apoptosis within SLOs are consistent with previous reports showing that activated CD8 T cells display reduced survival and enhanced apoptosis in these tissues (77–79). Moreover, splenectomy results in elevated levels of circulating effector CD8 T cells, further supporting the role of the spleen in their clearance (80).

One potential mechanism underlying this elimination may involve inhibitory receptors expressed on CD11c⁺T_E, such as PD-1, NKG2A, Tim-3, and Fas. Concurrently, expression of their corresponding ligands, FasL and Lgals9, was up-regulated in the spleen following αCD137 immunotherapy, whereas programmed cell death 1 ligand 1 (PD-L1) and PD-L2 levels remained unchanged, likely due to the high level of constitutive expression in this tissue (81). PD-1 and NKG2A blockade led to increased accumulation of CD11c⁺T_E in the spleen and LNs, respectively. These findings suggest a potential role for inhibitory signaling in the contraction of CD11c⁺T_E cells. Notably, the apoptotic rates of CD11c⁺T_E and CD11c⁻T_E cells were comparable in the liver but differed in the spleen and iLNs, where the CD11c⁺ subset exhibited higher apoptosis. This increased susceptibility may be partially explained by elevated expression of inhibitory receptors such as NKG2A on CD11c⁺T_E cells. However, contributions from additional, unexamined proapoptotic or prosurvival pathways cannot be excluded.

In addition, αCD137 stimulation up-regulated the expression of chemokines CXCL9 and CXCL10 in the spleen, which are known to influence effector CD8 T cell differentiation and limit their exposure to survival niches via the CXCR3 axis (82, 83). CXCL9 and CXCL10 are expressed in lymphoid tissues by FRCs, dendritic cells, and macrophages, which regulate T cell responses and maintain immune homeostasis through chemokine production as well as survival and inhibitory signals (84–89). Notably, αCD137-induced transcriptional changes were accompanied by structural remodeling of LNs, characterized by expansion of ER-TR7⁺ stromal cells and F4/80⁺ macrophages (90). We also observed pronounced structural alterations in the spleen, particularly disruption of B cell follicles and loss of CD169⁺ and SIGN-R1⁺ marginal zone macrophages. Thus, our findings suggest that SLOs foster a proapoptotic environment for differentiated effector CD8 T cells, facilitating their recruitment and elimination, likely through the production of inhibitory ligands by FRCs and macrophages.

Our findings identify the LT-LTβR axis as a regulator of the circulation and tissue accumulation of effector CD8 T cells in all examined organs, with a preferential effect on the CD11c⁺T_E subset. This global reduction supports a model in which LTβR signaling promotes effector CD8 T cell survival during circulation and transit through the spleen. In the absence of LTβR signaling, CD11c⁺T_E cells undergo accelerated contraction, thereby limiting their accumulation in peripheral tissues such as the liver. Our data argue against alternative mechanisms, such as phenotypic conversion or CD11c down-regulation. Specifically, we did not observe a reciprocal increase in other activated CD8 T cell populations that would be expected if CD11c⁺T_E cells broadly down-regulated CD11c and reemerged as CD11c⁻T_E or memory subsets. Instead, CD11c⁻T_E cells remained unchanged or were similarly reduced in LTβ- or LTβR–deficient mice, consistent with enhanced cell death rather than phenotypic conversion. Nevertheless, we cannot formally exclude the possibility that CD11c⁺T_E cells may down-regulate CD11c and transition into memory subsets, nor whether such processes are influenced by LTβR signaling. These important questions were not directly addressed in the present study and warrant further investigation.

Our data further identify the LT-LTβR axis as a regulator of effector CD8 T cell apoptosis in the spleen. We demonstrate that LTβR signaling in FRCs and LTβ expression on B cells inhibit the contraction of circulating CD11c⁺T_E cells by delaying cell death in the spleen. Although LTβR in stromal and myeloid cells is known to support CD8 T cell responses via type I IFN production (42, 91–93), our data indicate that αCD137 immunotherapy does not induce type I IFN expression in the spleen, and LTβR inhibition does not modify this response. Instead, LTβR-Fc treatment restored the expression of Il7, a prosurvival cytokine for T cells in lymphoid tissue (94). Paradoxically, this was accompanied by a reduction in CD11c⁺T_DPE cells, a subset that expresses IL-7Rα (CD127), suggesting that their loss is driven by mechanisms independent of IL-7 availability. In contrast to the minimal effects on cytokine expression, iLTβR^Δ mice exhibited pronounced alterations in splenic microarchitecture, characterized by increased accumulation of F4/80⁺ macrophages and enhanced ER-TR7 staining—markers normally restricted to the red pulp. These changes indicate a remodeling of the white pulp toward a red pulp-like organization in iLTβR^Δ mice. Both FRCs and red pulp macrophages are known to inhibit T cell responses (86, 95, 96). Accordingly, given the well-established role of LT-LTβR pathway in FRC differentiation and the maintenance of lymphoid tissue microarchitecture (51, 62, 97–99), we propose that LTβR blockade enhances the exposure of effector CD8 T cells to proapoptotic factors by altering their migration within the spleen or by concentrating inhibitory signals within the white pulp. Specifically, LTβR inhibition may reprogram FRCs, leading to the recruitment of red pulp macrophages into the white pulp, where they facilitate T_E cell apoptosis. Although the immunoregulatory benefits of LTβR inhibition were described in several CD8 T cell–mediated immunopathological settings (36–40), our results provide further insights into LTβR-dependent regulation of effector CD8 T cell responses during the contraction phase. However, the precise mechanism by which LTβR signaling inhibits the death of T_E cells in the spleen remains to be clarified.

Our data indicate that targeting LTβR with an LTβR-Fc antagonist prevents αCD137 antibody–induced hepatotoxicity by reducing the accumulation of IFN-γ–producing CD11c⁺T_E cells while preserving tumor-specific CD8 T cells and their antitumor activity. Although questions remain regarding the role of major histocompatibility complex (MHC) class I in liver tissue damage (24), αCD137 antibody–induced CD8 T cell expansion is antigen independent (8, 27). This expansion was shown to occur independently of MHC class I and requires CD137 expression on CD8 T cells (8). In the same study, antigen-experienced OT-I CD8 T cells exhibited markedly enhanced proliferation in response to CD137 stimulation compared to naïve OT-I cells, indicating that memory, but not naïve, CD8 T cells are the primary responders to CD137 engagement. Based on these findings, we propose that in nontumor-bearing mice, αCD137 treatment predominantly drives the expansion of CD137⁺ memory and virtual memory CD8 T cells in an antigen/TCR-independent manner. αCD137 immunotherapy has been reported to enhance the survival of tumor-specific CD8 T cells without inducing extensive clonal expansion (35). Thus, αCD137 stimulation likely promotes robust proliferation of nontumor-specific memory CD8 T cells while inducing only modest expansion of tumor-specific clones. This model provides a mechanistic explanation for why the substantial reduction in total effector CD8 T cells observed upon LTβR blockade does not compromise antitumor efficacy: Tumor-specific CD8 T cells represent a smaller population that undergoes slower expansion-contraction dynamics and is therefore less susceptible to accelerated elimination in the spleen. These observations also support our conclusion that hepatotoxicity in tumor-bearing mice is not primarily driven by tumor-specific CD8 T cells. In this context, our results suggest that LTβR blockade preferentially accelerates the elimination of rapidly expanding bystander CD11c⁺ effector CD8 T cells.

LTβR has a context-dependent role in cancer, either promoting (100–105) or suppressing tumor progression (106, 107). For example, LTβR signaling can enhance the antitumor response by promoting the development of tertiary lymphoid structures (TLSs) and high endothelial venules, thereby facilitating immune cell infiltration, activation, and lymphocyte entry into tumors (108–111). Consistent with this, LTβR stimulation induces antitumor immunity in TLS-forming Colon26 tumors when used as monotherapy, and enhances the efficacy of αCTLA-4 treatment in the subcutaneous MC38-OVA model, which does not spontaneously form intratumoral TLS (112). Despite these reported benefits of LTβR activation in the MC38 model, our data demonstrate that LTβR blockade does not compromise the overall therapeutic efficacy of combined αCD137 and αPD-1 treatment, nor does it reduce the accumulation of tumor-specific CD8 T cells in this model. This suggests that LTβR signaling is not essential for therapeutic benefits in the context of αCD137 immunotherapy, likely because the broad immune stimulation induced by CD137 engagement compensates for any potential LTβR-dependent antitumor effects. However, given the heterogeneity in tumor responses to LTβR activation, even within the same tumor type, the impact of LTβR inhibition on αCD137-based therapy should be validated in additional tumor models. In addition, the impact of systemic LTβR blockade on immune response, in combination with other immune checkpoint inhibitors, should also be validated. Conversely, in specific contexts, LTβR blockade may enhance antitumor responses by suppressing regulatory T cells and immunosuppressive macrophages (113, 114), suggesting that it may also enhance the efficacy of immunotherapy. Future studies are needed to define the tumor types and immune landscapes in which targeting LTβR signaling is beneficial to immunotherapy and where it is dispensable. Moreover, follow-up investigations should explore whether LTβR signaling limits CD8 T cell contraction in other T cell–driven immunopathologies and assess the translational potential of this strategy in human disease.

Based on our findings, we propose a model in which LTβR blockade uncouples the immunopathology associated with αCD137 antibody treatment from its antitumor efficacy (fig. S9). αCD137 immunotherapy induces a massive expansion of predominantly nontumor-specific T_E, which circulate in the blood and traffic to the liver, leading to IFN-γ–mediated damage. In this context of excessive expansion, SLOs function as contraction hubs, eliminating circulating effector CD8 T cells by inducing apoptosis. The apoptotic fate of circulating effector CD8 T cells passing through the spleen is regulated by LTβR signaling in FRCs, which engage LTβ expressed by B cells. Under sustained CD137 stimulation, homeostatic LTβR signaling in FRCs limits apoptosis and prolongs the persistence of CD11c⁺T_E cells in circulation, thereby facilitating their trafficking to the liver and inducing immunopathology. In contrast, LTβR blockade limits hepatotoxicity by accelerating the contraction of circulating CD11c⁺T_E cells within the spleen, thereby restricting their accumulation in the liver. Notably, tumor-specific CD8 T cells remain unaffected by LTβR blockade, likely reflecting their more moderate expansion and slower contraction dynamics within the spleen. Thus, targeting the LTβR pathway preferentially accelerates the contraction of rapidly expanding, largely bystander effector CD8 T cells while preserving tumor-specific clones following αCD137 immunotherapy. Collectively, these findings indicate that promoting effector CD8 T cell contraction within the spleen via LTβR blockade offers a promising approach to manage αCD137-induced hepatotoxicity and, more broadly, CD8 T cell–mediated immunopathology while preserving efficient antitumor immunity.

MATERIALS AND METHODS

Study design

This study aimed to investigate the role of LTβR signaling in regulating CD8 T cell responses and liver immunopathology during αCD137 antibody immunotherapy. We used genetic approaches, including constitutive and conditional deletion of LTβR and its ligands LTβ and LIGHT. Spectral flow cytometry was used to analyze CD8 T cells. To evaluate liver pathology, serum ALT levels, histological analysis, and gene expression profiling were performed. To study the elimination of effector CD8 T cells, we used the adoptive transfer of αCD137-activated CD8 T cells from congenic donor mice. Pharmacological inhibition of LTβR signaling was tested using an LTβR-Fc fusion protein in tumor-free and tumor-bearing mice treated with αCD137 and αPD-1 blockade to assess its therapeutic potential in mitigating toxicity. Experiments in this study were performed at least three times unless otherwise stated in the figure captions.

Mice

All animal studies were conducted in accordance with the University of Texas Health Science Center at San Antonio (UTHSCSA) Institutional Animal Care and Use Committee. All mice used in this research were housed under specific pathogen–free conditions in accordance with National Institutes of Health (NIH) guidelines. C57BL/6 (JAX #000664), TCRβδ^−/− (JAX #002122), LysM-Cre (JAX #004781), CD11c-Cre (JAX #007567), CD4-Cre (JAX #022071), CD19-Cre (JAX #006785), B6 CD45.1 (JAX #002014), and GREAT [IFN-γ–internal ribosomal entry site–enhanced yellow fluorescent protein (eYFP), JAX #017581] were purchased from the Jackson Laboratory (Bar Harbor) and bred at the UTHSCSA. LTβR^f/f (59), LTβ^f/f (62), LTβR^−/− (59), LTβ^−/− (115), LIGHT^−/− (116), CCL19-Cre (41), and Cre-ERT2 LTβR^f/f (iLTβR^Δ) (50) mice were described previously. To induce Cre recombination in Cre-ERT2 LTβR^f/f mice, tamoxifen was administered orally in corn oil containing 10% ethanol (5 mg/kg per day) for 4 consecutive days via oral gavage 2 weeks before use in experiments. CCL19-Cre LTβR^f/f (CCL19^ΔLTβR), CD11c-Cre LTβR^f/f (CD11c^ΔLTβR), LysM-Cre LTβR^f/f (LysM^ΔLTβR), CD19-Cre LTβ^f/f (CD19^ΔLTβ), and CD4-Cre LTβ^f/f (CD4^ΔLTβ) were generated by crossing LTβR^f/f or LTβ^f/f mice with corresponding Cre-transgenic mice. CCL19^ΔLTβR and CD19^ΔLTβ mice have been previously characterized (41, 62). Age- and sex-matched littermate LTβR^f/f or LTβ^f/f mice were used as wild-type controls. Eight- to 20-week-old male and female mice were used in the experiments.

In vivo interventions in mice

For αCD137 antibody–induced hepatotoxicity, mice were treated intraperitoneally with αCD137 agonistic antibody (clone 3H3, 3 mg/kg) or isotype control [rat immunoglobulin G2a (IgG2a), clone 2A3, 3 mg/kg] in sterile Dulbecco’s Phosphate-Buffered Saline (DPBS) on days 0, 4, 8, and 11, unless otherwise indicated. αNKG2A/C/E inhibitory antibody (clone 20D5, 5 mg/kg), αPD-1 inhibitory antibody (clone RMP1-14, 5 mg/kg), and LTβR inhibitor (LTβR-Fc, 3.5 mg/kg) (58) were administered intraperitoneally on the same days. Agonistic αLTβR antibody (clone ACH6, 3.5 mg/kg) (117) was administered intraperitoneally on days 4 and 8 following αCD137 treatment. To block lymphocyte egress, FTY720 (0.5 mg/kg) in saline was injected intraperitoneally daily from day 5 to day 12 after initiation of αCD137 antibody treatment.

Mouse tissue processing

Single-cell suspensions from perfused livers were prepared by mechanical dissociation through a 200-μm steel mesh, followed by centrifugation using 40% isotonic Percoll (Cytiva). Residual erythrocytes were removed by treatment with Ammonium-Chloride-Potassium (ACK) lysis buffer. Spleens and LNs (inguinal and mesenteric) were mechanically disrupted using 40-μm nylon cell strainers to generate single-cell suspensions. For blood processing, samples were centrifuged to separate the serum and cellular fractions. Erythrocytes were then lysed from the cell fraction using ACK lysis buffer. Cell number and viability were assessed using acridine orange and propidium iodide staining on an RWD-C100 automated cell counter (RWD Life Science). Tumor tissues were finely minced using a scalpel and digested in Hanks’ balanced salt solution containing collagenase D (0.5 mg/ml; Roche) and deoxyribonuclease I (40 U/ml; Sigma-Aldrich) for 60 min at 37°C on an orbital shaker.

Flow cytometry and cell sorting

Single-cell suspensions were first incubated with anti-CD16/32 antibody (20 μg/ml; clone 2.4G2) to block Fc receptors, followed by staining with Zombie near-infrared (NIR) fixable viability dye (BioLegend) to distinguish live and dead cells. Cells were then incubated with fluorescent dye–conjugated MHC class I tetramer and/or antibodies in DPBS containing 2% fetal bovine serum (FBS). The tetramer was obtained from the NIH Tetramer Core Facility, and antibodies were purchased from BioLegend and BD Biosciences. For cell sorting, CD8 T cells were sorted on a BD FACSAria II (BD Biosciences) using a 70-μm nozzle. For LTβ detection, cells were stained with an unconjugated polyclonal anti-LTβ rabbit antibody (Boster Bio), followed by a secondary allophycocyanin (APC)-conjugated polyclonal anti-rabbit IgG antibody. For intracellular staining, cells were fixed and permeabilized using the True-Nuclear Transcription Factor Buffer Set (BioLegend) according to the manufacturer’s protocol and then incubated with antibodies. To assess IFN-γ production, isolated intrahepatic cells were stimulated for 4 hours at 37°C with the Cell Activation Cocktail (BioLegend), which includes ionomycin, phorbol 12-myristate 13-acetate, and brefeldin A. After stimulation, cells were stained for surface markers, fixed, and stained intracellularly for IFN-γ. Apoptosis was analyzed by staining with phycoerythrin-conjugated annexin V (1:200) and Helix NP NIR (a DNA binding dye, 20 nM) to distinguish between live and dead cells. Flow cytometry data were acquired using an Aurora spectral cytometer (Cytek Biosciences) and analyzed with FlowJo v10 software. Uniform Manifold Approximation and Projection (UMAP) analysis of flow cytometry data was performed using OMIQ software. Antibodies are listed in table S1. Cell frequencies were calculated among live CD45⁺ or CD8 T cells.

Histological analysis

Tissues were fixed in 10% neutral buffered formalin (Thermo Fisher Scientific), paraffin-embedded, sectioned, and stained with hematoxylin and eosin (H&E) for histological evaluation. For immunofluorescence, fresh-frozen tissues embedded in Optimal Cutting Temperature (OCT) compound were cryosectioned using a CryoStar NX70 cryostat (Thermo Fisher Scientific). Sections were fixed in 4% paraformaldehyde in phosphate-buffered saline for 30 min at room temperature, stained with fluorophore-conjugated antibodies and 4′,6-diamidino-2-phenylindole, and imaged with a Keyence BZ-X800 microscope.

RNA extraction and quantitative reverse transcription PCR

RNA from tissue stabilized with RNAlater (Thermo Fisher Scientific) was extracted using the E.Z.N.A. Total RNA Kit I (Omega Bio-tek), while RNA from sorted cells was isolated using the RNeasy Micro Kit (QIAGEN) according to the manufacturer’s protocols. cDNA synthesis and real-time reverse transcription polymerase chain reaction (PCR) were performed as described previously (58) using Power SYBR Green master mix (Applied Biosystems) on a C1000 Touch Thermal Cycler (Bio-Rad). Relative mRNA expression of the target genes was determined using the comparative 2^−ΔΔCt method. Gene expression level was normalized to Hprt. Primer sequences are listed in table S2.

Aminotransferase activity

Serum ALT and AST activities were measured using Liquid ALT and AST Reagent Sets (Pointe Scientific) following the manufacturer’s instructions using a DS-11 Spectrophotometer (DeNovix).

Bone marrow chimeric mice

To generate BM chimeras, LTβR^−/− or LTβR^f/f mice were lethally irradiated with a fractionated dose of 2 × 500 cGy 4 hours apart, using a Cesium-137 source. A total of 10⁷ BM cells collected from the femurs and tibiae of LTβR^f/f mice were injected intravenously into the irradiated recipients on the same day.

Adoptive cell transfer of CD8 T cells

For total splenocyte transfer, 1 × 10⁸ splenocytes per recipient mouse were administered via intravenous injection. For purified CD8 T cell transfers, CD8 T cells were isolated from splenocytes by negative magnetic selection using the EasySep Mouse CD8 T Cell Isolation Kit (STEMCELL Technologies) according to the manufacturer’s protocol.

Tumor cell lines

MC38 cells, mouse colorectal carcinoma cells, and OVA-expressing MC38 cells (MC38-OVA) were provided by T. Curiel. Cells were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% FBS and penicillin-streptomycin at 37°C in a humidified atmosphere containing 5% CO₂. For passaging, cells were detached using 0.05% trypsin-EDTA (Thermo Fisher Scientific).

Tumor growth

C57BL/6 mice were subcutaneously injected in the right flank with 5 × 10⁶ or 2 × 10⁶ MC38 tumor cells. Seven days later, when tumors reached a mean size of 70 to 150 mm³, mice were randomized into treatment groups and administered the indicated therapies. The control group received an isotype control on the same days. Tumors were measured with a caliper (Scienceware), and volume was calculated using the formula: V = (L × W²/2), where W is the tumor width and L is the tumor length. Mice were euthanized when tumors reached 1500 mm³. To study the tumor-specific CD8 T cell response, C57BL/6 mice were subcutaneously injected with 2 × 10⁶ OVA-expressing MC38 tumor cells (MC38-OVA). When tumors reached 400 to 500 mm³, mice were randomized in a stratified manner based on tumor size and treated with αCD137 antibody and LTβR-Fc.

Bulk RNA sequencing and differential gene expression analysis

C57BL/6 mice received αCD137 antibody on days 0, 3, and 6, and single-cell suspensions were isolated from the livers on day 8. Intrahepatic CD11c⁺T_SLE and CD11c⁻T_SLE cells were sorted using a BD FACSAria cell sorter (n = 4 per group). Total RNA from eight sorted cell pools was processed at the Greehey Children’s Cancer Research Institute (GCCRI) Genome Sequencing Facility, and the quality of RNA samples was assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies, CA). RNA sequencing libraries were prepared according to the SMART-seq2 protocol (118), with the following modifications: PCR preamplification to 10 to 12 cycles, two rounds of beads cleanup with a 1:1 ratio after cDNA synthesis, and 0.6 to 0.8 dual beads cleanup for Nextera XT DNA-seq library purification. The sequencing libraries were pooled, and 50-bp single-read sequencing was performed on the HiSeq 3000 system (Illumina, CA).

The sequence reads were aligned to the mouse reference transcript (build mm9) using TopHat2 (v2.0.8b), with default parameters. The alignment (BAM) files were processed with HTSeq to quantify gene expression as expected counts and normalized expression as RPKM (reads per kilobase of transcript per million mapped reads). The read counts were further analyzed using the R/Bioconductor-based DESeq (v1.36.0) package to identify differentially expressed genes across different cell sorting criteria. We considered differentially expressed genes with significant (P < 0.05) by at least twofold change (i.e., |log₂ fold change| ≥ 1). Further analysis for gene ontology enrichment or gene set enrichment analysis was performed using the R package topGO and fgsea (R/Bioconductor), respectively.

Quantification and statistical analysis

All statistical analyses were performed using GraphPad Prism versions 9 and 10. P values were determined using Student’s t test, one-way analysis of variance (ANOVA), or two-way ANOVA with Sidak’s or Tukey’s multiple comparisons test, as appropriate. Statistical significance was set at P < 0.05. The specific tests used are indicated in the figure legend. P values are shown on the graphs. At least three biological replicates (e.g., independent animals) were included in each group for analysis. Data are presented as mean ± SD, unless otherwise noted. Each symbol on the graphs represents an individual mouse. No data or subjects were excluded from the analysis. When data from multiple independent experiments were pooled, entire datasets were included without post hoc exclusion or selective inclusion.

Supplementary Materials

This PDF file includes:

Figs. S1 to S12

Tables S1 and S2