T-αFGL2 treatment has limited antitumor cytotoxic T lymphocyte activity in vitro
To obtain superior FGL2 specific monoclonal antibodies (mAbs), we selected 75 clones of FGL2 mAbs through 3 independent hybridoma fusion and via an ELISA-based mouse FGL2 (mFGL2) binding assay (Supplementary Fig. 1a). 13 out of 75 clones showed strong binding activity to mFGL2 but not to his-tag. Human FGL2 (huFGL2) was then used to select mAbs that show bi-species binding reactivity (Supplementary Fig. 1b), and mouse FGL2 binding clone #4 also showed high binding affinity to human FGL2 (Supplementary Fig. 1a, 1b). Additionally, clone #4 showed the most linear association between binding capacity and dilution (Supplementary Fig. 1c). Western blotting, immunofluorescence staining, and ELISA assay further validated the binding activity of FGL2 mAb-clone #4 to both mouse and human FGL2 (Supplementary Fig. 1d-1f). To further test the effect of FGL2 blocking scFv, lentiviral constructs derived from FGL2 mAb-clone #4 were generated to arm T cells (Fig 1a). This construct contained scFv domains that aimed at recognizing and binding FGL2 (Fig 1a). To ensure that FGL2 scFv was expressed on the surface of T cells with the ability of movement, an EGFR-transmembrane (TM) domain was linked to the FGL2 scFv by a P2A linker (Fig 1a, 1b). The expression of FGL2 scFv on the T cell membrane was validated by the staining of the His tag domain (Fig 1c). The transduction efficiency of the activated mouse T cells was consistent and in the range of 15% to 25% (Fig 1c). To verify that T-αFGL2 cells can directly bind FGL2, we have established a microfluidics chip binding assay. As the data shown in Supplementary Fig.1g, T-αFGL2 cells can directly bind the FGL2 that is anchored on the chip via biotin-streptavidin covalent bond. The antitumor cytotoxic T lymphocyte activity of FGL2-scFv–armed T cells (T-αFGL2) against FGL2-expressing DBT cells, a mouse GBM cell line, was evaluated by measuring proportion of tumor cells, and granzyme B, interferon γ (IFNγ), and tumor necrosis factor α (TNFα) positive T cells (Fig. 1d). T-αFGL2 cells expressed higher granzyme B levels than did T cells transfected with a control construct (T-Ctr) when cocultured with DBT cells at an effector-to-target ratio of 1:1. However, no significant difference was found in the proportion of tumor cells when cocultured with T-αFGL2 and T-Ctr cells, and T-αFGL2 and T-Ctr cells had comparable levels of IFNγ and TNFα, suggesting that T-αFGL2 may have limited direct tumor cell killing activity effects in vitro.
T-αFGL2 treatment does not cause toxicity in immunocompetent mice
To evaluate the suitability of FGL2 as a target for T cell therapy with low risk of off-tumor on-target toxicity, we assessed the expression of FGL2 in human GBM tissues and normal human tissue arrays using FGL2 mAb-clone #4 from which the αFGL2 construct was derived. As shown in Fig. 1e, FGL2 was highly expressed in human GBM tissues but not in healthy medulla oblongata tissues. In healthy tissue arrays (Fig. 1e), major organs such as the brain, lung, breast, spleen, and muscle were FGL2 negative, while moderate expression of FGL2 was observed in the stomach, colon, and pancreas. (Fig. 1e). To assess the potential toxicity of T-αFGL2, we intravenously injected 5 million T-Ctr or T-αFGL2 cells into non-tumor-bearing 7-week-old immunocompetent Balb/c mice. Five days after T cell injection, we evaluated blood chemistry, organ toxicities, and immune cell populations in the spleen and bone marrow. As shown in Supplementary Fig. 2a and b, mice treated with T-αFGL2 exhibited no significant changes in immune cell counts in either the spleen or bone marrow. T-αFGL2 treatment caused no abnormalities in blood chemistry (Supplementary Fig. 2c), but mice treated with T-Ctr had significantly higher blood levels of albumin (P = 0.0264) and globulin (P = 0.0181) than did untreated mice (Supplementary Fig. 2c). Furthermore, a board certified veterinary pathologist (N.W.F.) observed no evident abnomality or aberrant T lymphocyte infiltration in tissue sections following T-αFGL2 cell infusion (Supplementary Fig. 3 and Supplementary Table 1). Taken together, these results show that T-αFGL2 therapy does not cause detectable organ toxicity in immunocompetent mice.
T-αFGL2 therapy induces superior antitumor activity in vivo
To test the efficacy of T-αFGL2 therapy in vivo, we first validated expression of FGL2 in mouse GBM tissue. Brain tissues from immunocompetent syngeneic mouse GBM model (DBT tumor-bearing mice) were cryosectioned and stained with FGL2 mAb-clone #4. As shown in Fig. 2a, both tumor cells and surrounding stroma were positively stained for FGL2. Next, DBT-bearing Balb/c mice were used to evaluate the antitumor effects of T-αFGL2. Mice were inoculated with tumor cells and then treated with standard chemotherapy temozolomide (TMZ) on days 3, 4, and 5 to assimilate standard care, before administering T-Ctr or T-αFGL2 cells via the tail vein on days 6 and 13 after tumor cell inoculation (Fig. 2b). DBT is a very aggressive GBM tumor, and most DBT-bearing mice died within 3 weeks in the no treatment or T-Ctr group. T-αFGL2 treatment effectively suppressed DBT tumor progression, and tumors were eliminated in about 30% of the T-αFGL2–treated mice. These mice remained tumor free for up to 70 days before being used for a rechallenge study. In contrast, tumors progressed rapidly in T-Ctr–treated mice (Fig. 2c-e).
To confirm the anti-tumor efficacy of T-αFGL2 treatment, we also took advantage of another syngeneic mouse GL261 model. As shown in Fig. 2f-g, T-αFGL2 treatment, compared with T-Ctr treatment, suppressed this GBM tumor growth and extended mouse survival. Overall, T-αFGL2 showed superior antitumor properties in syngeneic malignant brain tumor models.
T-αFGL2 treatment induces formation of tumor-specific CD8+TRM like cells in the brain
We next evaluated whether long-term survivors that had been treated with T-αFGL2 cells developed memory T cells that were reactive to tumor cells. We rechallenged T-αFGL2–treated survivors with an intracranial (i.c.) injection of DBT cells (Fig. 3a). The re-challenged DBT cells were cleared within 7 days in T-αFGL2–treated survivors (Fig. 3b), and local re-exposure to DBT cells induced a rapid, more than 18-fold increase in the number of CD8+ T cells in the brains of T-αFGL2–treated survivors compared to the number in naïve brains (Fig. 3d, 3f). To investigate the tumor specificity of the generated memory T cells, we rechallenged DBT tumor-rejecting mice induced by T-αFGL2 treatment with 4T1 tumor cells (i.c.), which also developed tumors in the naïve balb/c mice rapidly. We found these DBT-rejecting memory T cells failed to protect mice from 4T1 tumor cell challenge (Supplementary Fig. 4a). This rapid and intense tumor reactivity to DBT cells, but not to 4T1 cells, confirmed that tumor-specific memory CD8+ T cells had developed in the brains of T-αFGL2–treated survivors.
To determine whether these tumor-specific memory CD8+ T cells stayed in the vicinity of the tumor (ie, in the brain) or migrated throughout the body, we inoculated DBT cells subcutaneously into the flanks of T-αFGL2 survivors and naïve mice (Fig. 3a). Interestingly, both groups of mice developed tumors under the skin (Fig. 3c), suggesting that the tumor-specific memory CD8+ T cells in T-αFGL2–treated survivors were restricted to the brain. To confirm that tumor-reactive CD8+ T cells only existed in the brain, we assessed T cells from the brains and draining lymph nodes (dLNs) of naïve mice and T-αFGL2–treated survivors 7 days after the rechallenge with DBT cells. As shown in Fig. 3d and e, the ratio of CD8+ T cells to CD4+ T cells in the brain was up to 8-fold higher in T-αFGL2–treated survivors than in naïve mice. Moreover, the ratio of CD8+ to CD4+ T cells was 9-fold higher in the brains than in the LNs of T-αFGL2–treated survivors, suggesting that CD8+ T cells, but not CD4+ T cells, were the primary memory T cells controlling tumor cell growth, and that these CD8+ T cells were only resident in the brain. Taken together, these data strongly indicate that T-αFGL2 treatment induced development of brain-resident tumor-specific CD8+TRM like cells.
CD8+TRM like cells undergo recall expansion and reject tumor cells when being transplanted into naïve brains
To validate that CD8+ T cells in the brains of T-αFGL2–treated survivors were CD8+TRM cells, we sorted CD8+ T cells from the brains, draining lymph nodes (dLNs), and peripheral blood (PB) of T-αFGL2–treated survivors on day 7 after tumor cell inoculation and adoptively transplanted these T cells along with DBT cells directly into the brains (intracranially) of naïve recipient mice (Fig. 4a). In contrast to both dLN and PB CD8+ T cells, which failed to mount a recall response, brain CD8+ T cells underwent expansion even when reseeded in the brain tissue in low numbers (3000 cells) (Fig. 4b, c), confirming that the CD8+ T cells in the brains of T-αFGL2–treated survivors are bona fide TRM cells. As most of the CD8+T cells in tumor experienced brains were CD44+ memory T cells, while the CD8+T cells in peripheral were not, we then compared the anti-tumor effect of CD44+CD8+T cells in brain, dLNs, pLNs and PB to validate the results. In consistence with CD8+T cells, the CD44+CD8+T cells in peripheral tissues didn’t show protection against tumor cells in vivo (Supplementary Fig. 4b and 4c). Similar results was found in GL261 model (Supplementary Fig. 4d). To determine whether CD4+ T cells in the brain behaved in a similar manner as CD8+ T cells, we sorted CD4+ and CD8+ T cells from the brains of T-αFGL2–treated survivors and then co-inoculated with DBT cells into the brains of naïve recipient mice. As shown in Fig. 4b and c, CD4+ T cells did not have the same tumor-cell-eliminating capacity as CD8+ T cells, confirming that the induced brain resident CD8+ T cells, but not CD4+ T cells, in the brain provide immune surveillance of the previously encountered antigen.
To further investigate whether the adoptively transplanted CD8+ TRM cells could survive and remain in the brains of naïve recipient mice, we subsequently challenged the recipient mice with tumor cells on day 40 after adoptive CD8+ T cells transplantation. We observed that the recipient mice rejected the rechallenged tumor cells (Fig. 4d, e). These findings show that CD8+TRM cells were successfully transplanted into naïve brains, transforming naïve brains to become tumor rejecting brains. To further validate the function of transplanted CD8+TRM cells and exclude the effect of host T cells, we transplanted the lymphocytes from brain with TRM (TRM-BILs) into brain of naïve immunodeficient SCID mice and re-challenged these mice 35 days after the transplantation (Fig. 4f). The same as the parental TRM bearing mice, these TRM transplanted SCID mice showed anti-tumor capacity in the brain (Fig 4f-g); To verify the transplanted CD8+T cells are responsible for the protection, we re-challenged these SCID survivors with tumor cells combined with αCD8, αCD4 or αsialo GM1 antibodies to deplete CD8+T cells, CD4+T cells and NK cells respectively. Only depleting CD8+T cells, but not depleting CD4+T cells or NK cells, impaired this protection (Fig 4h and 4i). Taken together, CD8+TRM like cells, in brains of T-αFGL2–treated survivors, which fulfill both memory and reactive functions against tumor cells, are tumor specific brain CD8+TRM cells. Notably, these CD8+ TRM cells can be adoptively transferred into the naïve brains with or without host T cells.
CD8+TRM cells establish classical TRM phenotype
To determine whether these CD8+TRM like cells bear classical TRM phenotype, we checked CD69, CD103 and CD62L expression, which were used to identify TRM1,10,24. To verify that the isolated BILs-CD8+T cells were brain restricted, we performed i.v. injection of the CD8β antibody and found that over 90% of BILs-CD8+T cells were non-circulating brain resident T cells (Supplementary Fig. 5a). Here, we found that compared with CD44+CD8+T cells in PB, the CD44+CD8+T cells in TRM bearing brains with TRM were CD69+ (either CD103+ or CD103-), and CD62L-(Fig. 4j). Similar results were found for CD4+ T cells (Supplementary Fig. 5b). These findings show that CD8+TRM like cells in brains established a classical TRM phenotype of CD69+CD62L-. Together, both function and phenotype of the CD8+TRM like cells in brains of T-αFGL2–treated survivors further validated that they are tumor specific brain CD8+TRM cells.
The function of CD8+TRM cells is TCR-MHC-I dependent
Since TCR is generated through random rearrangement of genomic V(D)J segments and is the mediator of antigen recognition and binding by T lymphocytes, we next questioned whether CD8+TRM cells displayed a unique TCR repertoire that was distinct from that found in the dLNs. To this end, we sorted CD44+CD8+ T (memory CD8+T) cells from the brains and dLNs of T-αFGL2–treated survivors on day 20 after the third challenge with DBT cells via flow cytometric sorting, followed by TCRα and TCRβ deep sequencing (Fig. 5a). The most abundant T cell clones—those with a frequency of more than 5%—in TRM-bearing brains constituted more than 60% of the total TCRα and TCRβ repertoire, whereas no T cell clones with a frequency of more than 5% were found in the TCRβ repertoire of CD44+CD8+ T cells in the dLNs (Supplementary Fig. 5c). To further characterize the TCR repertoires of TRM cells and dLNs-CD44+CD8+ T cells, we analyzed sequences of complementarity determining region 3 (CDR3), which encompasses the V(D)J recombination junctions and encodes the vast majority of TCR variation. Moreover, all of the top 10 dominant CDR3 sequences in TRM cells encompassed the V/J recombination, but each dominant CDR3 sequence in dLNs-CD44+CD8+ T cells encompassed a unique V/J recombination (Fig. 5b and Supplementary Fig. 5d). Analysis of the V and J domain usage showed that, in one of the T-αFGL2–treated survivors, the most dominant clone of TCRβ in TRM cells was grouped by V17/J1-4, which was absent in dLNs (Fig. 5b). These data showed the presence and expansion of unique T cell clones and that there was no overlap of the highly occupied TCR clone in TRM cells with the TCR clone in dLNs-CD44+CD8+ T cells. Interestingly, each mouse of T-αFGL2–treated survivors bears different TRM clones against different antigens. These data suggested that these highly expanded TCR repertoires of CD8+ TRM cells were associated, thus, with the rapid and robust response of TRM cells against tumor cells.
To verify the robust response of CD8+TRM cells against tumor cells is associated with the interaction between expanded TCR and MHC-I, we blocked MHC-I in vivo using αMHC-I antibody when transplanted CD8+TRM cells into naïve mice. Blocking MHC-I abolished the anti-tumor efficacy of the transplanted CD8+TRM cells (Fig 5c-f), demonstrating TCR-MHC-I interaction is required for the proper function of CD8+TRM cells in vivo.
T-αFGL2 treatment-induced CD69 expression on CD8+ memory T cells is essential for CD8+TRM formation
To understand the cellular mechanisms by which the FGL2-blocking scFv induces the generation of CD8+TRM cells, on day 4 after the second T cell infusion, we performed high-dimensional profiling of brain-infiltrating lymphocytes (BILs) using time-of-flight mass cytometry (CyTOF) with a panel of 37 antibodies that illustrated different immune populations (Fig. 6a). CyTOF data analysis divided BILs into 15 immune cell populations (Fig. 6b). As CD8+ T cells are the primary functional cells rejecting tumor cells as shown in our transplant study (Fig. 4b), we focused on these cell populations. We found that the CD8+ T cell population was composed of 2 subpopulations: CD69+CD8+ memory T cells (CD69+CD8+ TM) and CD69−CD8+ memory T cells (CD69−CD8+ TM) (Fig. 6b). Notably, the subset of CD69+CD8+ TM cells was significantly larger in mice that underwent T-αFGL2 treatment than in those that received T-Ctr treatment (Fig. 6c, d). Indeed, CD69 has been reported to help in the retention of memory T cells in resident tissues through inhibiting expression of the S1P receptor, which can promote T cell circulation into the blood; high level expression of CD69 on T cells is an indicator of TRM cells24. To further determine the phenotype of these CD69+CD8+ TM cells, we compared their expression of T cell exhaustion markers with that of CD69−CD8+ TM cells. As shown in Fig. 6e, CD69+CD8+ TM cells had higher levels of Ki67, CD223 (LAG3), and CD279 (PD-1) than did CD69−CD8+ TM cells. This CD69HiPD-1HiLAG3Hi phenotype of highly proliferating CD69+CD8+ TM cells has been reported to be most prominent in cells with TRM characteristics in different kinds of tissues, including lung15,25,26, breast19, and skin6. These data indicated that T-αFGL2 treatment increased proliferating CD69+CD8+ TM subsets with TRM characteristics, which may promote the transformation of these CD69+CD8+ TM cells into TRM cells in the brain. To further validate the biological function of CD69 on CD8+TRM, we blocked CD69 in vivo by αCD69 antibody. When transplanted the CD69+CD8+TRM cells with tumor cells into naïve mice i.c., we found αCD69 antibody treatment didn’t disrupt the anti-tumor efficacy of CD8+TRM cells. However, when we re-challenge these mice with tumor cells i.c. on day 60 post the transplantation, the mice treated with αCD69 antibody lost the tumor-rejecting capacity (Fig. 6f-g), indicating that CD69 doesn’t affect the executive function of TRM but is required for the prolonged residence and function of CD8+TRM in brains.
Besides CD8+ T cells, we analyzed the other immune subpopulations, including helper T cells (Th cells) and regulatory T cells (Tregs), DCs, macrophages, monocytes, and neutrophils. However, no significant difference in these subpopulations was observed between the T-αFGL2 and T-Ctr groups (Supplementary Fig. 6), suggesting that the antitumor effect induced by T-αFGL2 treatment is different from antibody therapy and may work mainly through regulating CD69+CD8+ TM cells.
T-αFGL2 induced CD8+TRM formation is associated with CXCL9/10-CXCR3 axis
To further understand the molecular mechanism by which FGL2-blocking scFv induces CD69+CD8+TM cells generation, we analyzed the CYTOF data for the chemokine receptors (i.e. CCR2, CXCR3, CXCR2, and CX3CR1) that may affect the T cells infiltration and recruitment to tumor sites. Intriguingly, we found that T-αFGL2 treatment increased CXCR3 expression on CD69+CD8+TM cells in tumor bearing brains, compared with T-Ctr treatment (Fig. 7a). Flow cytometry data also verified that T-αFGL2 treatment, compared with T-Ctr, increased proportion of CXCR3+CD69+CD8+T cells among total CD8+T cells in tumor bearing brains (Fig. 7b), indicating that increased CXCR3 expression on CD8T cells may play a role in mediating T-αFGL2 induced CD69+CD8+TRM cells formation. To validate this notion, we compared the anti-tumor efficacy of T-αFGL2 in treating tumor bearing wild type (WT) mice and CXCR3 deficient (CXCR3-/-) mice. T-αFGL2 treatment didn’t show protective effect in CXCR3-/- mice as did in WT mice (Fig. 7c). Besides, the CD69+CD8+TM population was reduced in CXCR3-/- mice compared with WT mice (Fig. 7d), suggesting that CXCR3 play a critical role in mediating T-αFGL2 induced CD69+CD8+TM cells generation and thus CD8+TRM formation for protective function.
To characterize how the CXCR3 chemokine system mediates anti-tumor responses to T-αFGL2 treatment, we examined the expression of the CXCR3 chemokine ligands CXCL9 and CXCL10. Protein levels of CXCL9 and CXCL10 were markedly increased in tumor bearing brains after T-αFGL2 treatment compared to T-Ctr treatment (Fig. 7e). To determine the roles of CXCL9 and CXCL10 in T-αFGL2 induced tumor rejection and CD8+TRM formation, we used αCXCL9 and αCXCL10 antibodies to blocking CXCL9 and CXCL10 in vivo. Consistent with our earlier findings (Fig. 7c-d), the therapeutic benefits of T-αFGL2 was lost when blocking CXCL9 and CXCL10 (Fig. 7f), indicating a critical role for CXCL9 and CXCL10 in T-αFGL2 immunotherapy. Moreover, the percentage of CD69+CD8+TM was increased upon T-αFGL2 treatment in control IgG group but not in αCXCL9 and αCXCL10 antibodies treatment group (Fig .7g), indicating the functional importance of CXCL9/10-CXCR3 axis for response to T-αFGL2 therapy and CD8+TRM formation. Altogether, T-αFGL2 therapy induced tumor-reactive T cells’ proliferation, secretion of granzyme B to control tumor progression, and increased CXCR3, CD69 expression on memory CD8+T cells to help them retain in the brains, which will foster the formation of tumor specific brain resident CD8+TRM (Fig. 7h).