Graphical abstract

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

Highlights

     * •
       BRAFi+MEKi triggers glioma cell state transitions that foster
       immune evasion
     * •
       BRAFi+MEKi activates interferon response and simultaneously
       suppresses T cells
     * •
       High PD-L1 expression in BRAF-mutant GBM provides a criterion for
       anti-PD-1 therapy
     * •
       Concurrent BRAF/MEK and checkpoint inhibition enhance anti-tumor
       immunity and survival
     __________________________________________________________________

   Xing et al. report that combined BRAF and MEK inhibition (BRAFi+MEKi)
   in BRAF^V600E-mutant high-grade glioma shifts tumor cell states and
   upregulates PD-L1, via Galectin-3 secretion, contributing to T cell
   suppression. Concurrent but not sequential immune checkpoint inhibition
   in mice overcomes these tumor-intrinsic adaptations, highlighting its
   translational promise for glioma therapy.

Introduction

   Brain tumors cause the most years of life lost among cancers and are
   the leading cause of cancer-related death in
   children.[79]^1^,[80]^2^,[81]^3^,[82]^4 Unlike other cancers, targeted
   therapies for brain tumors remain limited, with notable exceptions
   including BRAF (v-raf murine sarcoma viral oncogene homolog B1) and IDH
   (isocitrate dehydrogenase) inhibitors now clinically used for BRAF- and
   IDH-mutant gliomas, respectively.[83]^5^,[84]^6 BRAF, a kinase in the
   mitogen-activated protein kinase (MAPK) pathway, links growth factor
   signals to transcriptional changes, thereby regulating cell
   proliferation, survival, and differentiation.[85]^7 It is the most
   mutated kinase in cancer, with the V600E mutation comprising 90% of
   BRAF alterations.[86]^8 BRAF^V600E occurs in 7% of pediatric and 4% of
   adult primary brain tumors, particularly epithelioid glioblastoma
   (GBM), pleomorphic xanthoastrocytoma (PXA), and
   ganglioglioma.[87]^9^,[88]^10^,[89]^11 In pediatric low-grade gliomas
   (pLGGs), BRAF^V600E confers poor outcomes with conventional therapies
   and increases the risk of malignant progression to high-grade gliomas
   (HGGs), especially with CDKN2A
   deletion.[90]^10^,[91]^12^,[92]^13^,[93]^14^,[94]^15 The BRAF^V600E
   mutation results in constitutive MAPK activation and loss of feedback
   inhibition, promoting oncogenic transformation.[95]^7^,[96]^16^,[97]^17

   ATP-competitive BRAF monomer inhibitors (BRAFi)—vemurafenib,
   dabrafenib, and encorafenib—effective in
   melanoma,[98]^18^,[99]^19^,[100]^20^,[101]^21^,[102]^22 were repurposed
   for patients with BRAF^V600E-mutant glioma alongside broader BRAF^V600E
   screening.[103]^7 However, BRAFi monotherapy shows limited efficacy due
   to multiple adaptive resistance mechanisms, including MAPK pathway
   reactivation.[104]^7 Sustained MAPK suppression correlates with
   clinical benefit,[105]^23 leading to strong preclinical and clinical
   support for combining BRAF with mitogen-activated protein kinase kinase
   (MEK) inhibitors (MEKi).[106]^24^,[107]^25^,[108]^26 Combined BRAF and
   MEK inhibition (BRAFi+MEKi) outperforms chemoradiation in LGG and shows
   encouraging response rates (∼47% in pLGG, ∼54% in adult LGG, and ∼33%
   in adult HGG) with reduced toxicity compared to
   monotherapy.[109]^27^,[110]^28^,[111]^29 In 2022, dabrafenib plus the
   allosteric MEKi trametinib received histology-agnostic, accelerated
   Food and Drug Administration approval for patients with unresectable or
   metastatic BRAF^V600E/K-mutant solid tumors who have progressed
   following prior treatment and lack satisfactory
   alternatives.[112]^30^,[113]^31 In 2023, approval expanded to pediatric
   patients with BRAF^V600E-mutant LGG who require systemic
   therapy.[114]^6^,[115]^32

   Despite promising objective response rates (ORRs), most
   BRAF^V600E-mutant gliomas escape BRAFi+MEKi, with residual tumor
   driving rebound, progression, resistance, and unbridled
   mortality.[116]^7^,[117]^27^,[118]^28^,[119]^33^,[120]^34^,[121]^35^,[1
   22]^36^,[123]^37^,[124]^38^,[125]^39 Resistance mechanisms, well
   characterized in melanoma, remain understudied in gliomas.
   Nevertheless, both tumor types exhibit intrinsic, acquired, and
   adaptive resistance to BRAFi, with shared resistance mechanisms
   including MAPK reactivation[126]^7^,[127]^33^,[128]^36^,[129]^40 and
   MAPK-independent mechanisms such as activation of the PI3K/AKT/mTOR
   pathway.[130]^26^,[131]^33^,[132]^40^,[133]^41 Unlike melanoma,
   BRAF^V600E-mutant HGGs exhibit a higher intrinsic (primary) resistance
   often due to concomitant genetic alterations,[134]^37 like CDKN2A
   deletion and PI3K-mTOR mutations.[135]^42^,[136]^43 High intra-tumoral
   heterogeneity contributes to this intrinsic resistance, with
   polo-like-kinase 1 (PLK1)-high subpopulation characterized by prolonged
   G2-M phase and increased asymmetric cell division, driving
   recurrence.[137]^44 Hence, preclinical studies have explored combining
   BRAF^V600E inhibitors with PLK1, MEK, mammalian target of rapamycin
   (mTOR), or epidermal growth factor receptor (EGFR)
   inhibitors,[138]^40^,[139]^44^,[140]^45^,[141]^46^,[142]^47^,[143]^48
   although only a few combinations have entered clinical trials
   ([144]NCT04201457 and [145]NCT03919071).

   Intrinsic resistance could also be attributed to the highly
   immunosuppressive glioma tumor microenvironment (TME), making
   immunotherapy extremely challenging, in contrast to melanoma’s
   immunologically active TME characterized by robust T cell
   infiltration.[146]^49^,[147]^50^,[148]^51^,[149]^52 Programmed cell
   death protein-1 (PD-1) and cytotoxic T-lymphocyte antigen-4 (CTLA-4)
   immune checkpoints, exploited by tumors to evade immunity, offer
   therapeutic targets for immune checkpoint inhibitors
   (ICIs)[150]^49^,[151]^50^,[152]^51 and were effective in
   BRAF^V600E-mutant metastatic melanoma.[153]^52^,[154]^53 However, ICIs
   have demonstrated limited efficacy in HGGs[155]^54^,[156]^55
   althoughencouraging responses in replication-repair-deficient HGG and
   evidence of immune synergism with PD-1 blockade and MEKi (trametinib)
   in a small number of patients have been reported.[157]^56 Moreover,
   ERK1/2 phosphorylation predicts survival following anti-PD-1
   immunotherapy in recurrent GBM.[158]^57

   Considering the variability in BRAFi+MEKi responses across
   tumors,[159]^37 and the distinct intra- and inter-tumoral heterogeneity
   and TME of BRAF^V600E-mutant HGGs, their treatment adaptations require
   deeper investigation.[160]^58^,[161]^59^,[162]^60 Evidence suggests
   that therapies tap into tumor plasticity, forcing cells to transition
   to distinct cell states that enable survival and resistance of solid
   tumors.[163]^61 Understanding BRAFi+MEKi-induced adaptations is
   critical yet largely unexplored in gliomas due to limited patient
   cohorts; therefore, mechanisms of cell state transitions in response to
   treatment in BRAF^V600E-mutant glioma remain unknown. Preclinical
   models are essential tools to investigate potential mechanisms of
   glioma cell adaptations and plasticity in BRAF^V600E-mutant HGGs.

   In this study, we investigated changes in HGG cell states following
   BRAFi+MEKi treatment and uncovered cell plasticity as a potential
   adaptive resistance mechanism of BRAF^V600E-mutant HGG. BRAFi+MEKi
   heightened cell state transitions along glial differentiation
   trajectories, giving rise to astrocyte (AC)- and oligodendrocyte
   (OL)-like states. PD-L1 upregulation in OL-like cells linked cell state
   transitions to immune evasion, possibly orchestrated by Galectin-3.
   BRAFi+MEKi-induced adaptive transcriptional changes included
   upregulation of interferon (IFN)-γ response- and antigen
   presentation-associated gene signatures, suggesting enhanced T
   cell-mediated anti-tumor immunity and potentially linking cytotoxicity
   to therapy response. Therapy-resistant murine tumors exhibited
   diminished cytotoxicity and impaired antigen presentation to T cells,
   indicative of immune evasion. This effect can be reversed by
   concurrently combining BRAFi+MEKi with ICI treatment, leading to
   significantly improved survival in a T cell-dependent manner. PD-L1
   expression was consistently elevated in BRAF-mutant versus
   BRAF-wild-type GBM patient cohorts, suggesting a pervasive immune
   evasion mechanism. Collectively, our parallel investigations of patient
   tumor samples, high-fidelity murine models, and patient-derived cell
   lines offer insights into BRAFi+MEKi modulation of tumor cell states
   and the tumor immune microenvironment and may hold promise for a
   rational strategy combining BRAFi+MEKi with ICIs to improve outcomes in
   BRAF^V600E-mutant HGGs.

Results

BRAF^V600E-mutant HGGs undergo cell state transitions along glial
differentiation trajectories following extensive treatment in patients

   Cancer treatment induces cellular adaptations via non-genetic
   mechanisms to drive resistance. Adaptive responses to BRAFi+MEKi in HGG
   are poorly understood. To investigate potential cellular adaptations to
   BRAFi+MEKi treatment, we analyzed the transcriptomes of paired
   BRAF^V600E-mutant HGG from patients before dabrafenib + trametinib
   treatment and at recurrence (Figure 1A) using gene ontology (GO)
   analysis of bulk RNA sequencing (RNA-seq). Magnetic resonance imaging
   (MRI) confirmed recurrence (Figure 1B), and immunofluorescence (IF) and
   immunohistochemistry (IHC) were conducted for validation ([164]Figures
   1E-F and [165]S1; [166]Table S1). GO analyses showed upregulation of
   genes engaged in gliogenesis, synapse organization, axonogenesis, and
   cell-cell adhesion among other pathways after BRAFi+MEKi treatment
   ([167]Figures 1C and [168]S1D).

Figure 1.

   [169]Figure 1
   [170]Open in a new tab

   Patient BRAF^V600E-mutant HGG undergoes glial differentiation after
   treatment

   (A) Schematic of longitudinal analyses in two patients.

   (B) MRI scans of patient 49 showing tumor (left), post-surgical cavity
   (middle), and recurrence (right) after BRAFi+MEKi.

   (C) GO analysis showing gliogenesis pathway upregulation at second
   recurrence after BRAFi+MEKi in patient 12.

   (D) Scatterplot of glial and progenitor expression changes before and
   after BRAFi+MEKi in patient 12.

   (E) Representative IF images for GFAP, MBP, and OLIG2 before and after
   BRAFi+MEKi and chemoradiation in patient 49. High-magnification images
   of boxed areas are shown on the right. Scale bars: 180 μm (left) and
   140 μm (right). Quantification of GFAP and MBP intensities and OLIG2+
   cells in patient 49 (n = 5 tumor-containing images). ∗p < 0.05 and ∗∗p
   < 0.01 (Student’s t test). Mean ± SEM.

   (F) Representative H&E- and NESTIN-stained tumor sections before and
   after BRAFi+MEKi in patient 49. High-magnification images of boxed
   areas are shown on the right. Scale bars: 150 μm.

   See also [171]Figure S1 and [172]Table S1.

   We further investigated pathways associated with gliogenesis for
   specific cell states and found that they are significantly enriched for
   markers for neural stem/progenitor cell (NSPC) (PROM1/CD133),
   pre-oligodendrocyte progenitor cell (pre-OPCs) (ASCL1, HES6, EGFR, and
   DLL1), AC (GFAP and AQP4), myelin-forming OL (PROM1/CD133, MBP, and
   PLP1),[173]^62 and—albeit at lower rates—mature OL (ASPA, CNP, and
   CLDN11) states ([174]Figures 1D, [175]S1D, and S1E). Increased
   PROM1/CD133+ expression aligns with previous findings showing a CD133+
   BRAF inhibitor-resistant cell population in BRAF^V600E-mutant
   GBM.[176]^44 PROM1/CD133 and NESTIN expression—typically associated
   with a stemness state in GBM—were not co-regulated since NESTIN
   transcript ([177]Figure 1D) and protein levels ([178]Figure 1F) were
   unchanged after BRAFi+MEKi. Further IF validation confirmed the
   upregulation of selected glial markers at the protein level, showing
   increased GFAP, MBP, and OLIG2 expression in post-treatment samples
   ([179]Figure 1E).

   In conclusion, heavily treated BRAF^V600E-mutant HGG retain NSPC
   markers and transition to AC-, pre-OPC-, and OL-like states, the latter
   reminiscent of pre-myelinating and mature OL-like cells.

Orthotopic BRAF^V600E-mutant mouse models show intra-tumoral cellular
heterogeneity

   To address small patient cohort limitations, particularly for
   BRAFi+MEKi treatment, we generated two BRAF^600E-mutant glioma models
   (BRAF-M34 and RCAS-BRAF), with distinct co-mutations on the C57BL/6
   genetic background, inducing gliomas at 54 and 70 days post-injection,
   respectively ([180]Figures 2A, 2F, [181]S2A, and S2B; [182]Table S2,
   STAR Methods). We subsequently isolated cells from these endogenous
   BRAF-M34 and RCAS-BRAF gliomas and re-injected dissociated cells into
   C57BL/6 mouse brains to generate large orthotopic masses with high
   tumor take rate (BRAF-M34: 95%; RCAS-BRAF: 90%) ([183]Figure 2A;
   [184]Table S2), as visualized by MRI ([185]Figures 2B and 2G).
   Histopathologic analyses of orthotopic masses revealed higher grade
   features, including high nuclear pleomorphism, necrosis, and
   microvascular proliferation, with subtle model-specific differences
   ([186]Figures 2C, 2H, [187]S2C, and S2I; [188]Table S2, STAR Methods).
   IF revealed high intra-tumoral heterogeneity in BRAF-M34 and RCAS-BRAF
   orthotopic glioma tested alongside BRAF-2341, which we previously
   generated on the FVB/N background.[189]^26 Expression of GFAP, OLIG2,
   NESTIN, PROM1/CD133, and OPC markers PDGFRα and NG2 in single tumor
   entities was detected ([190]Figures 2D, 2E, 2I, [191]S2C–S2H, and S2J;
   [192]Table S2). Glial and neuronal markers are no longer lineage
   restricted as in the normal brain but are frequently co-expressed in
   orthotopic gliomas, consistent with human HGGs ([193]Figure
   2I).[194]^63 Reactive ACs (GFAP+) and glioma-associated oligodendrocyte
   progenitor cells (OPCs) (PDGFRα/NG2+) and OLs (nuclear
   OLIG2/PROM1+)[195]^62^,[196]^64^,[197]^65^,[198]^66 identified by lack
   of GFP and CRE expression, were found in all three models as part of
   the TME ([199]Figure 2D, middle panel, [200]S2C, S2D, and S2H).
   PROM1/CD133 was co-expressed with NESTIN, potentially marking a
   stemness state ([201]Figure 2E), and with OLIG2, which encompasses
   highly proliferative OPC-like states in addition to the
   non-proliferative, mature OLs ([202]Figures S2F, S2H, and S2J) and,
   when cytoplasmic, also ACs ([203]Figure S2J).[204]^65 Staining with
   proliferation marker Ki67 revealed that PROM1/CD133-positive cells are
   less proliferative than OLIG2-single-positive cells, suggesting that
   PROM1/CD133 differentiate into OL-like cells.[205]^62 Our two murine
   models (BRAF-M34 and RCAS-BRAF) recapitulate features of human
   BRAF^V600E-mutant gliomas,[206]^10^,[207]^66^,[208]^67 including
   intra-tumoral cellular heterogeneity, with model-specific differences.
   These models provide robust preclinical tools, facilitating
   immunological studies due to their intact immune system and cross-model
   comparisons owing to their identical C57BL/6 genetic background.

Figure 2.

   [209]Figure 2
   [210]Open in a new tab

   Orthotopic BRAF^V600E-mutant mouse models show tumor heterogeneity

   (A) Schematic of BRAF-2341 and BRAF-M34 model generation.

   (B) MRI scans showing tumor mass in BRAF-2341 and BRAF-M34 HGG mice
   (yellow dashed outlines).

   (C) Representative low- and high-magnification images of H&E-stained
   coronal sections of BRAF-2341 and BRAF-M34 HGGs. Scale bars: 80 μm
   (top) and 60 μm (bottom).

   (D) Representative IF images of BRAF-2341. GFP and/or CRE expression
   mark tumor cells. DAPI (cell nuclei). Left: GFP and PROM1/CD133
   double-positive tumor cells (white arrowheads). Middle: OLIG2 and
   PROM1/CD133 double-positive/GFP-negative glioma-associated OLs (white
   arrowheads). Right: GFP and CRE double-positive tumor cells (white
   arrowheads); tumor cells co-express GFAP and OLIG2 (yellow arrowheads)
   or OLIG2 alone (blue arrowhead). Single-channel images of the
   white-boxed area are shown next to the merged images. Scale bars: 60 μm
   (left), 45 μm (middle), and 75 μm (right).

   (E) Representative IF images of BRAF-M34 HGG. Left: CRE, NESTIN, and
   PROM1/CD133 triple-positive NSPC/CSC-like tumor cell (white
   arrowheads). CRE and PROM1/CD133 double-positive (NESTIN negative)
   OL-like tumor cells (yellow arrowheads). Right: NeuN, OLIG2, and GFAP
   single-positive cells (white arrowheads) in the tumor core featuring
   hypercellularity. Single-channel images of the white-boxed area are
   shown next to the merged images. Scale bars: 110 μm (left) and 50 μm
   (right).

   (F) Schematic of RCAS (replication-competent avian retrovirus)-induced
   BRAF^V600E TP53-deleted (RCAS-BRAF) HGG and orthotopic, immunocompetent
   model development.

   (G) MRI scans showing the tumor infiltrating into the contralateral
   hemisphere (yellow dashed outlines).

   (H) Representative image of an H&E-stained coronal section of
   orthotopic RCAS-BRAF HGG showing hypercellular proliferation of
   malignant ovoid to spindled cells arranged in fascicles and sheets.
   Scale bar: 140 μm.

   (I) Representative IF images of RCAS-BRAF HGG. GFAP-positive
   astrocyte-like cells, infrequently double-positive for OLIG2 (green
   arrowhead). An abundance of NESTIN+ cells, frequently double-positive
   for OLIG2 (white arrowheads), and PDGFRα+ cells (white arrowheads).
   Single-channel images of the white-boxed area are shown next to the
   merged images. Scale bars: 90 μm (left) and 115 μm (right).

   See also [211]Figure S2 and [212]Table S2.

BRAFi+MEKi heightenscell state transitions along glial differentiation
trajectories in BRAF^V600E-mutant gliomas

   Our analyses of heavily treated BRAF^V600E-mutant HGG patient materials
   strongly suggested that BRAFi+MEKi treatment boosts cell states
   associated with glial differentiation ([213]Figures 1D and [214]S1D).
   To investigate these changes in additional samples, we assessed whether
   BRAFi+MEKi boosts similar cell state changes in murine
   BRAF^V600E-mutant gliomas. Moreover, we used single-cell RNA sequencing
   (scRNA-seq) of tumors in addition to bulk RNA-seq to analyze these
   changes on a more granular level. Established (BRAF-2341) and newly
   generated (RCAS-BRAF) mouse models were treated with dabrafenib +
   trametinib for up to 14 days, and tumors were analyzed by bulk RNA-seq
   or scRNA-seq and then flow cytometry (FC) and IF for validation
   ([215]Figures 3A and 3C). In BRAF-2341 cells, isolated by flow
   cytometry (FC) sorting for A2B5,[216]^67 BRAFi+MEKi increased the
   expression of transcripts associated with myelin-forming OL- (Plp1,
   Mbp, Cnp, Prom1/CD133, and Cldn11), mature OL- (Aspa), and AC-like
   states (Gfap and Clu) ([217]Figures 3B and [218]S1E). We confirmed that
   BRAFi+MEKi upregulated transcripts associated with AC- and OL-like
   states in two patient-derived BRAF^V600E-mutant glioma cell lines
   (aGBM5 and STN-49), albeit with quantitative differences in certain
   transcripts ([219]Figures S3A and S3B), consistent with their selected
   upregulation in patient tumors ([220]Figures 1D and 1E). Additionally,
   BRAFi+MEKi boosted the expression of transcripts associated with
   myelin-forming OL- (Mbp, Plp1, Pllp, and Prom1) and AC-like states
   (Gfap and Aqp4) in the RCAS-BRAF model, as analyzed by scRNA-seq and
   validated by IF ([221]Figures 3C–3E, [222]S3F, and S3G). GO analysis
   revealed significant upregulation of transcripts involved in
   gliogenesis ([223]Figure 3E), while IF confirmed upregulation of GFAP
   and MBP proteins, associated with glial differentiation (AC and OL)
   ([224]Figures 3F and 3G). Markers for NSPC- (Nes), pre-OPC- (Dll1,
   Egfr, and Hes6), and OPC-like (Pdgfrα and Cspg4) states were
   downregulated by BRAFi+MEKi in BRAF-2341 cells ([225]Figures 3B and
   [226]S1E), suggesting that NSPC-, pre-OPC-, and OPC-like cells are
   sensitive to short-term 3-day treatment. Decreased expression of
   transcripts for NSPC- and OPC-like states (Nes and Cspg4) was also
   detected, whereas some markers for pre-OPC-like states (Ascl1, Hes6,
   and Egfr) were unchanged in the RCAS-BRAF tumors ([227]Figure 3D), and
   the OPC-like state marker PDGFRα protein was reduced, as shown by IF
   ([228]Figure 3G).

Figure 3.

   [229]Figure 3
   [230]Open in a new tab

   BRAFi+MEKi treatment promotes glial differentiation in
   BRAF^V600E-mutant HGG

   (A) Schematic of BRAFi+MEKi treatment timeline and analyses (RNA-seq,
   flow cytometry, [FC[, and immunofluorescence [IF]) of BRAF-2341 HGG.

   (B) Heatmap from bulk RNA-seq data depicting average expression levels
   of stem/progenitor cell and glial differentiation-associated
   transcripts between two replicates in BRAF-2341 HGG after 3-day
   BRAFi+MEKi treatment.

   (C) Schematic of 14-day BRAFi+MEKi treatment timeline and analyses
   (scRNA-seq and IF) of RCAS-BRAF HGG.

   (D) Heatmap from scRNA-seq data showing transcripts in the RCAS-BRAF
   tumor cluster.

   (E) Cnetplot showing enrichment of glial-related transcripts from GO
   analysis of the scRNA-seq dataset from RCAS-BRAF HGG.

   (F) Representative IF images of control or BRAFi+MEKi-treated RCAS-BRAF
   HGGs stained for GFAP, NESTIN, and DAPI. Scale bar: 120 μm.
   Quantification of marker-positive cells within the tumor field (right)
   ∗∗p < 0.01 (unpaired t test). n = 4 mice/group, mean ± SEM.

   (G) Representative IF images of control/BRAFi+MEKi-treated RCAS-BRAF
   HGGs stained for MBP, PDGFRα, and DAPI. Scale bar: 170 μm.
   High-magnification images of the white-boxed areas are shown on the
   right. Quantification of marker-positive cells within the tumor field
   (right). ∗p < 0.05 (unpaired t test). n = 4 mice/group, mean ± SEM.

   (H) Schematic of BRAFi+MEKi treatment timeline and analysis of
   BRAF^V600E HGG cell lines in vitro.

   (I and J) Representative ICC images of murine BRAF-M34 (I) and
   RCAS-BRAF (J) cell lines treated with control/BRAFi+MEKi and stained
   for MBP, OLIG2, NESTIN, and DAPI. Scale bars: 40 μm. ∗p < 0.05
   (unpaired multiple t tests and Holm-Šídák method). n = 2 independent
   experiments, mean ± SEM.

   See also [231]Figure S3.

   In glioblastoma, NESTIN and PROM1/CD133 are markers associated with a
   stemness state that contributes to chemoradiation resistance and drives
   tumor recurrence.[232]^68^,[233]^69 We therefore investigated the
   effects of BRAFi+MEKi treatment on NESTIN and CD133/PROM1 cells in
   mice. Double-positive cell states decreased, as indicated by FC and IF,
   consistent with downregulated Nestin transcript with BRAFi+MEKi
   ([234]Figures S3C–S3E). The abundance of CD133/PROM1+ single-positive
   cells, however, increased while their overall abundance remained
   unchanged ([235]Figures S3C and S3D). This suggests that BRAFi+MEKi
   reduces NESTIN+ but not CD133/PROM1+ cell states; the latter appears to
   be intrinsically resistant to treatment, aligning with increased
   CD133/PROM1 transcript level in patient’s recurrent tumors after
   treatment ([236]Figure 1D) and prior reports of CD133+ cells involved
   in BRAFi resistance.[237]^44

   Lastly, we tested BRAFi+MEKi effects on tumor cell states in vitro by
   treating murine (BRAF-M34 and RCAS-BRAF) and human (STN-49 and aGBM5)
   glioma cell lines with BRAFi+MEKi for 13 days. STN-49, specifically
   generated for this study, was derived from patient 49 before treatment
   ([238]Figures 3H and [239]S3H, [240]STAR Methods). Tumor cells were
   subsequently analyzed by immunocytochemistry (ICC) for markers for
   OL-like (MBP), pan-OPC/OL-like (OLIG2), and NSPC/CSC-like states
   (NESTIN). BRAFi+MEKi reduced tumor cell viability in all cell lines and
   increased MBP+ and OLIG2+ cell abundance ([241]Figures 3I, 3J,
   [242]S3I, and S3J), consistent with a treatment-induced boost in glial
   differentiation states (AC- and OL-like) observed in vivo.

   NESTIN+ cells, abundant at baseline (vehicle treatment) in all four
   cell lines, were unchanged with 13-day BRAFi+MEKi treatment in murine
   cells, which is further corroborated by findings from the RCAS-BRAF
   model ([243]Figure 3F) and patients ([244]Figures 1D and 1F). These
   data indicated that cells with NSPC/CSC-like states initially decreased
   with treatment ([245]Figures 3B, 3D, and [246]S3C–S3E) and were
   reinstated with treatment over time ([247]Figures 3I, 3J, [248]S3I, and
   S3J). Notably, mRNA and protein levels were not concurrent in the
   RCAS-BRAF tumors, and model-specific differences are evident as Nestin+
   cells are more abundant in untreated RCAS-BRAF tumors ([249]Figure 2I)
   compared to BRAF-2341 and BRAF-M34 tumors ([250]Figures 2D and 2E;
   [251]Table S2).

   These findings demonstrate that BRAFi+MEKi treatment temporarily
   suppresses NSPCs and reduces cells in pre-OPC and OPC-like states,
   while consistently heightening transitions toward glial cell
   differentiation states reminiscent of myelin-forming OL and AC in
   BRAF^V600E-mutant gliomas.

BRAFi+MEKi boosts IFN response pathway gene expression signature in
BRAF^V600E-mutant HGG

   BRAF-altered glioma exhibits higher CD8^+ T cell infiltration and major
   histocompatibility complex MHC class I expression, when compared to
   BRAF-wild-type glioma,[252]^63^,[253]^64 suggestive of a robust
   anti-tumor immune response that could be exploited therapeutically in
   combination with BRAFi+MEKi. We therefore investigated whether
   BRAFi+MEKi alters immunomodulatory programs in BRAF^V600E-mutant glioma
   cells, using transcriptome analyses and quantitative
   reverse-transcription PCR (RT-qPCR) for validation. We first analyzed
   the whole transcriptomes of BRAF^V600E-mutant patient-derived (STN-49
   and aGBM5) and murine (BRAF-2341) glioma cell lines for differential
   gene expression and Reactome pathway analysis. We observed that 48-h
   BRAFi+MEKi treatment upregulated IFN-γ receptor-associated signature
   with 67 transcripts across human and mouse transcriptomes ([254]Figures
   4A, 4B, [255]S4A–S4C, and S4F). Transcripts for MHC genes were included
   in this signature, along with other inflammation-related transcripts.
   RT-qPCR confirmed upregulation of selected MHC class I/II genes (HLA-A,
   HLA-B, HLA-DRA, and CIITA) in patient-derived and murine cell lines
   ([256]Figures S4D, S4E, and S4G; [257]Table S3). These data suggested
   that BRAFi+MEKi alters the immunomodulatory activity of HGG cells,
   heightening antigen presentation by MHC class I/II molecules and
   potentially boosting anti-tumor immune responses.

Figure 4.

   [258]Figure 4
   [259]Open in a new tab

   BRAFi+MEKi treatment induces interferon response pathway signature and
   glial differentiation, with a PD-L1-expressing immunomodulatory
   subpopulation

   (A) Reactome pathway analysis revealed upregulated genes associated
   with interferon receptor signaling after 48-h BRAFi+MEKi treatment in
   the STN-49 cells. Numbers on the bar graph indicate the gene count.

   (B) Heatmap demonstrating expression changes of IFN-γ receptor-related
   transcripts after 24- and 48-h BRAFi+MEKi treatment to the STN-49 cell
   line.

   (C) Representative ICC images of murine and patient-derived cell lines
   treated with control/BRAFi+MEKi for 13 days. Scale bars: 40 μm.
   Quantification of (C). ∗p < 0.05 (multiple paired t tests). n = 2
   independent experiments, mean ± SEM.

   (D) Representative IF images of BRAF-M34 HGGs stained for PD-L1, BRAF
   V600E, PLP1, and DAPI. Control panel: BRAF^V600E-mutated tumor cell
   negative for PD-L1 (arrow). PD-L1-positive non-tumor cell (arrowhead).
   BRAFi+MEKi panel: differentiated BRAF^V600E-mutant glioma cells with
   PLP1 and PD-L1 co-expression (white arrows). PLP1-negative,
   PD-L1-positive glioma cells (yellow arrows). PD-L1 only positive cells
   (arrowheads). Scale bars: 70 μm (left) and 80 μm (right).

   (E) Representative IF image of BRAFi+MEKi-treated BRAF-M34 HGG stained
   for PD-L1, Cre, PLP1, and DAPI. Differentiated Cre-positive glioma cell
   co-expressing PLP1 and PD-L1. Scale bar: 20 μm.

   (F) Quantification of (D). Proportions of BRAF^V600E and PLP1
   double-positive differentiated glioma cells with OL-like state positive
   or negative for PD-L1 (left) and proportions of PLP1-positive cells
   co-expressing BRAF^V600E and/or PD-L1 (right) were measured and
   compared between groups. ∗p < 0.05 for BRAF^V600E+PLP1^+PD-L1^+
   triple-positive cells between treatment groups (two-way ANOVA with
   Sidak’s multiple comparisons test). n = 3 mice/group, mean ± SEM.

   (G) Representative IF images of human BRAF^V600E-mutant HGG (patient
   49) tissue before and after BRAFi+MEKi treatment, stained for PD-L1,
   Olig2, and DAPI. Scale bars, 180 μm.

   (H) Quantification of (G). n = 4 images from 2 consecutive sections of
   patient 49 tissue, ∗p<0.05 (unpaired t test), mean ± SEM.

   (I) The scatterplot of upregulated CD274 transcript (PD-L1) after
   BRAFi+MEKi treatment in patient 12.

   See also [260]Figure S4.

BRAFi+MEKi induces glial differentiation cell state transitionsand a
PD-L1-expressing immunomodulatory subpopulation

   IFN-γ signaling induces the PD-1 immune checkpoint pathway, which is
   frequently hijacked by cancer cells to circumvent anti-tumor immune
   responses and is an important immunotherapy target.[261]^70 We
   therefore investigated BRAFi+MEKi effects on PD-1 signaling through
   Reactome pathway analysis of whole transcriptomes, validated by ICC and
   IF. PD-1 signaling was one of the pathways significantly altered by
   BRAFi+MEKi treatment, in human and murine BRAF^V600E-mutant glioma cell
   line transcriptomes ([262]Figures 4A, [263]S4H, and [264]S5A). PD-1
   signaling components, including T cell receptor complex protein CD247,
   and the PD-1 ligand PD-L1 (CD274), were upregulated, as shown by gene
   network analysis ([265]Figure S5B). Next, we rigorously tested whether
   PD-L1 protein and gene expression aligned, using immunostainings.
   BRAFi+MEKi indeed increased PD-L1 protein expression in
   BRAF^V600E-mutant HGG cells in vitro ([266]Figure 4C), and in vivo
   ([267]Figures 4D–4F), consistent with the upregulation of PD-L1
   transcript. We also found upregulated PD-L1 expression at protein and
   transcriptomic levels in BRAF^V600E glioma patients post-BRAFi+MEKi
   treatment ([268]Figures 4G–4I). PD-L1 is also expressed in murine
   BRAF-2341 gliomas, accounting for 83% of NESTIN+ cells ([269]Figure
   S4I).

   Immunomodulatory roles for disease-associated OL-lineage cells have
   been reported, aside from their function in
   myelination.[270]^71^,[271]^72^,[272]^73^,[273]^74 We then investigated
   if OL-like cell states enriched in response to BRAFi+MEKi express the
   immunomodulatory PD-L1, by assessing the co-expression of PD-L1 with OL
   markers PLP1, MBP, and OLIG2, by immunostaining. BRAFi+MEKi
   significantly upregulated PLP1 expression in BRAF-M34 glioma
   (BRAF^V600E+PLP1^+) after 12 days of treatment ([274]Figures 4D–4F),
   consistent with an increase in OL-like cell states in the patient tumor
   ([275]Figure 3I). The vast majority (85.92%) of these OL-like tumor
   cells expressed PD-L1 ([276]Figure 4F, left panel), suggesting that
   treatment induced immunomodulatory function in OL-like glioma cells.
   Interestingly, 40.28% of PLP1^+ BRAF^V600-negative cells were
   immunoreactive for PD-L1 even before treatment (vs. 25.87% after
   treatment), indicating that these non-malignant OLs, also called
   glioma-associated OLs, might be hijacked by glioma cells to promote
   immunosuppression. Co-expression of MBP/Olig2 and PD-L1 was confirmed
   in vitro to complement our in vivo results ([277]Figure S4J). Positive
   fold changes for co-expression in murine and human cell lines are
   variable, suggesting model-specific differences ([278]Figure S4K, left
   panel). Lastly, co-expression of pan-OL marker OLIG2 and PD-L1 in
   BRAF^V600E -mutant patient tumors revealed a 10-fold increase in
   OLIG2/PD-L1 double-positive cells after BRAFi+MEKi treatment
   ([279]Figures 4G and 4H)

   In conclusion, our data provide compelling evidence that BRAFi+MEKi
   treatment induces a transition to differentiating glia cell states,
   with a subpopulation of OL-like glioma cells acquiring immunomodulatory
   properties via PD-L1 expression, establishing a direct mechanistic link
   between cell state transitions and immune evasion.

PD-L1 expression is elevated in BRAF-mutant gliomas

   Our data demonstrated that BRAFi+MEKi upregulated PD-L1 expression, a
   mechanism for T cell suppression and a key criterion for patient
   enrollment in PD-1 inhibition therapy.[280]^75 Therefore, we evaluated
   PD-L1 expression in a large cohort of patients with BRAF-mutant glioma,
   including those with BRAF^V600E. First, we evaluated the association
   between PD-L1 expression and BRAF mutation status in a large cohort of
   adult glioblastoma (IDH-wild-type) specimens (n = 3,126), undergoing
   different characterization analyses ([281]STAR Methods). PD-L1
   expression by IHC was significantly higher in BRAF-mutant (54%) than in
   BRAF-wild-type glioblastoma (17%, p < 0.0001) ([282]Figure 5A). This
   was orthogonally validated using whole-transcriptome sequencing data,
   showing increased PD-L1 (CD274) gene expression in BRAF-mutant (n = 17)
   versus BRAF-wild-type (n = 1,022) glioblastoma (p = 0.045) ([283]Figure
   5B). We found that PD-L1 and BRAFV600E showed a high degree of
   co-expression by IHC ([284]Figure 5C), further corroborating PD-L1
   expression in BRAF^V600E-mutant glioma cells. We also showed that PD-L1
   is expressed in various BRAF^V600E-mutant glioma types, including
   ganglioglioma, PXA, and malignant astrocytoma/glioma, albeit with a
   varied frequency score, as determined by IHC ([285]Figure S5C;
   [286]Table S4).

Figure 5.

   [287]Figure 5
   [288]Open in a new tab

   PD-L1 expression is elevated in BRAF-mutant gliomas

   (A) Bar graph depicting percentage of patients with PD-L1+ tumors in
   the BRAF mutation/fusion versus BRAF-wild-type (no alterations) groups.
   n = 59 patients with BRAF mutation/fusion and 3,126 total patients
   (chi-squared tests).

   (B) Boxplot representing CD274 (PD-L1) transcripts per million in the
   BRAF mutation/fusion (n = 17) versus BRAF-wild-type (no alterations; n
   = 1,022) groups, from whole-transcriptome sequencing (Wilcoxon rank-sum
   test).

   (C) Representative IHC images for BRAF V600E and PD-L1 in consecutive
   sections of BRAF^V600E-mutant HGG. Scale bars: 250 μm.

   See also [289]Table S4.

   Our findings demonstrate that BRAF mutations, including the BRAF^V600E
   variant, are associated with elevated PD-L1 expression in GBM and other
   glioma subtypes. This correlation suggests that patients with
   BRAF-mutant gliomas may harbor suppressed T cells that are
   reactivatable through anti-PD-1 therapy, potentially eliciting a robust
   anti-tumor immune response and driving a significant survival benefit.

BRAFi+MEKi treatment enhances tumor-infiltrating T cells

   The efficacy of T cell checkpoint therapy is significantly influenced
   by the presence of tumor-infiltrating T cells.[290]^76^,[291]^77 Since
   BRAF-altered glioma exhibits robust CD8^+ T cell
   infiltration,[292]^11^,[293]^78^,[294]^79 and our data showed that they
   express PD-L1 ([295]Figures 4C, 4D, 4F, and 4H), we closely assessed
   tumor-infiltrating T cells in BRAF^V600E-mutant murine glioma with and
   without BRAFi+MEKi treatment, using high-dimensional single-cell mass
   cytometry (CyTOF), scRNA-seq, and validation by IF ([296]Figure 6A,
   [297]STAR Methods). We identified robust infiltration of CD4^+ and
   CD8^+ effector T cells (T[eff]), and CD4+FoxP3+ regulatory T cells
   (T[reg]) within the tumor mass, as confirmed by CyTOF ([298]Figures 6B
   and [299]S6A–S6D), scRNA-seq ([300]Figures 6G, 6K and 6L), and IF
   ([301]Figures 6H-J, 6M, [302]S5D, and [303]S6G). The presence of
   tumor-infiltrating T cells in murine BRAF^V600E-mutant gliomas is
   consistent with a heightened frequency of CD3^+CD8^+ T cells in human
   BRAF-altered glioma compared to BRAF-wild-type
   astrocytoma.[304]^78^,[305]^79 BRAFi+MEKi further elevated the
   frequency of CD4^+ and CD8^+ T cells in all three models ([306]Figures
   6B and 6C, 6H–6J, 6M, and S6G), consistent with upregulation of the
   antigen presentation machinery ([307]Figures 4B and [308]S4B–S4G) and
   suggestive of anti-tumor immunity heightened by treatment.

Figure 6.

   [309]Figure 6
   [310]Open in a new tab

   BRAFi+MEKi enhances T cell infiltration and synergizes with ICI to
   improve T cell-dependent survival

   (A) Experimental timeline (green: 12–14 days treatment for CyTOF, IF,
   and scRNA-seq analyses; pink: continuous treatment for survival
   analysis) in all three murine models.

   (B) Frequency of populations of tumor-infiltrating lymphocytes in
   BRAF-2341 tumors by CyTOF analysis. 3 mice/group and two independent
   experiments. ∗∗p < 0.01, ∗∗∗p < 0.001 (two-way ANOVA with Sidak’s
   multiple comparisons test), mean ± SEM.

   (C) Graphs depicting the frequency of PD-1+CD8^+ and PD-1+CD4^+ (left)
   or CTLA-4+CD8^+ and CTLA-4+CD4^+ (right) T cells normalized to control
   by CyTOF analysis. 3 mice/group and two independent experiments. ∗p <
   0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 (two-way ANOVA with
   Tukey’s multiple comparisons test), mean ± SEM.

   (D) Kaplan-Meier survival curve of orthotopic BRAF-2341 (left) or
   BRAF-M34 (right) HGG-bearing mice of all four treatment groups.
   Treatment initiation at day 30 (BRAF-2341) or day 14 (BRAF-M34)
   post-injection (black arrow). Data representative of two independent
   experiments and 5–6 mice/group (BRAF-2341) and 4–5 mice/group
   (BRAF-M34). p values are indicated in the graph and [311]Tables S5 and
   [312]S6 when compared to the control group.

   (E) Kaplan-Meier survival curve in immunocompetent C57Bl/6 mice with
   orthotopic BRAF-M34 HGG treated with BRAFi+MEKi and ICI, either
   concurrently (BRAFi+MEKi+ICI) or sequentially. Treatment initiation at
   day 14 post-injection (black arrow). 4–5 mice/group. p values are
   indicated in the graph and [313]Table S7 when compared to the
   concurrent treatment group.

   (F) Kaplan-Meier survival curve in immunocompetent C57Bl/6 mice with
   orthotopic RCAS-BRAF HGG treated with BRAFi+MEKi and/or ICI for 14
   days. Treatment period from day 22 to day 36 post-injection (shaded
   area). Data representative of two independent experiments and 7–10
   mice/group. p values are indicated in the graph and [314]Table S5 when
   compared to the control group.

   (G) Heatmaps of T cell inhibition- and apoptosis-associated transcripts
   in CD4^+ and CD8^+ T cells in the RCAS-BRAF model.

   (H) Densities of CD4+PD-L1+ and CD8+PD-L1+ T cells in BRAF-M34 and
   RCAS-BRAF (IF analysis). ∗p < 0.05, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001
   (two-way ANOVA with uncorrected Fisher’s LSD). n = 4 mice/group, mean ±
   SEM.

   (I and J) Representative IF images of RCAS-BRAF tissue stained for
   PD-L1, CD4 (I) and CD8 (J). CD4^+ and CD8^+ cells positive for PD-L1
   (arrows); CD4^+ and CD8^+ cells negative for PD-L1 (arrowheads). Scale
   bars: (I) 20 μm (top), 15 μm (bottom), (J) 10 μm (top), and 20 μm
   (bottom). Quantifications of cell populations before versus after
   treatment. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 (two-way
   ANOVA with Sidak’s multiple comparisons test). n = 4 mice/group, mean ±
   SEM.

   (K) Heatmaps of T cell activation-associated transcripts in CD4^+ and
   CD8^+ T cells in the RCAS-BRAF model.

   (L) Heatmaps of MHC class I (green), II (red), and IFN-γ
   (black)-related transcripts associated with antigen presentation in
   CD4^+ and CD8^+ T cells in the RCAS-BRAF model.

   (M) Densities of CD4^+CD69^+ and CD8^+CD69^+ T cells from BRAF-M34 and
   RCAS-BRAF (IF analysis). ∗p < 0.05, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001
   (two-way ANOVA with uncorrected Fisher’s LSD). n = 3–4 mice/group, mean
   ± SEM.

   See also [315]Figures S5 and [316]S6 and [317]Tables S5–[318]S7.

Combination of BRAFi+MEKi with ICI enhances survival outcomes in a T
cell-dependent manner

   Our findings provided a strong rationale for combining BRAFi+MEKi with
   PD-1/PD-L1 ICI to boost T cell activation and improve survival. To
   explore this, we assessed the effects of BRAFi+MEKi combined with ICIs
   on the T cell compartment in our three mouse models ([319]Figure 6A).
   Mice bearing orthotopic BRAF^V600E-mutant gliomas were treated with
   dabrafenib and trametinib (BRAFi+MEKi) and/or anti-PD-L1 antibody.
   Since PD-1/PD-L1 inhibitors alone have shown no survival benefit in
   HGGs,[320]^54^,[321]^55 we also included CTLA-4 blockade
   (α-PD-L1+α-CTLA-4; referred to as ICI) ([322]Figures 6A and [323]S5E).
   Concurrent, quadruple treatment (BRAFi+MEKi+ICI) significantly improved
   survival over BRAFi+MEKi or ICI alone in all three models ([324]Figures
   6D–6F; [325]Tables S5 and [326]S6). This survival benefit is associated
   with T cell reactivation, as shown by reduced frequencies of PD-1- and
   CTLA-4-expressing CD4+/CD8+ T cells ([327]Figure 6C) and T[reg] cells
   ([328]Figures S6A–S6D).

   To assess model-specific differences for BRAF-M34 and RCAS-BRAF
   further, we analyzed the frequency of CD4^+ and CD8^+ T cells
   co-expressing inhibitory (PD-L1/PD-1) and activation (CD69) markers.
   BRAFi+MEKi increased PD-L1+ and PD-1+ T cell frequencies ([329]Figures
   6H–6J, [330]S6F, and S6H), suggesting that the PD-L1/PD-1 checkpoints
   become activated to escape BRAFi+MEKi-induced T cell activation. Only
   PD-1+ T cells in the RCAS-BRAF model showed the opposite trend, most
   likely undergoing apoptosis after treatment ([331]Figures 6G and
   [332]S6H).

   We observed differences in T cell inactivity across the models.
   RCAS-BRAF showed the highest T cell suppression, while BRAF-2341 had
   the least ([333]Figure S6F). This variability is expected and
   consistent with reported model-specific differences and differences in
   the immune responses, even in genetically identical mice ([334]Figures
   6D–6F). Interestingly, both BRAF-M34 and RCAS-BRAF models did not
   respond to ICI alone, suggesting their T cells were unresponsive to
   activation ([335]Figures 6D, 6F, and 6H). However, their T cells were
   responsive to BRAFi+MEKi alone, with the BRAF-M34 model showing
   significantly higher CD4^+CD69^+ and CD8^+CD69^+ T cells, indicative of
   early T cell activation, compared to the RCAS-BRAF model ([336]Figure
   6M). Furthermore, our in vitro data showed that BRAF-M34 glioma cells
   with OL-like states have less immunosuppressive effect on T cells
   ([337]Figure S4K).

   The survival benefits seen with therapeutic antibodies, either alone or
   in combination with BRAFi+MEKi, were completely abolished in athymic
   mice lacking T cells. Similar results were observed in BRAF-M34 mice
   that are depleted of T cells, confirming T cell-dependent anti-tumor
   efficacy ([338]Figure S6E; [339]Tables S5 and [340]S6). Although T
   cells are not strictly required for the anti-tumor effects of
   BRAFi+MEKi, increased T cell activation and anti-tumor immunity may
   contribute to the treatment’s effectiveness, particularly in the
   BRAF-M34 model.

   Lastly, we tested whether sequential therapy optimizes BRAFi+MEKi
   responses, similar to what has been reported for BRAF^V600E-mutant
   melanoma patients.[341]^80 ICI 1^st showed a trend for improved
   survival over BRAFi+MEKi 1^st, which provided no survival benefit over
   BRAFi+MEKi treatment alone. Importantly, concurrent quadruple treatment
   provided the greatest survival benefit and is superior to any
   sequential approach ([342]Figures 6D and 6E; [343]Table S7). These
   findings suggest that patients with BRAF^V600E-mutant gliomas are most
   likely to benefit from a combination treatment regimen integrating
   BRAFi+MEKi and ICIs concurrently.

Synergistic effect of BRAFi+MEKi with ICI on T cell reactivation

   Our data demonstrated that quadruple (BRAFi+MEKi+ICI) treatment
   overcomes resistance observed with dual therapies (BRAFi+MEKi and ICI
   alone) in the RCAS-BRAF model ([344]Figure 6F). To gain insights into
   potential differences in T cell dynamics with dual and quadruple
   treatments, we analyzed CD4+ and CD8+ T cell and tumor clusters in
   scRNA-seq data from the RCAS-BRAF model when mice reached their
   disease-related endpoint ([345]Figure 6A). BRAFi+MEKi moderately
   upregulated selected suppression markers on T cells, including Cd274
   (PD-L1), and robustly upregulated others, such as Cd276, Eomes, Socs,
   Fas, and Fasl in both CD4^+ and CD8^+ T cell clusters ([346]Figure 6G).
   Foxp3 was upregulated in CD4^+ T cells post-treatment, indicating the
   expansion of T[reg] cells. BRAFi+MEKi downregulated activation markers
   in CD4^+ T cells, including Gzmk, Gzmb, IL2ra, and others, while
   upregulating immunosuppressive markers, including Cd27, Cd7, Tnfrsf18,
   and Tnfrsf4 in CD8^+ T cells ([347]Figure 6K), suggesting the presence
   of dysfunctional CD8^+ T cells after BRAFi+MEKi treatment.[348]^81

   We also found that treatment reduced expression of MHC class I and II
   genes in CD4^+ and CD8^+ T cells ([349]Figure 6L). MHC class I gene
   expression was diminished in tumor cells and tumor-associated
   macrophages/microglia ([350]Figures S6J and S6K). This downregulation
   was suggestive of suppressed antigen presentation, potentially enabling
   immune evasion and BRAFi+MEKi resistance at the survival endpoint.

   ICI alone upregulated T[reg] expansion markers, including Foxp3 and
   Btla ([351]Figures 6G and 6K). Moreover, increased expression of
   inhibitory markers like Ctla-4, Lag3, and Pdcd1 ([352]Figures 6G and
   6K) in CD4^+ and CD8^+ T cells, and increased MHC and Ifng expression,
   indicated activation balanced with suppression of T cell activity
   ([353]Figures 6K and 6L). ICI also caused excessive pro-inflammatory
   cytokine production in CD4^+ T cells and tumor cells, resembling a
   “cytokine storm” ([354]Figure S6I).

   In contrast, quadruple treatment induced expression of genes linked to
   T cell activation and effector functions, including Itgae, Gzmk, Cd28,
   Tbx21, Itga1, Cd69, and Tnfrsf9 ([355]Figure 6K). Importantly, MHC
   class I/II gene expression was restored for effective antigen
   presentation without causing excessive cytokine production
   ([356]Figures 6L and [357]S6I). MHC class II gene upregulation in CD8^+
   T cells is associated with memory T cell differentiation ([358]Figures
   6L and [359]S6D).[360]^82 This may contribute to a more durable
   anti-tumor immune response.

   Our findings indicate that BRAFi+MEKi resistance is attributed to
   T[eff]cell suppression, T[reg] expansion, and impaired antigen
   presentation, which collectively could lead to immune evasion in the
   RCAS-BRAF model. While ICI activates T cells, it also induces severe
   cytokine release syndrome (cytokine storm) due to excessive cytokine
   production. In contrast, quadruple treatment effectively ensures full
   activation of CD4^+ and CD8^+ T cells, which could be pivotal for
   overcoming therapy resistance and conferring a significant survival
   benefit.

BRAFi+MEKi-induced Galectin-3 secretion links glial differentiation to PD-L1
upregulation

   To investigate the mechanism for cell state transitions and PD-L1
   expression by BRAFi+MEKi, we assessed the secretome of
   BRAF^V600E-mutant glioma cells in vitro using a high-throughput
   enzyme-linked immunosorbent assay (nELISA). STN-49 and aGBM5 cell lines
   were treated with BRAFi+MEKi over 13 days, mimicking the duration of in
   vivo treatments ([361]Figure 7A, [362]STAR Methods). Molecular factors
   measured by nELISA that were either upregulated or below the detectable
   threshold post-treatment are presented in a heatmap ([363]Figure S7A).
   BRAFi+MEKi significantly elevates the secretion of the
   carbohydrate-binding lectin family protein Galectin-3[364]^83
   ([365]Figure 7B), with levels peaking at day 13, coincidently with
   increased glial differentiation ([366]Figures S3I and S3J). RNA-seq
   confirmed upregulation of the Galectin-3-encoding gene LGALS3 with 48 h
   treatment of BRAFi+MEKi ([367]Figures 7C and 7D). In addition to
   Galectin-3, other factors interleukin (IL)-23 and IL-32 were also
   upregulated post-treatment, albeit the timing of upregulation varied
   between STN-49 and aGBM5 cell lines ([368]Figures S7B and S7C). This
   indicates that cytotoxic response rates to BRAFi+MEKi could be cell
   line specific.

Figure 7.

   [369]Figure 7
   [370]Open in a new tab

   BRAFi+MEKi-induced Galectin-3 secretion links glial differentiation to
   PD-L1 upregulation

   (A) Timelines of in vitro experiments using STN-49 and aGBM5 cell
   lines.

   (B) Relative Galectin-3 concentrations measured over time with nELISA
   in supernatants of patient-derived cell lines treated with
   control/BRAFi+MEKi (two-way ANOVA with multiple comparisons). Simple
   linear regression ± SEM.

   (C and D) Scatterplots representing Galectin-3 expression levels in
   STN-49 (C) and aGBM5 (D) cell lines treated with DMSO or BRAFi+MEKi at
   24 and 48 h

   (E and F) Representative ICC images of STN-49 (E) and aGBM5 (F) cell
   lines treated with BSA or recombinant Galectin-3 protein for 72 h and
   subsequently stained for PD-L1, MBP, or OPALIN, and DAPI. Scale bars:
   40 μm. Bar graphs showing cellular quantification (% of total DAPI+
   cells). ∗p < 0.05 (two-way ANOVA with Sidak’s multiple comparisons
   test). ns, not significant. n = 2–3 replicates from one independent
   experiment, mean ± SEM.

   See also [371]Figure S7.

   Next, we investigated a potential direct role for Galectin-3 in
   inducing glial differentiation cell state transitions and PD-L1
   expression in STN-49 and aGBM5 cell lines ([372]Figure 7A). Adding
   recombinant Galectin-3 to cells in vitro efficiently enhanced both
   glial differentiation cell states and PD-L1 expression in STN-49 cells
   ([373]Figure 7E), whereas aGBM5 exhibited only an upward trend in PD-L1
   expression in differentiated glial populations ([374]Figure 7F). These
   findings suggest that Galectin-3 secretion from glioma cells
   potentially provides a key link between tumor differentiation and PD-L1
   upregulation after BRAFi+MEKi treatment.

Discussion

   BRAF and MEK inhibitors have revolutionized treatment for
   BRAF^V600E-mutant cancers, including gliomas, where targeted therapies
   are otherwise limited. Nevertheless, treatment challenges persist,
   particularly in glioma patients who experience tumor rebound after
   treatment cessation and in patients who develop therapy
   resistance.[375]^84 Resistance mechanisms can be categorized as primary
   (intrinsic), adaptive, and acquired. While intrinsic and acquired
   resistance mechanisms to BRAFi and BRAFi+MEKi are well described in
   several solid tumors, like melanoma, the specific mechanisms underlying
   resistance in glioma remain poorly understood.[376]^33^,[377]^85 Our
   study addresses this gap, focusing on glioma lineage-specific
   resistance. Compared to melanoma, the higher intrinsic resistance to
   BRAFi+MEKi of BRAF^V600E-mutant HGG underscores the importance of
   cancer lineage-specific studies. The high rates for intrinsic
   resistance in HGG may partially stem from the high intra-tumoral
   heterogeneity compared to LGG.[378]^44^,[379]^67 Our data confirm that
   BRAF^V600E-mutant HGG express gene programs of both neuronal and glial
   lineages, suggesting diverse differentiation states within tumors.

   We found that BRAFi+MEKi treatment induces significant shifts in glioma
   cell state, favoring more differentiated phenotypes at the expense of
   stem- and progenitor-like states. Our in vitro data support this
   observation, showing that surviving cells adapt to treatment by
   acquiring a more differentiated state,[380]^86 which is associated with
   upregulated immunomodulatory PD-L1 expression. This phenotypic
   heterogeneity and cell plasticity may underlie the high intrinsic
   resistance and rapid adaptation to treatment observed in
   BRAF^V600E-mutant HGGs. Indeed, similar mechanisms are emerging as key
   features glioblastomas, where tumor cells transition among four major
   cellular states.[381]^63

   Lineage-specific responses to BRAFi+MEKi were also observed.
   CD133/PROM1, a marker for both CSCs and differentiating
   OLs,[382]^62^,[383]^87 remains stable or increases with BRAFi+MEKi;
   whereas NESTIN+ cells decrease after short-term BRAFi+MEKi but persist
   with prolonged exposure, suggesting distinct CSC populations with
   unclear roles in tumor recurrence. Notably, differentiated glioma cells
   have been shown to accelerate CSC-driven tumor growth,[384]^88 implying
   that BRAFi+MEKi-induced differentiation could paradoxically support
   tumor persistence by maintaining a CSC reservoir. Traditionally,
   therapy resistance in glioma is associated with cell de-differentiation
   and the acquisition of stem cell-like states. In contrast, we
   demonstrate that BRAFi+MEKi induces a shift toward a more
   differentiated, immunosuppressive state, which may be a unique
   adaptation and resistance mechanism in glioma.

   Analysis of paired pre-/post-treatment glioma samples identified 13
   putative resistance-associated alterations, including mutations in
   ERRF1, BAP1, ANKHD1, and MAP2K1.[385]^33 Despite a small patient cohort
   size (4 out of 15 with BRAFi+MEKi), only about half of the cases
   harbored de novo mutations, suggesting that non-genetic adaptive
   resistance mechanisms, such as tumor cell plasticity and cell state
   transitions described in our studies, may play an important role in
   resistance in BRAF^V600E-mutant glioma. Our work shows that BRAFi+MEKi
   induces cellular plasticity, which may serve as a therapeutic
   vulnerability. Further studies are needed to determine whether the
   adaptive mechanisms also extend to non-BRAF^V600E-mutant and
   BRAF-wild-type gliomas.

   Our data reveal that BRAFi+MEKi may induce adaptive resistance via
   increased Galectin-3 secretion from glioma cells. Galectin-3 promotes
   cell survival and proliferation[386]^89 and induces MAPK reactivation
   via K-Ras-Raf-ERK1/2 signaling,[387]^89 facilitating glioma cell
   differentiation while simultaneously upregulating PD-L1 as an adaptive
   response to BRAFi+MEKi.[388]^90^,[389]^91^,[390]^92 It is not well
   understood what drives cancer cell state transitions in glioma,
   although they are a well-recognized adaptive mechanism enabling immune
   evasion, therapy resistance, and metastasis in many cancer
   types,[391]^93 influencing both plasticity and the TME. We found that
   BRAFi+MEKi-induced Galectin-3 secretion promotes glioma cell
   differentiation into AC- and OL-like states, mirroring its role in
   normal OL differentiation with increased MBP
   expression.[392]^94^,[393]^95 This differentiation is associated with
   increased PD-L1 expression, fostering an immunosuppressive TME. Given
   the antigen-presenting capabilities of ACs and
   OLs,[394]^71^,[395]^72^,[396]^96^,[397]^97 these cell state transitions
   may actively shape immune responses in glioma.

   Gliomas are considered immune inactive due to their low mutational
   burden,[398]^98 but BRAF^V600E-mutant gliomas are more immune rich,
   with elevated CD8^+ T cell infiltration, MHC class I
   expression,[399]^78 and PD-L1 expression (shown here), suggesting
   intrinsic ICI sensitivity. However, clinical trials with PD-1 and/or
   CTLA-4 inhibitors show limited efficacy in glioblastoma. PD-1 and
   CTLA-4 blockade are effective as second-line treatment after BRAFi+MEKi
   in BRAF^V600E-mutant metastatic melanoma,[400]^51^,[401]^52 providing a
   strong rationale for exploring BRAFi+MEKi+ICI combinations in glioma.
   Our study demonstrates that this concurrent quadruple therapy
   significantly improves survival compared to BRAFi+MEKi or ICI alone
   across all preclinical models, by preserving antigen presentation,
   enhancing T cell responses, and conferring a significant survival
   advantage.

   While clinical trials of BRAFi+MEKi+ICI in BRAF^V600E gliomas are
   lacking, melanoma studies provide insights. Patients previously treated
   with BRAFi+MEKi showed similar outcomes when subsequently treated with
   PD-1 or PD-1+ipilimumab, though the combination treatment had higher
   toxicity.[402]^52 Subsequent BRAFi+MEKi treatment after anti-PD-1
   monotherapy doubled objective response rates compared to anti-CTLA-4
   treatment alone.[403]^99 This is consistent with our findings showing
   that sequential ICI followed by BRAFi+MEKi offers benefit, although
   less than concurrent quadruple treatment. Additionally, BRAF-wild-type
   melanoma patients responded better to anti-PD-1+MEKi, while BRAF-mutant
   patients had lower ORR due to acquired resistance to
   treatment.[404]^100

   Toxicity remains a concern, as BRAFi+MEKi causes fever, leukopenia, and
   hyponatremia, while ICIs induce immune-related adverse
   effects.[405]^27^,[406]^101 The principal reason for stopping
   sequential treatment (ICI 1^st, BRAFi+MEKi 2^nd) was disease
   progression rather than adverse effects, which is promising and could
   support quadruple treatment in glioma. Dose reductions for BRAFi+MEKi
   could be considered for patients on quadruple treatment when tumors
   respond.[407]^53 Quadruple treatment may increase toxicity, but
   pediatric patients - enriched for BRAF-mutant gliomas - may tolerate it
   better than adults, and dose reductions or sequential treatment
   combined with Galectin-3 inhibitors could help balance efficacy and
   safety. Clinical data in metastatic melanoma and glioma highlight the
   necessity for further preclinical studies to refine treatment timing
   and minimize toxicity in HGG.

   In our preclinical models, BRAFi+MEKi could synergize with ICI by
   increasing T cell activity and antigen presentation, thus enhancing
   tumor intrinsic sensitivity. We hypothesize that sequential treatment
   (BRAFi+MEKi 1^st, ICI 2^nd) is less effective because it allows time
   for tumor cells to suppress T cell activity to reduce ICI efficacy,
   potentially through Galectin-3 secretion[408]^102 ([409]Figure S7D).
   Future studies will address whether Galectin-3 inhibition could further
   improve therapeutic outcomes by simultaneously targeting adaptive
   resistance and immunosuppressive TME.

   In summary, our study underscores the necessity of implementing
   concurrent quadruple therapy to generate more durable responses,
   thereby mitigating resistance mechanisms in BRAF^V600E-mutant HGG and
   GBM. Incorporating Galectin-3 inhibitors into existing treatment
   regimens for these gliomas also offers a promising strategy to enhance
   therapeutic efficacy while managing toxicity, thereby improving the
   overall quality of life for patients.

Limitations of the study

   Due to the rarity of matched pre- and post-treatment BRAF^V600E-mutant
   glioma patient samples, we generated two patient-derived cell lines and
   two immunocompetent mouse models/murine cell lines that effectively
   recapitulate key disease features.[410]^103 Despite expected
   inter-individual variability in treatment response within
   models,[411]^104^,[412]^105^,[413]^106 concurrent quadruple treatment
   consistently improved survival across all models, except for limited
   response in the RCAS-BRAF model, potentially caused by a higher PD-L1+
   T cell proportion. Future studies are required to elucidate resistance
   mechanisms in this model. Additionally, future studies should fully
   characterize BRAFi+MEKi-induced cytokine alterations and their effects
   on glioma cells, and on CD4/CD8 T cell activation. Lastly, variability
   in drug responses across different patient-derived cell lines may
   reflect intrinsic differences in tumor cell types or Galectin-3
   receptor expression, highlighting the need for single-cell analyses and
   expanded biological replicates or patient-derived cell lines.

Resource availability

Lead contact

   Requests for further information, resources, and reagents should be
   directed to and will be fulfilled by the lead contact: Claudia K.
   Petritsch, cpetri@stanford.edu and clpetritsch39@gmail.com.

Materials availability

   Murine BRAF-M34 and RCAS-BRAF cell lines, and patient-derived cell line
   (STN-49) generated for this study, will be made available upon request
   and require a materials transfer agreement according to Stanford
   guidelines.

Data and code availability

     * •
       Data supporting the findings of this study are available in the
       article and its supplementary files. Bulk RNA-seq data have been
       deposited in NCBI’s Gene Expression Omnibus (GEO) under accession
       number [414]GSE287972
       . scRNA-seq data have been deposited in GEO and are accessible
       through the accession number [415]GSE288791.
     * •
       All original code for analyzing bulk and single-cell RNA-seq data
       is indicated in the [416]key resources table.
     * •
       Any additional information required to reanalyze the data reported
       in this paper is available from the [417]lead contact upon request.

Acknowledgments