ATG5-Dependent Autophagy Uncouples T-cell Proliferative and Effector Functions and Separates Graft-versus-Host Disease from Graft-versus-Leukemia

Macroautophagy is an adaptive response that permits cell response to stress and is involved in cellular differentiation and programed cell death. Its dysfunction is associated with a broad range of diseases (1, 2). During autophagy, cells engulf intracellular organelles, cytoplasmic aggregates, and proteins in a double-membrane vesicle called an autophagosome, that subsequently fuse with the lysosome and promotes their degradation and recycling (3, 4). ATG5 plays a key role in this process of autophagy and is a critical component of the membrane extension process of the autophagosome. It is conjugated to ATG12 in a complex that catalyzes the conjugation of microtubule-associated protein 1 light chain 3 (LC3–1) to phosphatidylethanolamine (LC3-PE or LC3-II) on the extension membrane (5–10). Mice deficient in ATG5 have been shown to succumb to neuronal dysfunction (11) in vivo, but its role in immunity and specifically with in vivo T-cell responses to clinically relevant antigens remain unknown. Several other components of the autophagic machinery, besides ATG5, have been shown to affect T-cell proliferation, cytokine production, and memory cell generation (5). Whether they differentially affect the various responses, and the impact on in vivo clinically relevant noninfectious immunity remains unknown.

Allogeneic hematopoietic cell transplantation (allo-HCT) is a curative therapy against hematologic malignancies. The curative potential is realized by the donor T-cell–mediated graft versus tumor (GVT) responses after allo-HCT. However, the concomitant occurrence of T-cell dependent graft-versus-host disease (GVHD), the major complication of allo-HCT, has precluded its efficacy. Strategies that can reduce GVHD while persevering sufficient GVT would be a major advance in the field. Donor T-cell responses that cause GVHD and GVT share critical pathways including shared antigens, proliferation, and effector functions. However, whether uncoupling these shared responses, such as proliferation versus cytotoxic effector functions, might provide a window of opportunity to meaningfully separate GVHD from GVT remains unclear.

Herein we show that allo-stimulation promotes autophagic response in T cells, in vitro and in vivo. Deficiency of ATG5 in donor T cells diminished induction of autophagy, decreased proliferation, increased effector cytokine, and CTL functions after in vitro and in vivo allo-stimulation. Mechanistic studies suggest that the lack of degradation of the CD8⁺ cytotoxic molecules, such as granzyme B, in the autophagosome enhanced the cytotoxicity of the Atg5^−/− CD8⁺ T cells. This uncoupling of the proliferation from effector responses reduced GVHD but still retained sufficient GVT responses.

Rapamycin (Millipore Sigma) was reconstituted in dimethyl sulfoxide (DMSO) at a concentration of 100 mmol/L, and further dilutions were made in RPMI media. Chloroquine diphosphate salt (Millipore Sigma) was reconstituted in PBS at a concentration of 100 mmol/L, and further dilutions were made in RPMI media. 3-Methyladenine (3-MA; Millipore Sigma) was reconstituted in DMSO at a concentration of 100 mmol/L with further dilutions made in RPMI media.

C57BL/6 (B6, H-2^b) and BALB/c (H-2^d) mice were purchased from the Jackson Laboratory and Charles River Laboratories As described previously (12, 13)^. B6-background ATG5^flox/flox mice (B6.129S-Atg5<tm1Myok>, #RBRC02975, Riken RBC) were bred to Lck-Cre (B6 Cg-Tg(Lck-Cre)1Cwi N9) transgenic mice (Taconic, #004197). All animals were cared for according to regulations reviewed and approved by the University Committee on Use and Care of Animals of the University of Michigan, based on University Laboratory Animal Medicine guidelines.

Bone marrow transplantations (BMT) were performed as described previously (14, 15). Briefly, splenic T-cells from donors were enriched using the Pan T Cell Isolation Kit II and manual MACS separation with LS columns (Miltenyi Biotec Inc.). BM was depleted of T cells by manual MACS separation utilizing CD90.2 microbeads (Miltenyi Biotec). We used the well-established MHC-mismatched B6→BALB/c BMT model (16). BALB/c animals were used as recipients and received 8 Gy (¹³⁷Cs source) on day −1 and then 0.5 × 10⁶ CD90.2⁺ T cells along with 5 × 10⁶ T-cell–depleted BM (TCD-BM) cells from either syngeneic (BALB/c) or allogeneic (B6 or Atg5^flox/flox Lck-Cre⁺) donors on day 0.

We monitored survival of transplanted animals daily and assessed the degree of clinical GVHD weekly, as described previously (Supplementary Table S1; ref. 17).

Human peripheral blood was obtained via routine venipuncture according to Institutional review board protocol UMCC 2001.0234 (HUM00043287), Long-Term Evaluation of the Biology and Outcomes of Hematopoietic Stem Cell Transplantation, which collects clinical data and blood samples serially throughout the transplant course as well as at the time of significant clinical complications, such as GVHD, at our center. Alternatively, heparinized whole blood was purchased from Innovative Research. Peripheral blood mononuclear cells (PBMC) were purified from heparinized whole blood via density centrifugation with Ficoll-Paque Premium (GE Healthcare) as per manufacturer's recommendations. PBMCs were then treated with red blood cell (RBC) lysis buffer (Millipore Sigma) as per the manufacturer's instructions, and cryopreserved in heat-shocked FCS (Gibco) containing 10% DMSO (Millipore Sigma). PBMCs were thawed in a 37°C water bath and immediately mixed with complete cell media (RPMI1640, 10% HS-FBS, penicillin, streptomycin, l-glutamine) supplemented with 50 U/mL of Benzonase (EMD Millipore). Cells were rinsed once with media and then T cells were isolated via negative magnetic selection with pan T-cell kits according to the manufacturer's protocol (Miltenyi Biotec).

For mixed lymphocyte reaction (MLR), with allogeneic activation, human T cells (2 × 10⁵/200 μL/well) were cocultured with an equal amount of lethally irradiated (30 Gy) allogeneic monocytes (pool of 2–3 unrelated donors) derived from the cell fraction bound to the Pan T cell negative selection column. Human MLRs were incubated for 6–7 days in complete cell media (RPMI1640, 10% heat-shocked FBS, penicillin, streptomycin, l-glutamine, 2-mercaptoethanol). For nonspecific T-cell receptor (TCR) stimulation, T cells (2 × 10⁵/200 μL/well) were cultured with soluble 2 μg/mL anti-CD28 (clone: 37.51) and plate-bound 10 μg/mL anti-CD3ϵ (clone: 145–2C11) purified anti-mouse antibodies (BioLegend) in 96-well flat-bottom plate for 48 hours. To assess T-cell proliferation, incorporation of 3H-thymidine (1 μCi/96 wells) during the final 6 hours (CD3/CD28 stimulation) or 15 hours (allogeneic stimulation) of culture was measured by a TopCount (PerkinElmer).

For in vitro experiments comparing B6-WT and B6-Atg5^−/− T cells, T cells were first isolated from the spleens using the Pan T Cell Isolation Kit II (Miltenyi Biotec) with manual MACS separation and LS columns. To increase the purity to >95%, T cells were then positively sorted with a CD90.2-PerCP/Cy5.5 and negatively for CD25 antibody (BioLegend) with MoFlo Astrios or Synergy Head cell sorters. For nonspecific TCR stimulation, sorted T cells (2 × 10⁵/200 μL/well) were cultured with soluble 2 μg/mL anti-CD28 (clone: 37.51) and plate-bound 10 μg/mL anti-CD3ϵ (clone: 145–2C11) purified anti-mouse antibodies (BioLegend) in 96-well flat-bottom plate for 48 or 60 hours. For MLR, 2 × 10⁵ Atg5^−/− and WT-B6 T cells (2 × 10⁵/100 μL/well) were cocultured at a 20:1 ratio with BM-derived dendritic cells (BMDC) from syngeneic B6 or allogeneic BALB/c mice (1 × 10⁴/100 μL/well) in 96-well flat-bottom plate for 4 or 6 days. To obtain BMDCs, BM cells were cultured with murine recombinant GM-CSF (20 ng/mL; PeproTech Inc.) for 7 days, then DCs were purified using CD11c microbeads and the autoMACS separator (Miltenyi Biotec). Before being put in culture, BMDCs were irradiated with 20 Gy. The incorporation of ³H-thymidine (1 μCi/well; PerkinElmer) by proliferating T cells during the final 6 hours (CD3/28 stimulation) or 16 hours (MLR) of culture was measured by a β-plate reader TopCount (PerkinElmer).

Mice underwent normal transplant procedures with 1 × 10⁶ C57BL/6 90.2 donor T cells and 5.0 × 10⁶ TCD-BM cells into C57BL/6 or Balb/c recipients. Twelve to 14 days posttransplant, the recipient mice were euthanized and spleens collected. T cells were purified from either C57BL/6 syngeneic or Balb/c allogeneic mice. After initial purification, donor T cells were then isolated with H-2K^b beads (Miltenyi Biotec) from allogeneic recipients. Donor T-cell samples were pooled, pelleted, and then washed with PBS. Cells were pelleted and fixed with glutaraldehyde and taken to the Biomedical Research Core Facilities Microscopy Core for processing and sectioning onto grids.

A total of 2 × 10⁶ cells/mL/well purified T cells from B6-WT and B6-Atg5^−/− spleens were incubated with 5 × 10⁶cells/mL/well whole splenocytes previously irradiated with 30 Gy in a bulk MLR in 24-well plate to produce activated CD8⁺ effector T cells. At day 6, cells were harvested, analyzed by flow cytometry for the percentage of CD4 and CD8⁺ cells, and resuspended at a concentration of 4 × 10⁶ CD8⁺ T cells/mL. P815 (H-2^d) mastocytoma cell line derived from a DBA/2 mouse was used as target. Briefly, 2 × 10⁶ tumor cells were resuspended in 200 μL RPMI with 10% FCS and incubated for 2 hours with 12 μCi ⁵¹Cr (PerkinElmer). After being washed three times, 4,000 tumor cells were incubated in 96-well round-bottom plate with activated CD8⁺ T cells at effector-to-target ratios from 100:1 to 12.5:1 in quadruplicates. Supernatants were harvested after 4 hours and ⁵¹Cr activity was determined in an autogamma counter (Packard). Maximal and background release were determined by the addition of 4% Triton X-100 or media alone to targets, respectively. The percentage of specific lysis was calculated as following: 100 × (sample count−background count)/(maximal count−background count).

Purified T cells were washed and suspended at a concentration of 1 × 10⁶ cells/mL in PBS containing 0.5% BSA and 2 mmol/L EDTA. Carboxyfluorescein diacetate succinimidyl ester (CFSE; Thermo Scientific) in DMSO (5 mmol/L) was added to the cell suspension at a final concentration of 2 μmol/L and incubated at 37°C for 10 minutes. The staining was quenched by the addition of ice-cold RPMI medium containing 10% FCS. The cells were then washed according to the manufacturer's instructions and used for the proliferation studies. At the end of the proliferation, T cells were stained with a PE-conjugated anti-Annexin-V antibody (BD Biosciences) in the dark for 10 minutes at room temperature in labeling buffer and analyzed using Accuri C6 cytometer (Accuri).

For the staining of LC3, purified T cells previously marked with CFSE were harvested after stimulation, washed twice with FACS wash buffer (0.2% BSA in PBS), fixed with 1× BD FACSTM Lysing Solution (BD Biosciences), and permeabilized for 10 minutes at 4°C with 1× Permeabilization Buffer (eBioscience) in the presence of 1:400 rat anti-mouse FcR mAb 2.4G2 to block nonspecific FcR binding of labeled antibodies. Primary rabbit-LC3 antibody (Novus Biologicals) diluted 1:200 in Permeabilization Buffer was incubated for 30 minutes at 4°C, cells were washed once, and secondary anti-rabbit Alexa Fluor 633 goat anti-rabbit IgG antibody (Thermo Scientific) was also incubated for 30 minutes at 4°C in Permeabilization Buffer at the same concentration. Cells were then resuspended in FACS wash buffer and analyzed using Accuri C6 cytometer. In assessing autophagy in unstimulated and stimulated human T cells, the CYTO-ID Autophagy Detection Kit 2.0 (Enzo Life Sciences) was used for flow analysis using the Attune NXT cytometer (Thermo Scientific) For other staining experiments, see the Supplementary Methods.

Acquisition of all samples were stained using identical conditions and mounted in Prolong Gold (RI = 1.5). Fluorescence was imaged onto GaAsP detectors using a Nikon A1 confocal scan head and 60 × 1.4NA oil objective. Identical image acquisition settings were used for all samples. Image analysis for “colocalization” was quantified using Mander's Overlap Coefficients, k1 and k2 (Manders, J Microscopy, 1993). This approach was chosen because we hypothesized that the amount of overlap in the spatial distribution of protein concentrations would change due to treatment. Calculations were performed on z-stack images using the FIJI JACoP plugin (Bolte, J Microscopy, 2006). Background pixels were excluded from the calculation based on a manual threshold. The signals were quantified using ImageJ software with a JaCoP plugin. The JaCoP plugin takes two signals A (eGFP) and B (RFP) that it labels as “Image A” and “Image B”. The interpretation of k1 is then “the amount (measured as intensity) of Signal B that overlaps with (occupies sample pixels) Signal A”. k2 is the amount of Signal A that overlaps with Signal B.

Concentrations of IFNγ, IL2, and IL17α were measured in the culture supernatants or in the serum by ELISA with specific anti-mouse MoAbs for capture and detection utilizing OptEIA (IFNγ, IL2; BD Biosciences), ELISA MAX (IL17α; BioLegend). Assays were performed according to the manufacturer's protocol and read at 450 nm using a microplate reader (Model 3550; Bio-Rad Labs).

T cells, unstimulated or stimulated for 24 hours (mouse) or 48 hours (human) with anti-CD3 (4 μg/mL) and CD28 (2 μg/mL), with or without chloroquine (10 μmol/L) or 3-MA (5 mmol/L), were harvested and washed twice with PBS. Then whole cells were lysed in RIPA lysis extraction buffer (Thermo Scientific). Equal amounts of protein lysates quantified with the BCA Protein Assay Kit (Thermo Scientific) were separated with 10% or 12.5% SDS-PAGE. Rabbit-LC3 antibody (Novus Biologicals) was used according to the manufacturer's instructions. Secondary antibody conjugated to horseradish peroxidase (Jackson ImmunoResearch) was used to detect the primary antibody. Densitometric analysis was performed using ImageJ software.

Survival curves were analyzed with the Wilcoxon test. An unpaired t test was performed for the statistical analysis of all other data. A P value < 0.05 was considered to be statistically significant (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001).

We first analyzed T cells for induction of autophagy following allo-stimulation. Purified splenic T (T_con) cells from C57BL/6 (B6-WT) mice were harvested and stimulated for 4 days with irradiated splenic cells from an allogeneic BALB/c mouse. Unstimulated B6-WT T_con cells, rapamycin (Rapa; induces autophagy), hydroxy-chloroquine (CQ; autophagy inhibitor), and 3-Methyladenine (3-MA) treatments were utilized as controls. The inhibitors, CQ and 3-MA, regulate distinct stages of autophagy (Supplementary Fig. S1A). Specifically, CQ is commonly used as a positive control for LC3-II staining because it blocks formation of the autophosomes when the conversion to LC3-II has already taken place and autophagosomes accumulate in the cell (Supplementary Fig. S1B), while 3-MA blocks autophagy proximally, thus preventing formation of LC3-II and leading to reduction in both LC3-II and LC3-I. This accumulation of autophagosomes can be observed with CQ inhibiting autophagy during B6-WT T_con cell stimulation with allogeneic BALB/c splenocytes by EM (Supplementary Fig. S1C).

B6-WT T_con cells activated by allo-BALB/c cells demonstrated greater conversion to the autophagosome-bound LC3-II form when compared with controls (Fig. 1A, right) and seen visually by EM of increased cytoplasmic activity of activated cells (Fig. 1A, left). This increase in LC3-II observed in mouse allogeneic stimulated T_con cells was further confirmed by confocal microscopy (Fig. 1B). The increase in autophagy in T_con cells increased in a time-dependent manner (Fig. 1C). A similar trend was observed when B6-WT T_con cells were stimulated with anti-CD3/CD28 antibodies, consistent with previous reports (Supplementary Fig. S1D; ref. 18). We next determined whether inhibition of autophagy in T_con cells alters their morphologic and functional response to allo-stimulation. As shown in Supplementary Fig. S1E, blocking autophagy with increased doses of CQ reduced T_con cell proliferation in a dose- and time-dependent manner while showing a reciprocal increase in apoptosis (Supplementary Fig. S1F).

We next analyzed the induction of autophagy on activation of human T cells and further determined the impact of inhibiting autophagy on their responses in MLR. Human T cells were isolated from whole blood from 3 individual healthy volunteers and stimulated in an allogeneic MLR as in Methods. Western blot (Fig. 1D) and flow cytometry analyses (Fig. 1E) 6 days after stimulation of human T cells in an MLR caused induction of autophagy in T cells, similar to murine T cells. We next assessed the functional relevance of autophagy induction in human T cells. We utilized 3-MA and CQ to inhibit distinct stages of autophagy (Supplementary Fig. S1A) to determine whether inhibition of autophagy regulates the proliferative responses of human T cells. As shown in Fig. 1F, inhibition of autophagy with both 3-MA and CQ significantly reduced the proliferation of human T cells. Collectively, these data demonstrate that T_con cells upon in vitro stimulation by allo-APCs undergo autophagy, which when inhibited, mitigated their proliferative responses, but increased their apoptosis.

We next explored whether autophagic induction following in vitro allo-stimulation was observed in vivo. To this end, we utilized donor T_con cells from the autophagy reporter CAG-RFP-eGFP-LC3 mice, which express a tandem RFP-eGFP LC3 fusion protein (19). Autophagosomes found in the cytoplasm would express both the RFP (red) and eGFP (green) simultaneously, thus producing a yellow color. Once the autophagosomes fuse with the lysosomes, indicative of autophagy, then the eGFP emission is lost due to the acidic pH of the autophagolysosome thereby leaving only the red color (RFP) to be visible (Fig. 2A). Splenic T_con cells from these mice were mixed with TCD-BM from WT-B6 and transferred to BALB/c recipients following myeloablative conditioning as described in Methods. Congenic C57BL/6 recipients served as syngeneic controls. On days 1 and 3, recipient mice from each group were sacrificed and their spleens were harvested and analyzed for donor T cells. After magnetic bead purification and cytospin preparations, cells were imaged with a confocal microscope (Fig. 2B) to quantitate the amount of combined fluorescence signaling of eGFP plus RFP verses RFP signaling alone. The imaging and their analysis were performed in an operator-blinded manner. As shown in Fig. 2B, when compared with syngeneic controls, donor T cells harvested from allo-recipients showed significantly greater loss of eGFP, that is, increased autophagy on both days 1 and 3 (Fig. 2C). These data collectively demonstrate that autophagy is induced in T_con cells following stimulation by allo-antigens in vitro and in vivo.

To study the impact of the loss of autophagy, specifically in T_con cells, we developed an Atg5 conditional knockout mouse (B6-Atg5^−/−; Supplementary Fig. S2A). We confirmed deficiency of ATG5 in T cells and also concurrently assessed the impact of loss of ATG5 on induction of autophagy in T cells following allo-stimulation. T_con cells from B6-WT and B6-Atg5^−/− animals were harvested and stimulated with splenic cells from BALB/c mice for 4 days and analyzed by immunoblot for ATG5 expression. Unstimulated T cells were used as controls. As shown in Fig. 3A, T cells from the Atg5 knockout mice demonstrated minimal ATG5. This loss, therefore, correlates with a reduction in autophagy when compared with B6-WT T cells.

We next determined the impact of ATG5 deficiency on T-cell development. T cells from B6-WT and B6-Atg5^−/− animals were harvested from thymus and spleen 8 weeks after birth. Deficiency of ATG5 did not affect thymopoiesis as determined by similar numbers and percentages of CD4⁻/CD8⁻, CD4⁺, CD8⁺, and CD4⁺/CD8⁺ subsets in the thymus (Supplementary Fig. S2B). T cells harvested from B6-Atg5^−/− spleens had similar splenic cellularity (Supplementary Fig. S2C), as well as CD4⁺ and CD4⁺Foxp3 T cells (Supplementary Fig. S2D). However, a reduction in absolute numbers of CD8⁺ T cells (Supplementary Fig. S2E) was observed, demonstrating that loss of autophagy led to loss of CD8⁺ T cells in the periphery without a significant change in CD4⁺ subsets and in lymphocytes.

We next analyzed the impact of ATG5-dependent autophagy on T-cell functional responses to allo-stimulation in vitro and in vivo. B6-Atg5^−/− T_con cells stimulated with allogeneic BALB/c showed a significantly reduced proliferation when compared with WT-B6 controls on days 4 and 6 (Fig. 3B). The impact on proliferation was also confirmed with CFSE (Fig. 3C). In Fig. 3A, B6-WT T_con cells demonstrated an increase in ATG5 expression, which correlates with an observed increase in autophagy via increased autophagosomes observed via electron microscopy. As expected, B6-Atg5^−/− T_con cells demonstrated decreased autophagy (Fig. 3D). To specifically confirm ATG5 deletion was associated with the autophagy machinery, we stained CFSE premarked T cells for intracellular LC3 after stimulation. B6-Atg5^−/− T_con cells showed a lower proportion of LC3^high T cells compared with B6-WT T_con (30% vs. 65%) in MLR at day 4 (Fig. 3E). Stimulation with anti-CD3/CD28 similarly demonstrated reduction in proliferation of B6-Atg5^−/− T_con cells as determined by H3-thymidine incorporation, CFSE dilution assays, and LC3 quantification (Supplementary Fig. S2F–2H).

We next analyzed whether ATG5-dependent autophagy in T_con cells altered activation-induced apoptosis. In B6-Atg5^−/− T_con cells following in vitro activation, these cells demonstrated significantly enhanced apoptosis when compared with WT-B6 T_con cells (Fig. 3F; Supplementary Fig. S2I). In vivo, the impact on T-cell autophagy could be confounded by T-cell–extrinsic factors. Therefore, we next validated the reduction in proliferation in the absence of ATG5 in B6-WT and B6-Atg5^−/− T_con cells labeled with CFSE and transplanted with TCD-BM from WT-B6 into either syngeneic B6 or allogeneic BALB/c mice after conditioning, as in Methods. T cells were harvested from allo-recipient's spleen on day 7 and assessed using CFSE. Consistent with the in vitro data, allo-stimulation of B6-Atg5^−/− T_con cell donor cells (CD3⁺H2Kb⁺) in vivo resulted in a significant reduction in proliferation when compared with donor WT-B6 T_con cell population (Fig. 3G). We further confirmed the enhanced apoptosis as shown in Fig. 3H.

We then determined whether the increase in apoptosis and the decrease in proliferation was secondary to impairment of T_con cell activation in the absence of ATG5. T_con cells from B6-WT and B6-Atg5^−/− were stimulated with CD3/CD28 antibodies and harvested at 1, 12, and 24 hours, and stained for NFATc1 and ZAP-70. Unstimulated splenocytes were used as controls. WT-B6 and B6-Atg5^−/− T_con cells showed a similar activation pattern with a progressive increase of their NFATc1 level at 12 and 24 hours (Fig. 3I; Supplementary Fig. S2J, left) and of ZAP-70 at 24 hours (Fig. 3I; Supplementary Fig. S2J, right). Furthermore, consistent with previous reports (13, 20), no differences were observed in expression levels of CD69 upon activation (Supplementary Fig. S2K). These data demonstrate that T_con cell activation cascade downstream of the TCR was not affected by the absence of ATG5 in T cells.

We next analyzed whether deficiency in autophagy altered the balance between pro- and anti-apoptotic factors. Pro-apoptotic proteins BIM, Bak, and Bax, and anti-apoptotic proteins Bcl-2 and Bcl-XL levels were determined 48 hours after activation. B6-WT T_con cells demonstrated greater increase in Bcl-2 anti-apoptotic protein upon stimulation when compared with B6-Atg5^−/− T_con cells while no difference in expression of Bcl-XL was observed (Fig. 3J). In contrast, the expression of all three proapoptotic proteins BIM, Bak, and Bax were also similar with a modest decrease in BIM in WT-B6 and B6-Atg5^−/− T_con cells after stimulation (Fig. 3K). Thus, absence of ATG5 in T cells altered the levels of apoptotic proteins.

Following allo-stimulation, T cells proliferate and undergo Th differentiation. We therefore examined the impact of T-cell ATG5 deficiency on T-cell differentiation. Supernatants from allo-stimulated T_con cells cultured for 4 days were collected and analyzed for the presence of Th1 (IFNγ and IL2) and Th17 (IL17α) signature cytokines. We made the surprising observation that production of IL17α by B6-Atg5^−/− T_con cells was significantly increased as well as IFNγ levels in the supernatants despite the change in proliferation (Fig. 4A). Similar results were found in the B6-Atg5^−/− T_con cells with an increase in intracellular IL17α and IFNγ after in vitro allo-stimulation (Fig. 4B). Similar effects were also observed following nonspecific activation of cells with CD3/CD28 (Supplementary Fig. S3A and S3B) demonstrating that ATG5-dependent autophagy increased Th17 differentiation despite reducing proliferation. We next determined whether the increased Th1 and Th17 levels were observed in vivo following B6 to BALB/c allo-HCT. Consistent with in vitro observation, serum levels of IL17α and IFNγ were significantly greater in allo-recipients of B6-Atg5^−/− than in recipients of B6-WT (Fig. 4C). These data collectively demonstrate that despite reduction in proliferation and numbers, B6-Atg5^−/− T_con cells demonstrated greater Th1/Th17 differentiation, thus uncoupling T-cell proliferation from differentiation following in vivo allo-stimulation.

The impact of autophagy on T cell CTLs is unknown. In light of normal T-cell activation and enhanced differentiation into helper cells, we hypothesized that the cytolytic function of B6-Atg5^−/− T cells may be enhanced on a per cell basis despite the reduction in their proliferative capacity. B6-Atg5^−/− and B6-WT T_con were activated in a bulk allogeneic MLR with BALB/c-derived irradiated splenocytes. On day 6, the T cells were harvested and their counts adjusted to the CD8⁺ fraction as effectors against allogeneic P815 target cells at various effector-to-target ratios. Consistent with the hypothesis, the cytotoxic capacity of Atg5^−/− CD8⁺ T cells was significantly greater than that of WT-B6 T cells at every one of the tested ratios (Fig. 5A).

Because P815 tumor cells express only MHC class I and are therefore targeted by CD8⁺ T cells, we reasoned that B6-Atg5^−/− CD8⁺ T cells would demonstrate increased killing through the perforin/granzyme B pathway (21). This result would be due to an increase in perforin and granzyme B content because of a reduction of their clearance by autophagy. To this end, we harvested some B6-WT and B6-Atg5^−/− cells following allo-stimulation in an MLR and analyzed CD8⁺ T cells for their intracellular levels of granzyme B. B6-Atg5^−/− CD8⁺ T cells demonstrated a trend to higher content of granzyme B than B6-WT controls at day 4 (Fig. 5B). Western blot analysis from these T cells confirm greater granzyme B and also greater intracellular perforin levels (Fig. 5C). We next determined whether the increased perforin and granzyme B following activation in B6-Atg5^−/− T cells is directly because of reduction in autophagy. We therefore analyzed for perforin/granzyme B levels following allogeneic stimulation in the presence of CQ, as well as with rapamycin and compared them with diluent treated controls. As shown in Supplementary Fig. S4, inhibition of autophagy caused an increase of perforin, granzyme B, and LC3 II.

We next further explored the mechanism for the increase in perforin/granzyme B in the absence of ATG5-dependent autophagy. We hypothesized that autophagic vacuoles were critical for degrading and reducing intracellular levels of perforin/granzyme B in T cells. We therefore next determined whether perforin/granzyme B were taken into autophagosomes after T-cell activation by analyzing for colocalization of LC3 and perforin/granzyme B. B6-WT and B6-Atg5^−/− T_con cells were stimulated in an allo-MLR and assessed with confocal microscopy for overlapping signaling from LC3 (green) and perforin (pink). While LC3 and perforin overlapped (white arrows, Fig. 5D), no such overlap was observed in the negative control stimulated B6-Atg5^−/− T cells. To further confirm, we next used perforin antibody for immunoprecipitation in allo-stimulated B6-WT cells and then immunoblotted for granzyme B and LC3B antibodies. As shown in Fig. 5E, both perforin and granzyme B coprecipitated with LC3B. Collectively, these findings demonstrate that perforin and granzyme B are taken up in autophagosomes and degraded. In contrast, an absence of autophagy in B6-Atg5^−/− CD8⁺ T cells leads to greater accumulation of perforin/granzyme B thus converting them into more efficient cytotoxic cells.

Allo-reactive T-cell numbers, Th differentiation, and CTL are critical for causing GVHD (22). Allogeneic T cells demonstrated increase in autophagy flux while ATG5-dependent autophagy causes uncoupling of proliferation from Th differentiation and CTL function. We next determined its relevance in well-established in vivo mouse models of experimental GVHD. We once again utilized the clinically relevant B6→BALB/c MHC-major mismatched BMT model as described in methods. Allogeneic recipients receiving total T cells from B6-Atg5^−/− donors (Atg5^−/− recipients) demonstrated a significantly greater survival and reduced clinical GVHD severity than those receiving WT-B6 T cells (WT-B6 recipients; Fig. 6A).

To exclude possible strain-related artifacts, we also tested in a second clinically relevant, MHC matched but multiple minor antigen-mismatched B6C3H.sw model of GVHD. Similar protection from GVHD was observed in allo-recipient of B6-Atg5^−/− T cells when compared with B6-WT cells (Fig. 6B).

Because T_con showed (i) in vitro and in vivo defects in absence of autophagy, (ii) phenotypic analyses of splenic T cells from a B6-Atg5^−/− donor demonstrated reduced numbers of CD62L⁺; CD25⁻ naïve T cells compared with B6-WT T cells (Supplementary Fig. S5), and (iii) because naïve T cells cause GVHD, we explored whether GVHD improvement was caused by a reduction of naïve T_con cells in the transplant inoculum. We transplanted lethally irradiated (800 cGy) BALB/c recipient with 5 × 10⁶ TCD-BM from B6 donors along with 0.5 × 10⁶ purified allogeneic naïve (CD62L⁺; CD25⁻) T cells from either B6-Atg5^−/− or B6-WT spleen. Again, B6-Atg5^−/− allo-donor T-cell recipients demonstrated significantly greater survival and reduced severity GVHD (Fig. 6C). These data demonstrated that inhibition of autophagy in donor T_con cells ameliorates GVHD.

GVHD is tightly linked to the beneficial GVT response. GVT is dependent on IFNγ CTL functions of donor T cells (23). Because autophagy-deficient T cells showed greater CTL response, we hypothesized that despite reduction in T-cell proliferation causing less GVHD, the uncoupling of proliferation responses from enhanced CTL function will lead to sufficient preservation of GVT. Utilizing the same mouse model as above for GVHD, we incorporated a mouse DBA/2 mastocytoma cell line, P185, to the BMT cell inoculum. P185 cells were transduced with lentivirus that contained a GFP and luciferase (luc⁺) construct to facilitate in vivo monitoring of the tumor burden by bioluminescence imaging (BLI) after allo-HCT. Mice were imaged weekly for quantitation of tumor load in each mouse. We found that at a dose of 100 P185 tumor cells mixed with the 0.5 × 10⁶ B6-Atg5^−/− T cells, all of the allo-B6-Atg5^−/− T-cell mice succumbed to tumor (Fig. 6D). Suggesting that at this ratio of tumor to effector cells, GVT was not sufficiently preserved by allo-BMT. Next, we utilized a less aggressive lymphoma cell line (A20) at a dose of 1 × 10⁵ tumor cells with 0.5 × 10⁶ donor T cells. B6-Atg5^−/− T cells (Fig. 6E) demonstrated greater reduction in tumor related mortality.

In vivo GVT responses are complex. They are, among other factors, likely a function of qualitative effects of T cells (such as individual T-cell cytotoxic functions) and also quantitative effects (dependent on the numbers of T cells, clinically demonstrated by DLI studies). Because B6-Atg5^−/− T cells have a proliferative/survival deficiency, we hypothesized that enhancing T-cell numbers will increase GVT without causing a similar magnitude of increase in GVHD as WT-B6 T cells were able to reduce the P185 aggressive tumor cells. Therefore, to assess the impact on magnitude of GVHD and GVT responses, Balb/c recipients were transplanted with either 0.5 × 10⁶ or 2.0 × 10⁶ doses of T cells from B6-WT or B6-Atg5^−/− (allogeneic) T cells while keeping the tumor targets, P815 tumor cells at the same dose per mouse. GVHD severity was determined by clinical score. While increasing the T-cell dose increased GVHD in both WT and KO T cell recipients, the Atg5^−/− T cells caused significantly reduced GVHD scores when compared with B6-WT animals (Fig. 6F) at similar T-cell numbers. When these animals were analyzed for GVT responses by quantifying tumor induced mortality and with BLI, the animals that received reduced numbers (0.5 × 10⁶) of Atg5 and WT T cells showed reduced GVT (greater tumor mortality and BLI) than those that were transplanted with higher numbers (2 × 10⁶) of T cells (Fig. 6G and H). However, at lower T-cell dose the Atg5^−/− T cells demonstrated a reduced magnitude of GVT when compared with WT T cells (Fig. 6G and H), but this reduction was not apparent at higher T-cell dose (Fig. 6G and H). Thus, ATG5-dependent autophagy uncoupled allo-reactive T_con cell proliferation responses from their Th differentiation and cytotoxic functions, allowing for in vivo reduction in GVHD without causing a similar magnitude of loss of GVT responses.

T cells are essential for the beneficial GVT and the undesirable GVHD effects of allo-HCT. Separation of GVHD from GVT remains an intractable clinical problem. We demonstrate that absence of autophagy uncouples T-cell functions. Specifically, it separated T-cell proliferation from its effector function such as cytokine release and CTL function. This uncoupling effect caused a reduction in GVHD, but retained sufficient GVT effects after allo-HCT. Previous studies have demonstrated that autophagy-deficient T cells show defects in proliferation and changes in release of cytokines (1, 6, 7, 10, 24–26). We confirm similar effect on proliferation and apoptosis in T_con cells and further extend the role of autophagy in T-cell function by demonstrating that ATG5-dependent autophagy causes an alteration in Th differentiation and causes a concomitant increase in cytotoxic (CTL) functions. To our knowledge, this is the first demonstration of uncoupling of critical T-cell responses after allo-stimulation, namely proliferation/apoptosis from effector function via release of cytokines and cytotoxic functions. T-cell subsets have varying effects on GVHD and GVT. T_mem cells have been shown not to induce GVHD but retain sufficient GVT effects in humans and mice (27). The induction of GVHD has been demonstrated to be caused primarily by naïve T cells (T_n; ref. 22). The allo-recipients of purified T_n cells from the ATG5-deficient donors showed significant reduction in GVHD severity when compared with recipients that were transplanted with WT T_n cells. Thus, the changes in the composition of T_n and T_mem cells in the absence of ATG5 when compared with WT donor was not the cause for decrease in GVHD.

GVHD and GVT are caused by T cells, but its severity is modulated by many other immune cells. Host dendritic APCs play an important role in GVHD. Reduction of autophagy by a hypomorph of ATG16L1 in host DCs has been shown to increase GVHD (28). This is in contrast to our observation of the reduction in GVHD when ATG5-dependent autophagy is absent in donor T cells. Similarly, another study demonstrates that autophagy in Treg-regulated GVHD (29). Thus, the net effect on GVHD severity by mitigating autophagy nonspecifically in various immune cells could depend on which cellular subset is most significantly affected.

Our data on T_con cells are consistent with previous reports demonstrating decreased proliferation of total T cells that lack ATG5-dependent autophagy. However, we also noticed an increase in apoptosis of these cells only after activation in proliferating cells. The increase in apoptosis was observed more in CD8⁺ than CD4⁺ T cells. The tight interplay between autophagy and apoptosis might explain the increase in apoptosis (5, 30). Notably, dysregulation of organelle homeostasis and the accumulation of mitochondria and endoplasmic reticulum have been linked to cell death, due to the release of reactive oxygen species (20). It has also been shown that autophagy can regulate apoptosis through the sequestration of the large caspase-8 subunit in autophagosomes and its subsequent lysosomal degradation (31). Similar results were obtained in T cells deficient in autophagy protein Beclin-1, where levels of pro-caspases, Bim and Bcl-2, were increased (32), demonstrating that the balance between pro- and anti-apoptotic proteins can be regulated by autophagy. We similarly observed change in the pro versus antiapoptotic levels in ATG5-deficient T_con cells that may also have contributed to increased apoptosis. In agreement with previous reports, we demonstrated that autophagy is induced upon TCR engagement as demonstrated by upregulation of LC3 (13, 18, 33–35). We also additionally observed that ATG5-deficient T_con cells displayed the similar activation profile of Zap-70 and NFAT as found in WT controls demonstrating no significant difference in TCR activation. Furthermore, the relevance of autophagy deficiency might be distinct for the survival of various T-cell subsets. However, the role of autophagy in induction of a Th response under in vivo physiologic conditions is variable and not well characterized (36). We observed a remarkable enhancement of the Th signature cytokines (INFγ, IL17α) in ATG5-deficient T_con cells, thus demonstrating uncoupling of T_con proliferation from Th differentiation following allo-stimulation. The increase in pathogenic Th1/Th17 differentiation did not cause an increase in GVHD in vivo, likely from the reduction in their total effector pool by the reduced proliferation and increase in apoptosis. Depending on the type of MHC disparity and the context of minor antigen presentation, GVHD can be caused by either CD4 or CD8 or both subsets. While the in vitro allo-stimulation proliferation studies are largely reflective of CD4 responses, stimulation with anti-CD3⁺/28⁺ assays reflect both CD4 and CD8 effects. Thus, the impact of autophagy inhibition regulates both CD4 and CD8 responses.

T-cell expansion regulates both GVHD and GVT. This GVT response relies on the cytotoxic T CD8⁺ T-cell response and the T_m cells following allo-HCT (37). We made the surprising observation that the absence of ATG5-deficient CD8⁺ T cells, in the periphery, demonstrated significantly greater killing or cytotoxicity in vitro against allogeneic MHC 1-restricted P815 tumor cells than the B6-WT controls. The increase in CTL was associated with increased granzyme B in CD8⁺ T cells from B6-Atg5^−/− compared with those from B6-WT. Mechanistic studies demonstrate that both granzyme B and perforin colocalize with LC3 demonstrating that they are selected targets for the autophagosomes. These data are consistent with a recent article reporting that breast cancer cells can degrade granzyme B released from NK cells through the activation of autophagy in a hypoxic environment (38). Thus, granzyme B degradation is likely regulated by autophagic machinery, which in the absence of ATG5-dependent autophagy, gets accumulated over time in B6-Atg5^−/− CD8⁺ T cells, facilitating their ability for cytolytic activity. Furthermore, it may also explain the greater loss of B6-Atg5^−/− CD8⁺ cells from fratricidal cell death. Consistent with our in vitro data, ATG5-deficient T cells demonstrated similar uncoupling of proliferation from cytotoxic response in vivo after allo-BMT. Specifically, we saw decreased proliferation, increased apoptosis, increased Th cytokines and increased cytotoxicity. The net reduction in allo-reactive T-cell numbers from decreased proliferation, increased apoptosis caused less GVHD. These observations were true after both MHC-major mismatch and MHC-matched multiple minor antigen-mismatch transplants, demonstrating that they were neither dependent on the mice strains nor on the degree of inflammation or MHC disparity. Despite decreased GVHD, a substantial preservation in GVT was observed, especially at higher T-cell doses. The likely reason for the sufficient preservation of GVT responses, especially when higher T-cell doses were utilized is likely from increased CTL/effector cytokine release despite a reduction in T-cell numbers. Furthermore, autophagy has been shown to modulate Th9 (IL9) release and affect antitumor responses (39). The relative contribution of IL9, cytotoxicity, and T-cell expansion kinetics will need to be assessed in future studies. However, our results should be interpreted with caveats that are inherent to these model systems. It could be a reflection of the tumor and T-cell expansion kinetics, because when tumor-to-effector cell ratios were changed, the enhanced GVT effect was lost. These data suggest the critical nature of the cellular kinetics in addition to the critical role of CTL/cytotoxicity pathways in controlling tumors.

In summary, our study thus demonstrates that autophagy plays a differential role in T-cell functions, separating the proliferative responses from effector Th and cytotoxic responses. Thus T-cell–specific inhibition of autophagy may be a therapeutic strategy to improve allo-BMT outcomes. Importantly, global inhibition of autophagy in all immune cell subsets, the host tumors and target tissues, may result in variable impact. Future studies will need to systematically dissect the relevance of autophagy in various immune cells subsets and nonimmune cell targets after allo-HCT.

No disclosures were reported.

K. Oravecz-Wilson: Data curation, formal analysis, supervision, investigation, writing-original draft, writing-review and editing. C. Rossi: Conceptualization, formal analysis, investigation, writing-original draft, writing-review and editing. C. Zajac: Data curation, methodology. Y. Sun: Data curation. L. Li: Data curation, methodology, writing-review and editing. T. Decoville: Data curation. H. Fujiwara: Methodology. S. Kim: Methodology. D. Peltier: Methodology. P. Reddy: Conceptualization, supervision, funding acquisition.

This work was supported by the NIH grants CA-203542, HL-152605, and HL-149633 (to P. Reddy). This research was also supported by the NCI of the NIH with award number CA-217156. This research was supported with the help of image acquisition and analysis by the University of Michigan's Biomedical Research Core Facilities Microscopy Core.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.