Graphical abstract

   graphic file with name fx1.jpg
   [125]Open in a new tab

Highlights

     * •
       CD38^hiCD8^+ T cells are associated with ICB resistance in melanoma
     * •
       CD38 expression is induced by chronic TCR stimulation and type I
       IFN signaling
     * •
       Blocking CD38 restores NAD^+ pools and mitochondrial function
     * •
       Dual PD-1/CD38 blockade overcomes ICB resistance in patient-derived
       3D tumor models
     __________________________________________________________________

   Revach et al. show that CD38 is elevated in dysfunctional T cells in
   melanoma, driving an impaired metabolic state, and that disrupting CD38
   improves T cell function and restores immune checkpoint blockade (ICB)
   sensitivity in ICB-resistant patient-derived melanoma organotypic tumor
   spheroids, supporting further clinical evaluation of CD38-directed
   therapy in melanoma.

Introduction

   Cancer immunotherapy with immune checkpoint blockade (ICB) has
   revolutionized cancer treatment, providing durable responses and cures
   for a subset of patients with advanced metastatic disease.[126]^1
   Despite the success of ICB in melanoma and other cancers, resistance
   remains a central challenge and is ultimately fatal.[127]^2 Dual
   PD-1/CTLA-4 blockade demonstrated a modest objective response rate
   (28%) in the second-line setting and is associated with significant
   toxicities.[128]^3 An improved response rate (31.4%) was observed in
   ICB-resistant melanoma patients treated with adoptive T cell therapy
   (lifileucel),[129]^4 but a durable benefit was observed in a minority
   of patients, underscoring the need for rational therapeutic strategies
   to overcome ICB resistance.

   Tumor-infiltrating CD8^+ T lymphocytes (TILs) are key determinants of
   anti-tumor immunity and successful response to
   ICB.[130]^5^,[131]^6^,[132]^7 Single-cell sequencing has revealed that
   CD8^+ T cells defined by expression of stem-like transcription factors
   (e.g., TCF7) are enriched in melanoma patients sensitive to ICB,
   whereas exhausted CD8^+ T cells, defined by high expression of
   co-inhibitory receptors, are enriched in ICB-resistant melanoma
   patients.[133]^8 Exhausted T cells develop during chronic antigen
   stimulation, are characterized by loss of proliferative potential and
   diminished effector function, and may acquire immunosuppressive
   properties.[134]^7^,[135]^8^,[136]^9^,[137]^10 Therapeutic strategies
   directed at promoting stem-like T cell expansion, enhancing effector
   functions, and preventing (or reversing) T cell exhaustion would
   represent rational approaches to overcome ICB resistance.

   CD38, an ecto-enzyme involved in NAD^+ catabolism, is upregulated
   during T cell activation and is associated with T cell exhaustion in
   cancer and chronic viral infection.[138]^11^,[139]^12 CD38^hiCD8^+ T
   cells are induced following unsuccessful response to ICB contributing
   to ICB resistance, while depletion of CD38^hiCD8^+ cells improved
   anti-tumor immunity in murine tumor models.[140]^13 Yet the precise
   role of the CD38-NAD^+ axis in governing the balance of
   tumor-infiltrating CD8^+ T cell states is unknown. Importantly, the
   potential impact of targeting CD38 to overcome T cell dysfunction and
   ICB resistance in human melanoma tumor models has not been tested.

   Here, we show that the upregulation of CD38 in CD8^+ T cells is
   strongly associated with T cell exhaustion and ICB resistance in
   melanoma. CD38^+CD8^+ T cells accumulate specifically in the tumor
   microenvironment (TME) and exhibit features of T cell dysfunction, with
   low TCF7 expression, decreased proliferative capacity, and altered
   mitochondrial bioenergetics. Pharmacological or genetic disruption of
   CD38 limits the development of these functional and metabolic defects,
   resulting in increased TCF7 expression, improved T cell function, and
   restored ICB sensitivity. Importantly, we demonstrate the therapeutic
   potential of CD38 blockade to overcome resistance to ICB using a cohort
   of patient-derived organotypic tumor spheroids
   (PDOTS),[141]^14^,[142]^15 which retain autologous tumor-infiltrating
   immune and stromal cells and are amenable to ex vivo 3D microfluidic
   culture and drug sensitivity testing.

Results

CD38^+CD8^+ TILs predict ICB resistance

   We first examined the expression of CD38 in CD8^+ TILs using published
   single-cell RNA sequencing (scRNA-seq) data from patients with advanced
   melanoma treated with ICB ([143]Figures 1A, [144]S1A, and S1C).[145]^8
   We observed increased CD38 expression in exhausted CD8^+ TILs (clusters
   1–3), which closely mirrored the expression of PD-1 (PDCD1)
   ([146]Figures 1A–1C). CD8^+ T cell states associated with early T cell
   activation or effector memory function (clusters 4–6) were marked with
   low CD38 expression and high expression of the transcription factor
   TCF7 ([147]Figures 1B and 1D). As TCF7-expressing CD8^+ T cells are
   associated with ICB response,[148]^8 we sought to determine whether
   CD38-expressing CD8^+ T cells were associated with ICB resistance in
   human cancer. Consistent with prior observations,[149]^8^,[150]^13 we
   observed a higher proportion of CD38 expressing CD8^+ TILs in
   ICB-resistant (ICB-NR) melanoma tumors ([151]Figures 1E and 1F).
   CD8-specific expression of CD38 also demonstrated high predictive power
   for ICB resistance in melanoma patients (area under the curve [AUC]
   0.87, p = 1.53 × 10^−5) ([152]Figure 1G), and higher proportions of
   CD38-expressing CD8^+ T cells were observed in both baseline
   (pre-treatment) and post-treatment tumors from ICB-resistant melanoma
   patients, though more robustly in the post-treatment setting
   ([153]Figures S1D and S1E). Furthermore, CD8^+ T cell-specific CD38
   expression was associated with resistance to both single-agent PD-1
   blockade and dual PD-1/CTLA-4 blockade ([154]Figures S1F and S1G).

Figure 1.

   [155]Figure 1
   [156]Open in a new tab

   CD38^+CD8^+ TILs predict ICB resistance

   (A–H) scRNA-seq of CD45^+ immune cells from melanoma patients.[157]^8
   (A) CD8^+ T cell clusters (n = 6,350). (B–D) Expression of (B) CD38,
   (C) PD-1 (PDCD1), and (D) TCF7. (E) CD8^+ T cell clusters by ICB
   response, (F) Proportion of CD38 expressing CD8^+ T cells in ICB
   responders (ICB-R) and non-responders (ICB-NR). Two-sided unpaired t
   test. Means (bars) and individual values (open circles) are shown. (G
   and H) Receiver-operating characteristic (ROC) curves, demonstrating
   (G) the predictive power of CD38^+CD8^+ TILs in melanoma tumors and (H)
   the specific performance of cluster 6 exhausted CD8^+ T cells for ICB
   resistance. FPR, false-positive rate; TPR, true-positive rate.

   (I–L) scRNA-seq analysis of CD8 T cells from melanoma validation
   cohort.[158]^16 (I) Uniform manifold approximation and projection
   (UMAP) of CD8 T cells (n = 20,210). (J) Proportion of CD38 expressing
   CD8^+ T cells. Two-sided unpaired t test. Means (bars) and individual
   values (open circles) are shown. (K) CD38 expression and (L) TCF7
   expression.

   (M) Proportion of CD38 expressing CD8^+ T cells from NSCLC[159]^17
   (MPR, major pathological response; ICB-R, n = 23, non-MPR; ICB-NR, n =
   34, two-sided unpaired t test). Means (bars) and individual values
   (open circles) are shown.

   (N) ROC curve demonstrating the predictive power of CD38^+CD8^+ T cells
   for lack of ICB treatment benefit in NSCLC.

   See also [160]Figures S1 and [161]S2; [162]Table S1.

   Further subcluster analysis demonstrated that exhausted CD8^+ T cells
   (cluster 6 in [163]Figure S1A) showed the highest predictive ability
   for ICB resistance (AUC 0.83, p = 5.3 × 10^−4) compared to the
   exhausted/cell-cycle cluster (cluster 11 in [164]Figure S1A, AUC 0.64,
   p = 0.068) and exhausted/heat-shock cluster (cluster 9 in [165]Figure
   S1A, AUC 0.62, p = 0.300) ([166]Figures 1H, [167]S1H, and S1I). CD38
   expression in other T cell and immune clusters also showed lower
   predictive power for ICB resistance and were mostly not significant
   ([168]Figures S1J–S1Q). While tumor-specific expression of CD38 has
   been observed in non-small cell lung cancer (NSCLC) and other cancers
   and is associated with ICB resistance,[169]^18 CD38 expression in human
   melanoma tumor cells was rarely observed compared to CD45^+ immune
   cells ([170]Figures S1R and [171]S2A–S2D).[172]^19^,[173]^20

   To validate these findings, we examined an independent cohort of
   ICB-treated melanoma patient tumors and evaluated tumor-infiltrating
   CD8^+ T cells using scRNA-seq ([174]Figures 1I and [175]S2E; [176]Table
   S1).[177]^16 Enrichment of CD38-expressing CD8^+ T cells was again
   observed in ICB-NR compared to ICB-R patients ([178]Figure 1J), with
   CD38 expression enriched in exhausted T cell clusters and
   anti-correlated with TCF7 expression ([179]Figures 1K and 1L). To
   determine whether the association between CD38-expressing CD8^+ T cells
   and ICB resistance could be observed in other cancers, we examined the
   expression of CD38 in CD8^+ T cells from patients with NSCLC.[180]^17
   Consistent with our observations in melanoma, CD38 expression in CD8^+
   T cells was associated with a lack of ICB treatment benefit in NSCLC
   and was predictive for lack of response (AUC 0.75, p = 0.002)
   ([181]Figures 1M and 1N). Taken together, these analyses indicate that
   CD38-expressing tumor-infiltrating CD8^+ T cells are strongly
   associated with ICB resistance in multiple cancer types.

Intratumoral CD38^+CD8^+ T cells accumulate during tumor progression

   To determine the temporal dynamics of CD8^+ T cell-specific CD38
   upregulation and examine their tumor specificity, we examined
   time-resolved scRNA-seq data from B16 murine melanoma tumors with
   matched tumor draining lymph nodes (dLNs) and normal lymph nodes
   (nLNs).[182]^21 While Cd38 expression was observed in other immune
   populations in tumors, dLNs, and nLNs ([183]Figure S3A),
   Cd38-expressing CD8^+ T cells were observed primarily within tumors
   ([184]Figures 2A and 2B). Further, Cd38 expression was largely absent
   from nLNs and dLNs but was detectable in tumors by day 7 with elevated
   expression at days 10 and 16 with increasing tumor burden ([185]Figure
   2C). In contrast, Tcf7 expression was observed largely in CD8^+ T cells
   in nLNs and dLNs, with decreased expression in CD8^+ TILs ([186]Figure
   2C). Further, while modest residual Tcf7 expression was observed at
   early time points (day 7), Tcf7 was absent by day 10 and remained
   undetectable at day 16 as tumor burden increased ([187]Figure 2C).
   Therefore, while CD38^+CD8^+ T cells can be detected in the tumor
   periphery, Cd38 upregulation and the parallel loss of Tcf7 expression
   in CD8^+ T cells occurs specifically in the TME during tumor
   progression.

Figure 2.

   [188]Figure 2
   [189]Open in a new tab

   Intratumoral CD38^+CD8^+ T cells accumulate during tumor progression

   (A) scRNA-seq[190]^21 of T/NK cells from B16 tumors (Tum),
   tumor-draining lymph nodes (dLN), and normal lymph nodes (nLN).

   (B and C) Dotplots indicating (B) Cd38 expression and (C) Cd38 and Tcf7
   expression at days 7, 10, and 16.

   (D–G) CyTOF analysis of CD38^+CD8^+ T cells in peripheral blood from
   melanoma patients (D) by response to ICB; before (E) and after (F) ICB
   treatment. (G) CD8^+CD38^+IgG4^+ by response to ICB. Two-sided unpaired
   t test. Means (bars) and individual values (open circles) are shown.

   (H–M) scRNA-seq of T/NK tumor-infiltrating leukocytes from control/IgG
   (n = 3) and αPD-1 (n = 4) B16-ova tumors.[191]^14 (H) UMAP of T/NK cell
   clusters by condition; (I and J) Cd38 expression and (K) proportion of
   Cd38 expressing terminal effector CD8^+ TILs per condition. Two-sided
   unpaired t test. Median (line) and individual values (open circles) are
   shown. (L and M) UMAP and track plots showing Cd38 gene expression.

   Immune population statistics can be found in [192]Figure S3D. ∗p <
   0.05. See also [193]Figure S3 and [194]Table S2.

   Higher circulating levels of CD38^+CD4^+ T cells have been associated
   with diminished ICB sensitivity in melanoma.[195]^9 Multiplexed
   cytometry by time of flight (CyTOF) analysis of peripheral blood from
   melanoma patients (before and after initiating ICB treatment) revealed
   higher levels of CD38^+CD8^+ T cells in ICB-NR patients ([196]Figure 2D
   and [197]Table S2). In contrast to tumors, pre-treatment circulating
   levels of CD38^+CD8^+ T cells were not significantly different between
   responders and non-responders ([198]Figure 2E), whereas higher levels
   of CD38^+CD8^+ T cells were observed in post-treatment ICB-NR samples
   ([199]Figure 2F). Increased IgG4^+ detection was also observed on
   peripheral blood CD38^+CD8^+ T cells in ICB-resistant patients,
   identifying cells bound by clinically administered IgG[4] isotype
   anti-PD-1 antibodies ([200]Figures 2G and [201]S3B).[202]^22 In
   agreement with these observations, circulating plasma levels of CD38
   were increased in melanoma patients 6 weeks after initiating ICB
   treatment but remained elevated at 6 months only in non-responder
   patients ([203]Figure S3C).

ICB treatment increases CD38^+CD8^+ T cells

   Ineffective ICB treatment promotes the induction of dysfunctional CD8^+
   T cells that contribute to ICB resistance.[204]^13 To evaluate the
   accumulation of Cd38 expressing TILs following ICB treatment, we
   examined scRNA-seq data of tumor-infiltrating CD45^+ immune cells from
   B16-ova tumors treated with ICB.[205]^14 Anti-PD-1 treatment resulted
   in the expansion of early-effector and terminal exhausted/effector
   CD8^+ T cells with a reduction of progenitor CD8^+ T cells
   ([206]Figures 2H–2J and [207]S3D). Yet Cd38 expression was
   significantly increased only in terminal exhausted/effector CD8^+ T
   cells following PD-1 blockade ([208]Figures 2K–2M, [209]S3E, and S3F).
   Thus, while Cd38 is expressed in various T/natural killer (NK) cell
   populations (and multiple immune cell types), its upregulation
   following ICB treatment is observed primarily in exhausted CD8^+ T
   cells.

CD38^hiCD8^+ T cells are dysfunctional

   We next sought to examine the association between CD38 and T cell
   exhaustion. Increased expression of multiple co-inhibitory receptors
   (PDCD1, HAVCR2, ENTPD1, CTLA4, and LAG3) was observed together with
   CD38 in exhausted CD8^+ T cells (clusters 1–3, see [210]Figure 1A) from
   ICB-treated melanoma patients,[211]^8 whereas CD38 expression was
   comparatively low in effector-memory CD8^+ T cells (clusters 4–6, see
   [212]Figure 1A) expressing TCF7 ([213]Figure 3A). Analysis of
   co-expression deviation proportions (CDPs) in our validation melanoma
   scRNA-seq dataset[214]^16 (see [215]Figures 1I–1L) also demonstrated
   co-expression of CD38 with multiple co-inhibitory receptors and an
   inverse correlation with TCF7 expression ([216]Figure 3B). Examination
   of differentially expressed genes in CD38^+ and CD38^−
   tumor-infiltrating CD8^+ T cells from melanoma patients treated with
   ICB[217]^8 revealed increased expression of co-inhibitory receptors
   (e.g., PDCD1 and HAVCR2) in CD38^+ CD8 T cells and increased expression
   in T cell stem/memory markers (e.g., TCF7, IL7R, and CCR7) in
   CD38^−CD8^+ T cells ([218]Figure 3C and [219]Table S3). Further,
   analysis of TILs from B16-ova tumors[220]^14 revealed similar
   upregulation of co-inhibitory receptors and exhaustion-related
   molecules in CD38^+ TILs, whereas increased expression of Tcf7, Il7r,
   and Ccr7 was observed in CD38^− TILs ([221]Figure 3D and [222]Table
   S4).

Figure 3.

   [223]Figure 3
   [224]Open in a new tab

   CD38^hiCD8^+ T cells are dysfunctional

   (A) Expression of exhaustion and effector/memory-related genes in CD8^+
   TILs from human melanoma tumors.[225]^8

   (B) Co-expression deviation proportion plot demonstrating co-expression
   of exhaustion-related genes and TCF7 from melanoma validation
   cohort.[226]^16

   (C and D) differentially expressed genes based on CD38 expression in
   (C) CD8^+ TILs from human melanoma[227]^8 and (D) CD3^+ TILs from
   B16-ova murine melanoma.[228]^14

   (E and F) CD38^hi and CD38^lo B7-H3.CAR-T cells (E) surface staining of
   PD-1^+CD39^+TIM-3^+ (n = 3; two-sided paired t test) and (F) TCF7
   expression (n = 3; two-sided unpaired t test). Means (bars) and
   individual values (open circles) are shown.

   (G) Scheme depicting acute and chronic TCR stimulation.

   (H–J) (H) Acute and chronic B7-H3.CAR-T proliferation assay (n = 3;
   two-way ANOVA with Sidak correction for multiple comparisons). Means ±
   SEM (shaded area) are shown. Staining of (I) CD38^+CD39^+ and (J)
   PD-1^+TIM-3^+ (n = 3, two-sided paired t test). Means (bars) and
   individual values (open circles) are shown.

   (K) Cytotoxicity assay toward 10164 patient-derived melanoma cell line.
   A representative experiment out of three is presented; two more are in
   [229]Figures S4E and S4F. (n = 3 biological replicates; three
   independent experiments; two-way ANOVA with Sidak correction for
   multiple comparisons). Means ± SEM (shaded area) are shown.

   (L–N) Analysis of chronically stimulated control sgRNA and CD38 sgRNA
   B7-H3.CAR-T cells. (L) Proliferation assay (n = 3 biological
   replicates; three independent experiments; two-way ANOVA with Sidak
   correction for multiple comparisons). Means ± SEM (shaded area) are
   shown. (M) TCF7 intracellular staining (n = 4; two-sided paired t
   test). Means (bars) and individual values (open circles) are shown. (N)
   Cytotoxicity assay against 10164 melanoma cells (n = 3 biological
   replicates; three independent experiments; two-way ANOVA with Sidak
   correction for multiple comparisons). Means ± SEM (shaded area) are
   shown.

   ∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.0001. See also [230]Figure S4;
   [231]Tables S3 and [232]S4.

   Chimeric antigen receptor T (CAR-T) cells are increasingly utilized to
   study T cell biology, providing a renewable source of antigen-specific
   effector T cells.[233]^23^,[234]^24 To further study the mechanism of
   CD38 function in human T cells, we examined human B7-H3-directed CAR-T
   cells (B7-H3.CAR-T).[235]^25^,[236]^26 Following isolation of CD38^hi
   and CD38^lo B7-H3.CAR-T cells, CD38^hi CAR-T cells exhibited increased
   expression of exhaustion-related surface markers (PD-1, TIM-3, and
   CD39) and decreased expression of TCF7 and FOXO1 compared to CD38^lo
   B7-H3.CAR-T cells ([237]Figures 3E, 3F, and [238]S4A–S4C). To model the
   development of T cell dysfunction, we performed repetitive in vitro T
   cell receptor (TCR) stimulation (every 2–4 days) with αCD3/αCD28 beads
   for 10–14 days (“chronic” stimulation) compared to a one-time
   activation with αCD3/αCD28, followed by interleukin-2 (IL-2) expansion
   (“acute” stimulation) ([239]Figure 3G). Chronic TCR stimulation of
   CAR-T cells resulted in reduced proliferative capacity ([240]Figure
   3H), increased levels of CD38^+CD39^+ and PD-1^+TIM-3^+ cells
   ([241]Figures 3I and 3J), and reduced ability to lyse target cancer
   cells compared to standard (acute) stimulation ([242]Figures 3K and
   [243]S4D–S4F), recapitulating key features of T cell exhaustion.
   Similar observations were obtained with patient-derived CD8^+ TILs and
   matched primary tumor cells ([244]Figures S4G and S4H).

   To determine whether CD38 is necessary for the development of T cell
   dysfunction, we deleted CD38 in B7-H3.CAR-T cells using CRISPR-Cas9
   gene editing ([245]Figure S4I) followed by chronic TCR stimulation.
   CD38 single-guide RNA (sgRNA) B7-H3.CAR-T cells displayed a clear
   proliferative advantage over control sgRNA CAR-T cells ([246]Figure 3L)
   accompanied by an increased proportion of TCF7^+ CAR-T cells
   ([247]Figure 3M), indicating a higher T cell stemness potential.[248]^8
   Further, CD38 sgRNA CAR-T cells exhibited increased cytotoxic capacity
   toward patient-derived melanoma cell lines compared to control sgRNA
   CAR-T cells ([249]Figure 3N), indicating that CD38 deletion prevents T
   cell dysfunction in a human T cell exhaustion model.

CD38^+ T cells exhibit altered bioenergetics

   To identify key cellular processes associated with upregulation of CD38
   in T cells, we performed gene set enrichment analysis (GSEA) using
   CD8^+ TILs from ICB-treated melanoma patients[250]^8 and CD3^+ TILs
   from B16-ova tumors.[251]^14 Top scoring gene sets included type I/II
   interferon signaling (IFN-α/IFN-γ response) as well as oxidative
   phosphorylation (OXPHOS), which correlated strongly between mouse and
   human data ([252]Figures 4A and [253]S5A–S5D; [254]Tables S5 and
   [255]S6). Given these findings and previous reports linking CD38 to
   altered mitochondrial fitness,[256]^12^,[257]^27 we examined the
   mitochondrial mass and mitochondrial membrane potential (MMP) of
   CD38^lo and CD38^hi T cells. Increased mitochondrial mass and MMP were
   observed in CD38^hi B7-H3.CAR-T cells and human melanoma CD8^+ TILs
   ([258]Figures 4B, 4C, and [259]S5E–S5G), consistent with previous
   reports.[260]^12^,[261]^28 Importantly, CD38 deletion resulted in
   reduced mitochondrial mass and lower MMP compared to control sgRNA
   following chronic TCR stimulation ([262]Figures 4D and 4E), indicating
   enhanced T cell mitochondrial fitness.[263]^12^,[264]^28^,[265]^29 In
   line with these observations, CD38 sgRNA CAR-T cells exhibited
   increased extracellular acidification (ECAR) ([266]Figure S5H),
   indicating enhanced glycolysis[267]^30 and increased mitochondrial
   spare respiratory capacity following the addition of the mitochondrial
   uncoupling agent, FCCP (carbonyl
   cyanide-p-trifluoromethoxyphenylhydrazone), suggesting improved T cell
   fitness under conditions of increased metabolic stress ([268]Figure
   4F).[269]^31

Figure 4.

   [270]Figure 4
   [271]Open in a new tab

   CD38^+ T cells exhibit altered bioenergetics

   (A) Correlation analysis between the GSEA of CD38^+/−CD8^+ T cells in
   human melanoma[272]^8 and CD38^+/−CD3^+ T cells in B16-ova murine
   melanoma.[273]^14

   (B–E) Flow cytometry of (B and D) mitochondrial mass and (C and E) MMP
   in indicated groups (n = 3; two-sided paired t test). Means (bars) and
   individual values (open circles) are shown.

   (F) Oxygen consumption rate (OCR) under basal condition and in response
   to inhibitors (n = 5, two biological replicates, two-way ANOVA with
   Sidak correction for multiple comparisons). Data are presented as mean
   ± SEM.

   (G and H) Relative levels of (G) NAD^+ (nicotinamide adenine
   dinucleotide) and (H) NADP^+ (nicotinamide adenine dinucleotide
   phosphate); log[2] fold change (L2FC) from control is shown (n = 6
   biological replicates; two independent experiments; two-sided unpaired
   t test). Means (bars) and individual values (open circles) are shown.

   (I) scheme of NAD^+ metabolism and L2FC of indicated analytes.

   (J) OCR as in (F) of B7-H3.CAR-T cells ± CD38i (n = 5, two biological
   replicates, two-way ANOVA with Sidak correction for multiple
   comparisons). Data are presented as mean ± SEM.

   (K and L) staining of (K) TIM-3^+PD-1^+ and (L) CD39^+TIM-3^+ in
   B7-H3.CAR-T ± CD38i (n ≥ 4; two-sided paired t test). Means (bars) and
   individual values (open circles) are shown.

   (M) TCF7 expression by RT-qPCR in B7-H3.CAR-T cells in indicated groups
   (n = 4; two-sided unpaired t test). Means (bars) and individual values
   (open circles) are shown.

   (N and O) Relative levels of (N) NAM (nicotinamide) and (O) ADPR
   (adenosine diphosphate ribose) in B7-H3.CAR-T cells ± CD38i (n = 6
   biological replicates; two independent experiments; two-sided unpaired
   t test). Means (bars) and individual values (open circles) are shown.

   ∗p < 0.05, ∗∗p < 0.01. See also [274]Figures S5–S7; [275]Tables S5,
   [276]S6, and [277]S8.

Disrupting CD38 restores NAD^+ in exhausted T cells

   As NAD^+ boosting has been shown to restore mitochondrial fitness and
   improve T cell function,[278]^29 we next sought to determine whether
   disrupting CD38 could improve mitochondrial metabolic fitness by
   restoring cellular NAD^+ levels. Genetic deletion or pharmacological
   inhibition of CD38 with 78c (CD38i)[279]^32^,[280]^33 in B7-H3.CAR-T
   cells resulted in increased NAD(H) levels that were further enhanced
   with nicotinamide riboside (NR) supplementation to boost cellular NAD^+
   levels ([281]Figures S6A and S6B).[282]^29 To define the broader
   metabolic consequences of targeting CD38 in T cells, we performed
   intracellular metabolite profiling using liquid chromatography-mass
   spectrometry (LC-MS) of acute (CD38^hi) and chronically TCR-stimulated
   (CD38^low) B7-H3.CAR-T cells (see [283]Figure 3I). Chronic TCR
   stimulation was associated with changes in a number of metabolites,
   including intermediate-chain acylcarnitines, glycerol-3-phosphate
   shuttle metabolic intermediates (glycerol-3-phosphate and
   dihydroxyacetone phosphate), malate-aspartate shuttle intermediates
   (malate and aspartate), and NADPH ([284]Figures S6C and S6D; [285]Table
   S7). Of note, NADPH can be generated from NAD^+ as part of the pentose
   phosphate pathway (PPP) via phosphorylation by NAD^+ kinase to NADP^+
   and subsequent reduction of NADP^+ to NADPH by glucose-6-phosphate
   dehydrogenase (G6PD), an enzyme also enriched in exhausted CD8^+ T
   cells in melanoma.[286]^34

   Further metabolomic analysis of CD38i-treated, chronically
   TCR-stimulated B7-H3.CAR-T cells demonstrated increased levels of both
   NAD^+ and NADP^+ ([287]Figures 4G–4I and [288]Table S8), showing
   restoration of cellular NAD^+ pools in exhausted T cells upon CD38
   inhibition. CD38i also increased the mitochondrial spare respiratory
   capacity of chronically TCR-stimulated B7-H3.CAR-T ([289]Figure 4J),
   similar to CD38-knockout T cells. Restoration of cellular NAD^+ pools
   was accompanied by decreased proportions of PD-1^+TIM-3^+ and
   CD39^+TIM-3^+ B7-H3.CAR-T cells and increased TCF7 expression
   ([290]Figures 4K–4M), whereas NAD^+ depletion using the nicotinamide
   phosphoribosyltransferase inhibitor FK866[291]^35 (NAMPTi) reduced
   NAD^+ levels and TCF7 expression ([292]Figures S6E and S6F), increased
   CD8^+PD-1^+TIM-3^+ and CD8^+CD38^+ T cells ([293]Figures S6G and S6H),
   and impaired proliferation and cytotoxicity ([294]Figures S6I and S6J).
   Further, NAD^+ boosting during chronic stimulation by NR
   supplementation (mimicking the effect of CD38i)[295]^36 ([296]Figure
   S6B) resulted in increased T cell proliferation and cytotoxicity
   ([297]Figures S6K and S6L), similar to the effect of CD38 inhibition
   (see [298]Figures 3L and 3N), with no significant changes in CD38
   expression ([299]Figure S6M).

   Compared to CD38 inhibition or deletion, the effect of NR
   supplementation on TCF7 expression was modest ([300]Figure S6N),
   suggesting that TCF7 regulation might also be governed by other
   CD38-regulated metabolites. In addition to NAD^+ catabolism, CD38
   promotes the production of cyclic ADP-ribose (cADPR), which was
   recently shown to regulate the expression of TCF7 via the
   ADPR/Ca^2+/RyR2/AKT pathway.[301]^36^,[302]^37 CD38 inhibition in
   chronically stimulated T cells resulted in decreased levels of the
   NAD^+ precursor nicotinamide (NAM)[303]^38 and ADPR[304]^39
   ([305]Figures 4N and 4O; [306]Table S8). Further, we observed
   elevations in cytoplasmic Ca^2+ levels following chronic stimulation
   (CD38^hi) ([307]Figure S6O) that was reduced upon CD38 inhibition
   ([308]Figure S6P). These findings implicate CD38 in the regulation of
   TCF7 expression via modulation of both NAD^+ and cADPR levels.

   ADPR can be further metabolized to the immunosuppressive metabolite
   adenosine[309]^40 by the sequential action of CD203a and CD73
   ([310]Figure 4I).[311]^41 While adenosine was not reliably detected in
   our analysis, related metabolites (e.g., AMP, ATP, ADP,
   S-adenosyl-L-methionine [SAM], and deoxyadenosine) were detectable and
   not significantly altered upon CD38i treatment ([312]Figures S7A and
   S7E). It is, therefore, possible that other ectoenzymes, such as CD73,
   CD203a, and CD39 expressed on chronically stimulated CAR-T cells
   ([313]Figures S7F and S7G), compensate for the inhibition of CD38 and
   contribute to the maintenance of these metabolites.[314]^41 Further,
   modulation of adenosine levels might depend on T cell extrinsic
   processes involving multiple cell types in the
   TME.[315]^9^,[316]^10^,[317]^41 LC-MS evaluation of conditioned media
   of control and CD38i-treated CAR-T cells failed to detect extracellular
   NAD(P)^+, and no significant changes in extracellular levels of ADPR,
   NAM, SAM, or deoxyadenosine were observed ([318]Figures S7H–S7K and
   [319]Table S9), consistent with prior reports that CD38 primarily
   governs intracellular NAD^+ levels.[320]^42 Using surface and
   intracellular immunofluorescence staining of CD38 and PD-1 in acute or
   chronically stimulated T cells, we showed that CD38 can be localized
   both in the cell membrane and in the cytoplasm of T cells ([321]Figures
   S7L–S7N), as previously shown.[322]^43^,[323]^44 CD38 cytoplasmic
   aggregates were more prevalent in chronically stimulated T cells and
   specifically in PD-1-positive T cells ([324]Figures S7L–S7N).

Type I IFN induces CD38 expression in T cells

   GSEA also revealed a strong association between CD38 expression and
   type I/II IFN signaling ([325]Figures 4A, [326]S5A, and S5B;
   [327]Tables S5 and [328]S6). While important for anti-tumor immunity,
   recent reports have suggested a role for type I IFN signaling in the
   development of T cell dysfunction,[329]^45^,[330]^46^,[331]^47 and
   exposure of healthy CD8^+ T cells to type I IFN promoted mitochondrial
   and T cell dysfunction via upregulation of CD38 and enhanced NAD^+
   consumption.[332]^48 Analysis of scRNA-seq of human melanoma CD8^+ TILs
   showed co-expression of CD38 and multiple type I IFN-stimulated genes
   (ISGs)[333]^49 in clusters 1–3 (exhausted T cells, see [334]Figure 1A)
   ([335]Figure 5A) with higher expression of numerous ISGs in CD8^+ T
   cells from ICB-NR compared to ICB-R patients ([336]Figure 5B). Type I
   IFN has well-established anti-tumor properties and can be secreted by
   tumor cells, macrophages, and dendritic cells in the
   TME.[337]^45^,[338]^50 Analysis of the expression of type I IFN
   signaling machinery (IFNAR1/2, STAT1/2, and IRF9)[339]^51 in CD45^+
   cell clusters in human melanoma showed high expression of these genes
   in macrophages and dendritic cells (clusters 3 and 4, see [340]Figure
   S1A) and T cells (clusters 6, 7, and 11, see [341]Figure S1A)
   ([342]Figure S8A). More importantly, this signature was associated with
   resistance to ICB ([343]Figure S8B).

Figure 5.

   [344]Figure 5
   [345]Open in a new tab

   Type I interferon induces CD38 expression in T cells

   (A and B) Expression of type I IFN-stimulated genes in CD8^+ TILs from
   human melanoma (A) by cluster and (B) by response to ICB.[346]^8

   (C–E) Staining of CD38 in human CD8^+ TILs (n = 3; in C, two-sided
   paired t test; in E, two-way ANOVA with Sidak correction for multiple
   comparisons). Means (bars) and individual values (open circles) are
   shown.

   (F) Analysis of relative NAD(H) in control or IFN-β-treated CD8^+ TILs
   (n = 3; two-sided unpaired t test). Means (bars) and individual values
   (open circles) are shown.

   (G and H) staining of TIM-3 in CD8^+ TILs in indicated groups (n > 4;
   two-sided paired t test).

   (I and J) Flow-cytometry analysis of (I) MMP and (J) mitochondrial mass
   of CD8^+ TILs ± IFN-β (n = 3; two-sided unpaired t test). Means (bars)
   and individual values (open circles) are shown.

   ∗p < 0.05; ns, not significant. See also [347]Figure S8.

   To test the hypothesis that exposure of CD8^+ TILs to type I IFN in the
   TME can reinforce the expression of CD38 and drive T cell dysfunction,
   CD8^+ TILs expanded from melanoma patients were treated with IFN-β
   (type I IFN)[348]^50 with and without TCR activation. CD38 expression
   was highly induced following IFN-β treatment alone and was increased
   further when combined with TCR activation ([349]Figures 5C, 5D, and
   [350]S8C). Interestingly, low-dose IFN-β was sufficient to induce
   significant elevation of CD38-positive cells ([351]Figure S8D).
   Further, the effect was specific to type I IFN, as IFN-γ (type II) did
   not induce CD38 upregulation ([352]Figure S8E). Ruxolitinib, a JAK1/2
   inhibitor,[353]^52 prevented upregulation of CD38 expression observed
   following stimulation of CD8^+ TILs with IFN-β ([354]Figure 5E),
   confirming a key role of IFN sensing in the T cell-specific
   upregulation of CD38. Consistent with the observations in TILs, IFN-β
   induced CD38 expression in a JAK1/2-sensitive manner in B7-H3.CAR-T
   cells, which was not observed following IFN-γ treatment ([355]Figures
   S8F and S8G). In both CD8^+ TILs and CAR-T cells, IFN-β-driven CD38
   upregulation resulted in decreased cellular NAD^+ ([356]Figures 5F and
   [357]S8H) and was accompanied by increased surface expression of PD-1
   and TIM-3 ([358]Figures 5G and [359]S8I), while CD38i treatment blunted
   IFN-β-driven upregulation of TIM-3, indicating a less exhausted T cell
   phenotype ([360]Figures 5H and [361]S8J). CD38 expression itself was
   not affected by CD38i treatment ([362]Figure S8K). Along with the loss
   of cellular NAD^+, we observed increased MMP and mitochondrial mass
   following IFN-β treatment with chronic TCR activation compared to
   chronic stimulation alone ([363]Figures 5I, 5J, [364]S8L, and S8M).
   These mitochondrial changes required longer IFN-β stimulation, together
   with TCR activation, to occur (over 6 days) and were not observed after
   short-term IFN-β stimulation (3 days) ([365]Figures S8N and S8O), as
   was shown in T cells from patients with systemic lupus
   erythematosus.[366]^48 These results indicate that both chronic TCR
   stimulation and type I IFN signaling can drive CD38 expression,
   resulting in increased NAD^+ consumption and an aberrant mitochondrial
   phenotype.

CD38 blockade overcomes ICB resistance

   Given that upregulation of CD38 in CD8^+ T cells is associated with T
   cell dysfunction and ICB resistance and that disrupting CD38 is able to
   restore T cell effector function, we next examined CD38 blockade as a
   strategy to overcome ICB resistance. CD38 blockade was previously shown
   to enhance in vivo response to PD-1 blockade in murine tumor
   models.[367]^13^,[368]^37 Yet clinical trials of dual PD-(L)1 blockade
   plus CD38 blockade have failed to demonstrate meaningful clinical
   responses in NSCLC, prostate cancer, and multiple
   myeloma,[369]^53^,[370]^54^,[371]^55 for which CD38 targeting was
   largely directed at the cancer cells themselves. Importantly, the
   effect of dual PD-1/CD38 blockade in ICB-resistant melanoma, where CD38
   expression in CD8^+ T cells correlates with lack of response to ICB
   ([372]Figures 1F and 1J), has not been tested.

   To examine the therapeutic potential of CD38 blockade in enhancing
   tumor immunity and overcoming resistance to ICB using a clinically
   relevant model of human melanoma, we performed ex vivo profiling of
   PDOTS.[373]^14^,[374]^15^,[375]^56 PDOTS, unlike traditional organoid
   models, are short-term 3D cultures of multicellular spheroids derived
   from freshly explanted patient tumors and retain tumor-infiltrating
   immune and stromal cells, including CD8^+ TILs.[376]^15 PDOTS derived
   from patients with advanced melanoma (n = 27) were treated with
   anti-PD-1 (pembrolizumab) ± anti-CD38 (daratumumab), compared to
   untreated controls ([377]Figure 6A and [378]Table S10). Following 5–7
   days of treatment, PDOTS viability was analyzed using a
   microscopy-based viability assay.[379]^15^,[380]^57 We observed a
   significant reduction in PDOTS viability following a dual PD-1/CD38
   blockade, which demonstrated superior anti-tumor activity compared to
   anti-PD-1 or anti-CD38 alone ([381]Figures 6B, 6C, and [382]S9;
   [383]Table S10). Examination of patient-specific responses to PD-1 ±
   CD38 blockade revealed an 18% response rate (5/27) to anti-CD38
   treatment and a 55% response rate to dual PD-1/CD38 blockade (15/27),
   compared to a 7.4% response rate to single-agent PD-1 blockade (2/27)
   ([384]Figure 6C). Exceptional ex vivo response to dual PD-1/CD38
   blockade was observed in several PDOTS from patients with clinical ICB
   resistance (60% of responders, 9 of 15), including PDOTS 10101 and
   10213 derived from patients with ICB-resistant cutaneous and mucosal
   melanoma, respectively ([385]Figures 6D–6F and [386]S10A). Importantly,
   53% of the PDOTS (8/15) that responded to dual PD-1/CD38 blockade did
   not respond ex vivo to PD-1 or CD38 blockade alone (e.g., PDOTS 10095,
   10101, and 10190) ([387]Figures 6C and 6E). Dual PD-1/CD38 blockade was
   also active in ICB-resistant Merkel cell carcinoma[388]^58 and showed
   significant tumor killing with the combination ([389]Figure S10B). For
   select PDOTS specimens, we were able to show that dual PD-1/CD38
   blockade was more effective than dual PD-1/CTLA-4 in reducing tumor
   growth ([390]Figure S10C). Multispectral flow-cytometry analysis (prior
   to PDOTS culture) and immunofluorescence staining of PDOTS “on-chip” in
   the 3D microfluidic culture confirmed the expression of CD38 and PD-1
   on tumor-infiltrating immune cells ([391]Figures S10D and S10E). Post
   hoc multivariate analysis of multispectral flow-cytometry data,
   mutational status, and selected clinical features failed to identify
   any statistically significant association with PDOTS response
   ([392]Figures S11A–S11C).

Figure 6.

   [393]Figure 6
   [394]Open in a new tab

   CD38 blockade overcomes ICB resistance

   (A) Scheme of PDOTS preparation.

   (B) Viability assessment of melanoma PDOTS (n = 27) following treatment
   with anti-PD-1 (pembrolizumab), anti-CD38 (daratumumab), or the
   combination. Individual values (open circles) indicate the mean for
   each PDOTS specimen; one-way ANOVA with Greenhouse-Geisser correction
   for multiple comparisons.

   (C) Waterfall plot of melanoma PDOTS (n = 27). Response is defined as a
   30% reduction from control (lower dashed lines); growth is defined as a
   20% increase from control (upper dashed lines). Viability percentages
   of controls are in [395]Table S10.

   (D and E) PDOTS viability assessment with indicated treatments (n = 3
   biological replicates per PDOTS specimen, one-way ANOVA with Tukey
   correction for multiple comparisons). Means (bars) and individual
   values (open circles) are shown.

   (F) Representative images of PDOTS in (E). HO, Hoechst (blue); PI,
   propidium iodide (dead cells [red]). Scale bars, 100 μm.

   (G) scRNA-seq analysis of CD8^+ T cells (n = 39,621) from tumors in (B)
   and (C) (n = 21), showing CD38 expression.

   (H) CD38 expression in indicated clusters (n = 21, one-way ANOVA with
   Tukey correction for multiple comparisons).

   (I) GSEA of CD38^+ PD-1 blockade responding tumors in Prolif.CD8 and
   CXCL13^+ T cell clusters (n = 12).

   (J) Module score of IFNG and GZMA in Prolif.CD8 and CXCL13^+ T cells
   discriminate PDOTS responsive (R) and non-responsive (NR) to dual
   PD-1/CD38 blockade (R, n = 12; NR, n = 8; individual graphs are in
   [396]Figures S11H and S11I).

   (K and L) Viability assessment of (K) B16-ova MDOTS (n = 6 biological
   replicates, two independent experiments) and (L) CT26-GFP MDOTS treated
   with indicated treatments (n = 12 biological replicates, four
   independent experiments, one-way ANOVA with Tukey correction for
   multiple comparisons). Means (bars) and individual values (open
   circles) are shown.

   ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001; ns, not
   significant. See also [397]Figures S9–S12; [398]Tables S10, [399]S11,
   [400]S12, and [401]S13.

   To identify CD8^+ T cell-intrinsic features associated with response or
   resistance to dual PD-1/CD38 blockade, scRNA-seq analysis was performed
   on CD8^+ T cells from matched tumor digests for selected PDOTS
   sensitive (n = 12) and resistant (n = 9) to dual PD-1/CD38 blockade
   ([402]Table S10). No differences were observed in the proportions of
   the four identified CD8^+ T cell clusters or in the expression of CD38
   and TCF7 between PDOTS responsive or resistant to dual PD-1/CD38
   blockade ([403]Figures 6G and [404]S11D–S11F; [405]Table S11). The
   proliferating T cell cluster (Prolif.CD8^+) and the CXCL13^+ T cell
   cluster exhibited the highest expression of CD38 ([406]Figures 6G and
   6H), the latter of which has previously been shown to be enriched for
   tumor-reactive T cells.[407]^59^,[408]^60 Gene sets associated with
   effector T cell function,[409]^61 including TCR signaling, antigen
   presentation, T cell co-stimulation, and mitosis, were enriched in
   Prolif.CD8 and CXCL13^+ T cell clusters from tumor explants responsive
   to dual PD-1/CD38 blockade ([410]Figure 6I and [411]Table S12).
   Further, genes associated with cytotoxic potential (GZMA,
   IFNG),[412]^61 were up-regulated in Prolif.CD8 and CXCL13^+ T cells
   from tumor explants responsive to dual PD-1/CD38 blockade ([413]Figure
   S11G and [414]Table S13). Accordingly, the module score of these two
   genes in Prolif.CD8 and CXCL13^+ T cells could accurately discriminate
   responding from non-responding samples with an accuracy score of 85%
   (17 samples out of 20) ([415]Figures 6J, [416]S11H, and S11I;
   [417]Table S13). Interestingly, CD8^+ T cells from matched tumor
   digests from PDOTS resistant to dual PD-1/CD38 blockade exhibited
   higher expression of genes associated with Ca^2+ signaling and related
   pathways (RYR2, PLCG2, and AKT2)[418]^37^,[419]^62 and amphiregulin
   (AREG), which has been shown to promote the immunosuppressive effects
   of regulatory T cells[420]^63 ([421]Figure S11G and [422]Table S13).
   These results suggest that while specific CD8^+ T cell states and genes
   (e.g., CD38 and TCF7) do not clearly associate with sensitivity to dual
   PD-1/CD38 blockade, relative enrichment of CD38-expressing,
   tumor-reactive CD8^+ T cells with high expression of effector molecules
   is associated with response.

   To confirm the efficacy of dual PD-1/CD38 blockade in ICB-resistant
   murine tumor models, we performed ex vivo profiling of B16-ova and
   CT26-GFP murine-derived organotypic tumor spheroids (MDOTS).[423]^15 To
   model a more immunosuppressive microenvironment with reduced ICB
   sensitivity,[424]^64 tumors were harvested for MDOTS profiling on days
   10–14 when the proportion of CD38^+CD8^+ T cells exceeded 50%
   ([425]Figures 2C and [426]S12A). MDOTS profiling revealed a modest
   response to single-agent PD-1 or CD38 blockade but showed significantly
   improved tumor control with combined PD-1/CD38 blockade in both models
   ([427]Figures 6K and 6L), supporting the effectiveness of CD38 blockade
   in overcoming ICB resistance. Lastly, using B16-ova MDOTS ± anti-CD8b
   treatment, we confirmed that CD8^+ T cell activity was required for the
   combinatorial effect of dual PD-1/CD38 blockade ([428]Figure S12B).
   Taken together, these findings demonstrate that CD38 blockade can
   overcome ICB resistance in human and murine tumor models.

Restoring NAD^+ to overcome ICB resistance

   CD38 blockade with anti-human CD38 (e.g., daratumumab) or anti-mouse
   CD38 antibodies can partially inhibit CD38 enzymatic activity but also
   results in enzyme internalization.[429]^65 To determine whether
   pharmacological inhibition of CD38 enzymatic activity and restoration
   of cellular NAD^+ pools is sufficient to enhance sensitivity to PD-1
   blockade, we performed MDOTS profiling of explanted B16-ova and
   CT26-GFP tumors treated with or without CD38i. CD38i exhibited modest
   single-agent activity but significantly increased the response to PD-1
   blockade in both B16-ova and CT26-GFP MDOTS ([430]Figures 7A–7C).
   Consistent with this observation, NR supplementation[431]^36 enhanced
   the response to PD-1 blockade to a similar extent as disrupting CD38 in
   CT26-GFP MDOTS, while NR by itself had a modest effect on MDOTS
   viability more similar to CD38 blockade alone ([432]Figures 7D and
   [433]S12C). Further, NAD^+ depletion using FK866[434]^35 (NAMPTi)
   blunted the effect of dual PD-1/CD38 blockade ([435]Figure S12D),
   confirming a key role for the CD38-NAD^+ axis in mediating response to
   ICB. Further evaluation of CD38i ± PD-1 blockade in an additional
   cohort of melanoma PDOTS showed that similar to daratumumab treatment,
   inhibition of CD38 enzymatic activity is sufficient to sensitize
   ICB-resistant melanoma PDOTS to PD-1 blockade ([436]Figure 7E and
   [437]Table S14), an effect that could be further enhanced with NR
   supplementation ([438]Figure 7F and [439]Table S14). Taken together,
   these findings demonstrate the therapeutic potential of targeting CD38
   in CD8^+ T cells in melanoma to overcome ICB resistance and the role of
   chronic TCR and type I IFN signaling in its upregulation while
   underscoring the importance of NAD^+ as a vital metabolite for
   maintaining and restoring T cell function ([440]Figure 7G).

Figure 7.

   [441]Figure 7
   [442]Open in a new tab

   Disrupting CD38 restores NAD^+ and overcomes ICB resistance

   (A and B) Viability assessment of (A) B16-ova MDOTS (n = 6 biological
   replicates, two independent experiments) and (B) CT26-GFP MDOTS (n = 12
   biological replicates, four independent experiments). One-way ANOVA
   with Tukey correction for multiple comparisons. Means (bars) and
   individual values (open circles) are shown.

   (C) Representative images of MDOTS in (B). HO, Hoechst (blue); PI,
   propidium iodide (dead cells [red]); tumor, GFP-tumor cells.

   (D) Viability assessment of CT26-GFP MDOTS with indicated treatments (n
   = 6 biological replicates, two independent experiments, one-way ANOVA
   with Tukey correction for multiple comparisons). Means (bars) and
   individual values (open circles) are shown.

   (E) Viability assessment of melanoma PDOTS with indicated treatments.
   (n = 18, six independent specimens; mixed-effects one-way ANOVA with
   Tukey correction for multiple comparisons).

   (F) Viability assessment of melanoma PDOTS treated with indicated
   treatments (n = 3 biological replicates; one-way ANOVA with Tukey
   correction for multiple comparisons). Means (bars) and individual
   values (open circles) are shown.

   (G) Scheme demonstrating the effect of targeting CD38 in T cells by
   increasing NAD^+ and TCF7 expression along with restoring mitochondrial
   function and overcoming resistance to ICB.

   ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001; ns, not
   significant. See also [443]Figure S12 and [444]Table S14.

Discussion

   Here, we show that CD38^+CD8^+ T cells accumulate during tumor
   progression and, following ineffective ICB treatment, exhibit features
   of T cell exhaustion and correlate with ICB resistance. Disrupting CD38
   restores cellular NAD^+ pools and TCF7 expression, thereby improving T
   cell function and sensitizing tumors to ICB, underscoring the
   importance of NAD^+ as a vital metabolite necessary for sustaining T
   cell-mediated tumor immunity.

   CD38-mediated ADPR generation was recently shown to contribute to T
   cell dysfunction via AKT-mediated repression of TCF7
   expression.[445]^37 Interestingly, NAD^+-dependent signaling was also
   shown to regulate TCF7 via the NAD^+/SIRT1/FOXO1 pathway.[446]^66
   Following CD38i treatment of chronically stimulated T cells, we
   observed increased NAD^+ levels with a concomitant reduction in ADPR.
   Thus, upregulation of CD38 can regulate T cell function by
   simultaneously depleting a driver of TCF7 expression (NAD^+) while
   promoting ADPR formation that represses TCF7 expression via increased
   intracellular Ca^2+ levels, RyR2 calcium channel activation, and
   downstream AKT activation.[447]^37 Interestingly, increased expression
   of genes associated with Ca^2+ signaling (e.g., PLCG2, RYR2, and AKT2)
   was observed in CD8^+ T cells from PDOTS matched tumor digests
   resistant to dual PD-1/CD38 blockade. Importantly, we showed that CD38
   can be localized in small cytoplasmic aggregates that are enriched in
   chronically stimulated T cells, particularly those expressing PD-1.
   Those aggregates suggest an increased intracellular activity of CD38 in
   exhausted PD-1^+ T cells that can both decrease intracellular NAD^+ and
   enhance intracellular Ca^+2 signaling. Further studies are required to
   elucidate the intersecting metabolic pathways governed by CD38 and the
   effect of its cellular locations on T cell function.

   Chronic TCR stimulation and type I IFN[448]^50 promote upregulation of
   CD38, leading to mitochondrial dysregulation and T cell dysfunction.
   The role of type I IFN in CD38 upregulation in CD8^+ TILs and human
   CAR-T cells can provide another link between type I IFN signaling and T
   cell exhaustion and dysfunction[449]^48 as well as a better
   understanding of the dual role of type I IFN in cancer, providing
   tumor-suppressive effects as well as immune suppression and tumor
   progression in different contexts.[450]^47 Interestingly, recent
   clinical trials in patients with Hodgkin lymphoma and NSCLC have shown
   that JAK inhibitors can restore ICB sensitivity with an associated
   increased proportion of effector-memory T cells with a concomitant
   reduction in exhausted T cells.[451]^52^,[452]^67

   Resistance to ICB in melanoma and other cancers remains a significant
   clinical challenge.[453]^3 The observed efficacy of dual PD-1/CD38
   blockade, particularly in PDOTS from ICB refractory melanoma patients,
   supports further clinical development of CD38-directed therapeutic
   strategies to overcome ICB resistance. As CD38 monoclonal antibodies
   are approved by the Food and Drug Administration for other
   indications,[454]^33 there is a clear path to rapid translation of our
   observations to evaluate dual PD-1/CD38 blockade in patients with
   ICB-refractory advanced melanoma. Based on recent findings[455]^68 and
   our observations using B7-H3.CAR-T cells, targeting CD38/NAD^+
   signaling may also be an appealing approach to prevent the development
   of CAR-T cell dysfunction. Thus, disrupting CD38 represents an emerging
   approach to enhance response to ICB and cellular cancer immunotherapy.

Limitations of the study

   In this study, we tested the efficacy of dual PD-1/CD38 blockade using
   ex vivo profiling of melanoma PDOTS, many of which were derived from
   patients with clinical ICB resistance. Despite the conceptual
   advantages of using an established ex vivo immunocomponent human model,
   PDOTS are short-term (5–7 days) cultures that incompletely recapitulate
   every aspect of the tumor-immunity cycle, such as recruitment of
   peripheral immune cells from the lymph nodes and blood. As such, the
   biological consequences of PD-1 ± CD38 blockade in PDOTS reflect the
   impact of targeting these molecules within the model TME and within the
   time frame of the assay. A second limitation of the study is that we
   were not able to dissect the specific contributions of type II
   (extracellular-facing catalytic domain) and type III
   (intracellular-facing catalytic domain) CD38 using commercially
   available reagents. As disrupting CD38 exclusively influenced
   intracellular NAD^+ levels, this suggests that the catalytic domain of
   CD38 in exhausted T cells is in a type III orientation. Future studies
   using metabolic labeling, select disrupting of downstream metabolic
   pathways, and novel chemical probes are expected to provide new
   insights into the subcellular compartmentalization and downstream
   metabolic consequences of disrupting CD38 in exhausted T cells.

Resource availability

Lead contact

   Any information and requests for resources and reagents should be
   directed to and will be fulfilled by the lead contact, Russell W.
   Jenkins (rjenkins@mgh.harvard.edu).

Materials availability

   Human melanoma cell lines will be made available on request, but we may
   require a payment and/or a completed materials transfer agreement
   especially if there is potential for commercial application. There are
   restrictions to the availability of human melanoma cell lines because
   of the lack of an external centralized repository for their
   distribution and our need to maintain the stock. We are glad to share
   cell lines with reasonable compensation by the requestor for its
   processing and shipping.

Data and code availability

     * •
       This paper analyzed publicly available scRNA-seq available at GEO:
       [456]GSE120575
       , [457]GSE173351, [458]GSE225713, and [459]GSE217160. New scRNA-seq
       data have been deposited at GEO: [460]GSE296471 and [461]GSE296926
       and are publicly available as of the date of publication.
       Metabolomics data have been deposited at NMDR: ST003779 and
       ST003830 and are publicly available as of the date of publication.
     * •
       All original code has been deposited at Zenodo at DOIs 15108338,
       15120392, 15164848, 15107287, 15115199, and 15381446 and are
       publicly available as of the date of publication.
     * •
       Any additional information required to reanalyze the data reported
       in this paper is available from the [462]lead contact upon request.

Acknowledgments