Introduction

Microglia are resident immune cells of the brain that play a key role in maintaining brain homeostasis and health. They perform diverse functions such as synaptic elimination, debris clearance, immune surveillance, and sculpting neuronal activity and plasticity^1–7. Microglial dysfunction is a key feature of neurodegenerative and myelin-related diseases such as Alzheimer’s disease (AD) and multiple sclerosis (MS)^4–6,8,9. For example, in AD, microglia can be maladaptive, exhibiting changes in gene expression and function such as impaired clearance of myelin debris and amyloid-f], chronic pro-inflammatory responses, and excessive elimination of neuronal synapses. Notably, the expression of many risk genes for AD are highly enriched in microglia^2,10,11; however, the signals and mechanisms that regulate the transition of microglia between homeostatic, adaptive, and pathological states remain poorly understood^1,4,8,9.

G-protein coupled receptors (GPCRs) mediate various cell-environment interactions and represent the largest set of currently druggable therapeutic targets. Several GPCRs are selectively expressed in microglia, but their functions are poorly understood, despite their potential relevance for therapeutic intervention^12,13. One such receptor is GPR34, which is a member of the rhodopsin-like family of GPCRs. GPR34 is highly and specifically expressed in microglia and macrophages, but its biological role remains largely undefined^13,14.

GPR34 expression in microglia is downregulated under various disease conditions such as AD and MS^14,15. Further, GPR34 is reported to play a role in maintaining microglial homeostasis, regulating lipid responses, and modulating microglia states and disease pathology^16–23. However, the identity of its endogenous ligand(s) remains uncertain; lysoPS has been proposed as a candidate ligand, but because it activates multiple receptors, its biological effects cannot be specifically attributed to GPR34^24–28. Studies report conflicting functions for GPR34: some protective, others detrimental against AD pathology^{16,17,19,21–23,29}. Several reports show that GPR34 inhibition alleviates neuroinflammation and promotes amyloid clearance in mice^17,21,22, whereas another reported that GPR34 agonism can enhance in vivo amyloid clearance¹⁶. These discrepancies highlight the limited understanding of the upstream signals and downstream pathways regulating GPR34 activity in microglia, as well as its context-dependent functions.

To address the in vivo function of GPR34, we first studied transcriptomic changes in Gpr34 KO mouse brains by RNA sequencing (RNA-seq). To investigate GPR34’s role in an Alzheimer’s disease model, we studied the brain and microglia of Gpr34 KO mice crossed to the 5xFAD mouse model of pathology^30,31. Human iPSC-derived induced microglia-like cells (iMGLs) null for GPR34 were generated and used to dissect downstream signaling of GPR34 in response to various ligands, and to evaluate the resulting effects on gene expression and phagocytosis. We find that Gpr34 deletion in mice shifts microglia toward inflammation-and DAM-related gene expression signatures, with some resemblance to 5xFAD microglia. In 5xFAD mice, Gpr34 KO exacerbates the disease-associated features, including heightened glial reactivity. In iMGLs, GPR34 deletion abolishes downstream signaling in response to both pharmacological agonists and myelin, and diminishes functional responses, including myelin-induced transcriptional effects and phagocytosis. Thus, by utilizing both in vivo and in vitro models, we were able to clarify how GPR34 influences microglial transcriptional states, signaling, and function in both normal physiologic and chronic disease settings.

Results

Gpr34 knockout in mice shifts microglial transcriptional profiles toward DAM- and inflammation-associated signatures

GPR34 is expressed highly and selectively in microglia within the brain in humans and mice (Supplemental Figure 1)^13,14. To determine how GPR34 loss affects microglial transcriptional programs, we performed single-cell RNA sequencing (scRNA-seq) of isolated microglia and brain border-associated macrophages (BAMs) from hippocampal tissue from 6-month-old wild-type (WT) and Gpr34 KO mice, on either a normal C57BL/6 background or in 5xFAD transgenic mice, resulting in 4 conditions: WT, Gpr34 KO (which refers to Gpr34 -/Y, as Gpr34 gene is X-linked), 5xFAD, and Gpr34 KO;5xFAD (Figure 1A-B; see Methods). We also performed single-nucleus RNA sequencing (snRNA-seq) of hippocampal tissue from the same mice (contralateral hemisphere) to study the effect of GPR34 loss on the non-microglial cells. Clustering, quality control, and cell type annotation were performed on both the single-cell and single-nucleus data (Figure 1B, Supplemental Figure 2 and 3). After standard quality control across samples of the microglia scRNA-seq data, we identified distinct subpopulations as described from previous studies^9,32, including a homeostatic state, DAM (which, more specifically, Amyloid-DAMs, as associated with amyloid-disease in 5xFAD mice)¹, interferon-responsive microglia, and a Ms4a7+ macrophage population representing brain-border associated macrophages¹, and cycling microglia (Supplemental Figure 3A-B). We then first focused on the comparison between WT and Gpr34 KO mice to define the transcriptional consequences of Gpr34 loss in the absence of 5xFAD pathology.

Figure 1:

In Gpr34 KO mice, we verified data quality across samples (Supplemental Figure 2B and Supplemental Table 1) and confirmed the absence of Gpr34 transcripts in Gpr34 KO microglia (Supplemental Figure 3C). Differential expression analysis of scRNA-seq from all isolated microglia comparing Gpr34 KO to WT revealed 60 differentially expressed genes (DEGs) with an FDR cutoff of 0.05 (Figure 1C and Supplemental Table 2), with many of the DEGs persisting even when performing differential expression analysis between Gpr34 KO and WT mice within a given subtype of microglia (Supplemental Table 2). Among the DEGs from the analysis with cells from all microglial subtypes (Figure 1C and 1D), we observed upregulation of known DAM and inflammatory genes, such as Anxa5, Axl, Itgax, Cst7, Cadm1, and Rftn1. Gene set enrichment analyses (GSEA) of Gpr34-deficient microglia revealed transcriptional changes across immune, Alzheimer’s disease-related, myelination, cell death, and cholesterol metabolism related modules (Figure 1E-F). DAM-related gene sets (more specifically Amyloid-DAM gene set) based on previously published data^1,9 (see Methods for details) were significantly enriched among upregulated genes of Gpr34 KO microglia, including genes such as Spp1 and Cst7, while “homeostatic” microglia gene sets were enriched among the downregulated genes (Figure 1E). Consistent with these changes, Gpr34 KO microglia showed increased enrichment of immune and inflammatory gene sets among upregulated genes, particularly cytokine and chemokine signaling, complement activation, and responses to lipopolysaccharide and interferon gamma, along with modest enrichment of Alzheimer’s disease-related genes (Figure 1E). Functional modules associated with microglial activity such as response to lipids, myelination, cell killing, and GPCR pathways were also enriched among upregulated genes (Figure 1F), suggesting altered lipid handling and signaling capacity.

Quantitative analysis of subcluster proportions in the microglia single cell data (Figure 1G) revealed a slight but significant reduction of homeostatic microglia and a corresponding expansion of DAM microglia in Gpr34 KO mice, with other subclasses not significantly changed.

Additionally, we used snRNA-seq to investigate the impacts of Gpr34 loss on other cell types in the hippocampus (Supplemental Figure 4A). Cell type composition analysis indicated a significant increase in the proportion of astrocytes that are disease-associated astrocytes (expressing previously defined markers genes such as Osmr1, Ggta1, and Gsn) and an increase (not reaching statistical significance) in the proportions of disease-associated microglia (DAMs; p-value=.083) and disease-associated oligodendrocytes (with previously reported markers such as C4b and Serpina3n; p-value=.29) (Figure 1H)^33–35. Notably, these changes occur despite the absence of Gpr34 expression in non-microglial cells, suggesting that the cell state changes in astrocytes and oligodendrocytes (ODCs) are secondary to the loss of Gpr34 in microglia. No significant changes in neuronal abundance or other major cell classes were observed, though we were able to identify differentially expressed (DE) genes in numerous cell types, most notably C4b in ODCs (Supplemental Figure 4B-G and Supplementary Table 3). Note, unlike in the single-cell data, we do not see any DEGs in the microglia in the single-nucleus data. This is likely due to the rarity of microglia in the single nucleus data, leading to low power to detect DEGs^7,36.

Together, the transcriptomics data demonstrate that the deletion of Gpr34 results in a small but significant shift in microglia from a homeostatic state towards a more DAM-like and inflammatory state, correlated with an expansion of DAM proportions and accompanied by (presumably secondary) transcriptomic changes in astrocytes and ODCs. These findings suggest that GPR34 functions to maintain microglia in a homeostatic state in vivo and that loss of Gpr34 may prime microglia toward activation, inflammation, and disease!Zlike states.

Gpr34 KO microglia partially resemble microglia of the 5xFAD Alzheimer’s model in transcriptomic profile

To further evaluate the relationship between GPR34 loss and AD-related microglial transcriptomic profiles, we directly compared DE results of Gpr34 KO vs WT microglia to those for the 5XFAD mouse model of AD vs WT at 6 months. 5xFAD is a well-established mouse model for amyloidosis that exhibits early and aggressive β!Zamyloid accumulation and robust microgliosis³¹. The original characterization of DAM and homeostatic transcriptional programs of microglia came from a study of the 5xFAD mouse line⁹. Changes in gene expression in Gpr34 KO microglia were positively correlated with those observed in 5xFAD microglia. 26 of the 60 genes that were significantly altered in Gpr34 KO vs WT microglia (Figure 1C) were also significantly altered in 5xFAD vs WT microglia, including many upregulated genes (Cst7, Itgax, Anxa5) and downregulated genes (including Gimap1, Gimap6, and Gimap9) all changing in the same direction (20 upregulated genes, 6 downregulated genes) (Figure 2A). Directly comparing the long fold changes (logFCs) from the DE analysis (Figure 2B) demonstrated positive correlation across the datasets (R = 0.53 among genes that reach significance in either comparison), particularly for DAM!Zassociated genes such as Apoe, Spp1, Cst7, and Axl. Although well correlated between Gpr34 KO and 5xFAD, the effect size (log₂FC) of the transcriptomic changes was generally larger for the DEGs in 5xFAD than in Gpr34 KO microglia (Figure 2B). Since both DE gene analyses used the same set of WT samples as controls, overestimation of correlation values can occur. To address this, we applied a permutation-based approach which found the correlation to be significant, confirming that the observed correlation was not driven artifactually by shared controls. As expected, analysis of scRNA!Zseq from purified microglia showed a reduction in Gpr34 expression in 5xFAD mice (Figure 2C), reaching significance based on the raw p-values reported by EdgeR³⁷ but not significant after applying fdrtools³⁸ as described in the methods.

Figure 2:

Gpr34 deletion potentiates DAM and inflammatory programs in the 5xFAD Alzheimer’s model

Given that Gpr34 KO by itself induces DAM-like and inflammation-related transcriptomic changes that partially overlap with that of 5xFAD microglia, we hypothesized that Gpr34 KO would further exacerbate the microglial DAM phenotype in the 5xFAD mouse model of amyloidosis. To address this, we crossed Gpr34 KO mice with 5xFAD mice and performed scRNAIZseq of isolated microglia and snRNAIZseq of hippocampal brain tissue from 6-month-old animals to assess gene expression and cell-type-specific changes across genotypes (Figure 1A).

Comparing Gpr34 KO;5xFAD to Gpr34 WT; 5xFAD by scRNA-seq revealed amplification of DAM and inflammatory-associated transcriptional responses in the Gpr34-deficient 5xFAD microglia. Differential expression analysis (Figure 2D) identified 27 DEGs that were upregulated in Gpr34 KO;5xFAD microglia relative to 5xFAD microglia, including Anxa5 and Gpnmb, reflecting an exacerbation of the DAM gene signature. There were only 8 significantly downregulated genes comparing Gpr34 KO;5xFAD to Gpr34 WT; 5xFAD microglia. Gene set enrichment analysis (Figure 2E) showed enrichment among upregulated genes of immune and microglial functional pathways, apoptosis-related terms, and gene sets associated with DAM and neurodegeneration, along with further downregulation of homeostatic gene sets. Cell-type composition analysis showed an increased proportion of DAMs in Gpr34 KO;5xFAD microglia relative to 5xFAD, though this did not reach significance (Supplemental Figure 5A).

snRNAseq of hippocampal tissues (Figure 2F) confirmed that Gpr34 deletion in the 5xFAD model significantly increased the proportion of disease-associated astrocytes and disease-associated oligodendrocytes with a non-significant increase in DAMs^33–35. As with the Gpr34 KO vs WT analysis, we were able to identify many DEGs in various cell types other than microglia (Supplemental Figure 5B-F and Supplemental Table 3).

Taken together, Gpr34 loss alone induces DAM and inflammation signatures in microglia (Figure 1) and its deletion in 5xFAD mice further amplifies, rather than reverses, the DAM and inflammatory transcriptomic profiles of microglia in 6-month-old 5xFAD mice (Figure 2). These findings suggest that GPR34 plays a modulatory role, restraining disease-associated and inflammatory pathways in response to βIZamyloid pathology.

Gpr34 loss exacerbates microglial reactivity and increases amyloid beta engulfment in 5xFAD mice

We examined the effects of Gpr34 knockout on other aspects of AD pathology (amyloid-beta (Aβ) and glial activation) by immunohistochemistry (IHC) of cortical sections of 6-month-old 5xFAD mice with or without Gpr34 (Gpr34 KO;5xFAD). Using confocal z-stack imaging and Imaris 3D surface analysis, we observed a non-significant increase in total Aβ-positive burden in the cortical ROIs in Gpr34 KO;5xFAD mice compared to 5xFAD (Figure 3A and 3B). The total Iba1+ immunoreactivity within cortical ROIs was unaltered between these genotypes (Figure 3C). However, we found that the CD68 signal (a marker of lysosome/phagocytic activity) within Iba1+ microglial compartments, normalized to total Iba1 signal, was significantly increased in Gpr34 KO;5xFAD mice (Figure 3D-E). Notably, Aβ signal within CD68+ structures (phagosomes/lysosomes) inside Iba1+ microglia was significantly increased in Gpr34 KO;5xFAD mice (Figure 3F). We also measured GFAP (a marker of astrocyte activation) in the cortex whole-slide immunostaining and 2D image analysis. We found a 45.7% increase in GFAP+ area in Gpr34 KO;5xFAD mice relative to 5xFAD, although that did not reach significance (p =0.08) (Figure 3G and 3H). Together, these immunohistochemical findings align with our transcriptomic results and suggest that Gpr34 loss in 5xFAD mice enhances microglial reactivity, amyloid beta engulfment, with a trend towards increased astrocyte reactivity.

Figure 3:

Generation and pharmacological validation of GPR34-deficient human iPSC-derived microglia (iMGLs)

We next sought to establish a robust platform for mechanistic and pharmacological studies at the cellular level. However, cultured primary murine microglia were inappropriate for these purposes, as Gpr34 expression is swiftly downregulated upon removal from the CNS^12,39(data not shown). We therefore utilized human iPSC-derived microglia-like cells (iMGLs), following a previously established 40-day cytokine-based protocol^32,40. We used a CRISPR/Cas9 strategy to introduce precise 2-base-pair deletions into the GPR34 coding sequence of WTC-11 iPSCs and confirmed an indel mutation in two different GPR34 KO clones (designated KO-1 and KO-2) by Sanger sequencing and RNA-sequencing (Figure 4A; Supplemental Figure 6A). For each KO clone, a matched isogenic control clone (WT-1 and WT-2) that underwent the identical CRISPR editing and clonal expansion workflow but retained the wild-type GPR34 sequence was used as a paired control. After differentiation into iMGLs, both KO clones responded similarly to a known GPR34 agonist, Compound 4B (Supplemental Figure 9A)⁴¹, so we focused on one pair (KO-1 and WT-1) for subsequent experiments. GPR34 KO-1 iMGLs displayed ∼2-fold elevated GPR34 transcript level relative to WT-1 iMGLs (Supplemental Figure 6B). However, quantitative mass spectrometry showed that GPR34 KO-1 iMGLs indeed lacked detectable GPR34 protein (Figure 4B).

Figure 4:

To evaluate any gross morphological effects of GPR34 loss, we performed immunostaining on iMGLs that were dissociated on day 42 and replated at equal cell density. Nuclear staining confirmed comparable cell numbers across all wells and genotypes (Supplemental Figures 7A-B). Cell area was measured by immunostaining using a panel of cell-surface and intracellular microglial markers consisting of: P2RY12, IBA1, C1q, C3aR, and CD45. All markers exhibited robust staining, validating that these were fully differentiated iMGLs^42,43 (Supplemental Figure 7A). GPR34 KO-1 iMGLs showed similar immunostaining intensities for all tested markers and no obvious differences in cell morphology were observed between KO and WT iMGLs (Supplemental Figures 7C-D). Additionally, flow cytometry conducted on Day 40, upon completion of the microglial differentiation protocol, showed comparable expression of CD45, CD11b, P2RY12, and CX3CR1 microglial markers in WT-1 and GPR34 KO-1 cells. (Supplemental Figures 8A-B).

To determine the functional consequences of GPR34 loss on downstream pathways and ligand responses, we used a set of pharmacologically validated tool compounds in our iMGL platform and examined downstream GPCR signaling pathways associated with GPR34⁴⁴. (Figure 4C). GPR34 activation by a validated agonist, Compound 4B, elevated intracellular calcium levels with an EC₅₀ of approximately 3 nM (Figure 4D and Supplemental Figure 9A) and increased ERK1/2 phosphorylation with an EC₅₀ of approximately 0.5 nM (Figure 4E). Both responses were abolished in GPR34 knockout iMGLs (Figure 4D-E) and by the GPR34 antagonist YL-365⁴⁵. (Supplement Figure 9B-C). GPR34 is thought to be a G αi-coupled receptor ⁴⁴. However, upon treatment with Compound 4B, the level of forskolin-induced cAMP was unchanged, suggesting that GPR34 does not significantly modulate cAMP signaling under these conditions (Supplemental Figure 9D). These data validate our iMGL platform for assessing GPCR signaling, demonstrating robust and receptor-dependent activation of canonical downstream signaling pathways.

Lysophosphatidylserine (LysoPS), which has been proposed as an endogenous GPR34 ligand, stimulated calcium and pERK responses in wildtype iMGLs¹⁷. Responses to LysoPS were reduced in the GPR34 knockout iMGLs, albeit incompletely and with a rightward shift in the dose response curve (Figure 4F-G), suggesting the presence of additional LysoPS-responsive receptors in iMGLs besides GPR34. Myelin debris, which is lysoPS-rich, has recently been reported to promote microglial dysfunction, neuroinflammation, and AD-related pathology via GPR34^19,22. Consistent with this, we found that myelin also induced calcium elevations in microglia, which was blocked in GPR34 KO-1 iMGLs (Figure 4H-I), or by a GPR34 antagonist, (S)-E49 (Figure 4H) ((S)-E49 - Itoh et. al. Patent application publication US2010/0130737). By contrast, calcium signals elicited by agonists of other microglia GPCRs (P2Y12 and C3aR), remained largely intact in GPR34 KO-1 iMGLs (Figure 4J-K).

Together, these findings show that GPR34 primarily regulates calcium and pERK signaling but not cAMP in human iPSC-derived microglia and that LysoPS, the proposed endogenous ligand, acts, at least in part, in a GPR34-dependent manner. Importantly, myelin debris also stimulates calcium signaling in microglia in a GPR34-dependent manner.

GPR34 deficiency selectively impairs myelin phagocytosis in iPSC-derived microglia

Phagocytosis is a core microglial function essential for homeostatic maintenance of the central nervous system (CNS)^2,5,6. Our data indicate that myelin, a key in vivo phagocytic substrate, can signal through GPR34 (See Figure 4H-I), leading us to examine whether GPR34’s interaction with myelin also regulates its engulfment. To directly investigate the role of GPR34 in this fundamental process, we performed in vitro phagocytosis assays using iMGLs from WT-1 and GPR34 KO-1 iMGLs, along with the GPR34 antagonist YL-365⁴⁵. These experiments build on prior findings suggesting that GPR34 contributes to microglial phagocytosis in vivo²⁰ and substrate selectivity in vitro¹⁶.

We assessed the phagocytic uptake of three substrates: myelin debris, amyloid fibrils, and E. coli (Figure 5A, schematic). Each substrate was conjugated to pHrodo Deep Red, a pH-sensitive fluorophore that becomes fluorescent in acidic compartments such as the phagolysosome. After exposure of WT-1 and GPR34 KO-1 iMGLs to each pHrodo-labeled substrate, we measured median fluorescence intensity (MFI) using two complementary approaches: live confocal imaging 5 and 24 hours after substrate addition (Figure 5A), and flow cytometry at the 25-hour terminal time point (Supplemental Figure 10). Both methods showed a significant 30% reduction of myelin uptake in GPR34 KO-1 microglia compared to WT-1 cells (95% CI: 12-49%) (Figures 5B and Supplemental Figures 10A-B). In contrast, phagocytosis of amyloid fibrils and E. coli bioparticles was not significantly affected (Figure 5C-D and Supplemental Figures 10C-D). To test whether pharmacological inhibition of GPR34 could replicate this selective impairment, iMGLs were pre-treated with the GPR34 antagonist YL-365 (10 µM) for 30 minutes prior to substrate exposure. Antagonist-treated cells showed a similar reduction in myelin phagocytosis, with no effect on uptake of the other substrates (Figures 5E and Supplemental Figures 10E-H). Together, these results indicate that, at least in vitro, GPR34 plays a selective and critical role in microglial phagocytosis of myelin debris.

Figure 5:

Multiomic analysis of GPR34-deficient iMGLs highlights dysregulation of immune, replicative, and lysosomal programs

To investigate the molecular consequences of GPR34 loss in human iMGLs, we performed bulk RNA sequencing and quantitative mass spectrometry (MS) proteomics on WT-1 versus GPR34 KO-1 iMGLs. Principal component analysis (PCA) of RNA-seq data demonstrated clear separation of WT-1 and KO-1 iMGLs on PC1 across multiple rounds of differentiation (Supplemental Figure 11A).

Differential expression analysis revealed broad transcriptional differences between GPR34 KO-1 and WT-1 iMGLs (Figure 6A, Supplement Table 4). Gene set enrichment analysis (GSEA) highlighted enrichment of DAM-related genes, and of pathways related to phagocytosis, endocytosis, and lysosomal function among the upregulated genes, consistent with increased expression of genes such as PLIN2, CD36, LAMP2 and SYT11^46–49. Conversely, pathways associated with DNA replication, cell cycle, and proliferation were downregulated with genes such as MCM4, and TOP2A, and CBX5 contributing to these changes^50–52 (Figure 6B-C).

Figure 6:

Consistent with the transcriptomic analysis, PCA analysis of the proteomics dataset demonstrated distinct clustering of WT-1 and KO-1 samples (Supplemental Figure 11B). GPR34 deletion produced a pronounced effect on the iMGL proteome, with thousands of differentially expressed proteins (DEPs) surpassing the significance threshold (FDR-adjusted P < 0.05) (Figure 6D). GSEA indicated enrichment of proteins related to T-cell activation and MHC Class II complex formation (FDR adjusted P < 0.05) among the upregulated DEPs, whereas pathways related to the MCM complex, DNA replication initiation, multivesicular body organization, endocytic regulation, and ESCRT I complex were downregulated (Figure 6E).

Within the vesicular trafficking pathways, VPS37B, a core ESCRT component required for endosomal sorting⁵³, was strongly downregulated at both the transcript and protein levels (log₂FC = -2.77, adjusted p = 6.8 x10^-15 for RNA; log₂FC = -1.23, adjusted p = 8.9 x10^-6 for protein) (Figure 6A, Figure 6D). The broader proteomic landscape also emphasized endosomal and lysosomal dysfunction, evidenced by reduction of ESCRT complex components (MVB12A, VPS37B, VPS37C, VPS28), lysosomal enzymes (CTSL, CTSB, CTSC, GNS), lysosomal membrane proteins (OSTM1), and regulators of endosomal trafficking (BLOC1S1, FLOT1, VPS13C, DYNLRB1, CLTCL1) (Figure 6F). Notably, these reductions were accompanied by a striking increase in GPNMB at the protein level, a putative biomarker of lysosomal impairment, and a canonical DAM marker^54,55 (Figure 6F). Together, these data highlight vesicular trafficking and lysosomal biology as microglial processes likely disturbed by GPR34 loss of function.

Consistent with the transcriptional downregulation of proliferative pathways in RNAseq data, we observed reduced abundance of DNA replication proteins, including multiple MCM family members (MCM2-7) in GPR34 KO-1 iMGLs (Figure 6G). These molecular phenotypes suggest reduced proliferation of GPR34-deficient cells, which is supported by our findings that KO-1 cultures yielded fewer cells than WT-1 during the iMGL differentiation protocol (Supplemental Figure 11C). In parallel, antigen presentation pathways were strongly upregulated, driven by increased protein levels of MHC class II molecules such as HLA-DPB1, HLA-DPA1, HLA-DQB1, and HLA-DRB4, highlighting a shift from proliferative to immune-related programs^56–58.

Finally, integration of transcriptomic and proteomic datasets demonstrated significant albeit modest correlation between respective log₂FCs of RNAs and proteins (r = 0.22, P=2.96e-17) (Figure 6H). In general, the magnitude of change was much greater on the proteome than on the transcriptome, implying that many of the changes driven by loss of GPR34 occur through post-transcriptional mechanisms, which suggest that proteomics can detect biologically relevant changes that are not present at the mRNA level. Nonetheless, several prominently altered DEGs exhibited directionally consistent and pronounced changes in RNA and protein levels, including downregulation of VPS37B, a core ESCRT-I component required for lysosomal degradation, and COASY, which encodes coenzyme A synthase⁵⁹.

Together, our omics data in human iMGLs indicate that loss of GPR34 impacts broadly microglial RNA and protein expression, with upregulation of pathways related to immune activation, and suppression of pathways related to DNA replication, phagocytosis, lysosomes and vesicle trafficking.

GPR34 loss blunts microglial transcriptional responses to myelin challenge

Myelin is a potent stimulus for microglia, inducing DAM and inflammatory gene expression signatures³². Since GPR34-deficient microglia show impaired calcium signaling in response to myelin (Figure 4H-I) and selective deficits in myelin phagocytosis (Figure 5B-E), we next investigated whether GPR34 alters microglial transcriptional responses to myelin. To capture both early and sustained responses, WT-1 and GPR34 KO-1 iMGLs were exposed to myelin in culture for 2 hours or 24 hours, followed by bulk RNA sequencing.

Principal component analysis (PCA) revealed that the transcriptome of 24-hour myelin-treated samples clustered distinctly from the 2-hour and untreated (no-myelin) groups, reflecting transcriptional changes induced by prolonged myelin exposure (Figure 7A). The 2-hour and no-myelin samples clustered more closely, consistent with modest early transcriptional responses. In all treatment conditions, WT-1 and KO-1 samples clustered separately from each other, in line with our previous analyses of GPR34 KO-1 microglia (Figure 7A). Differential expression analysis and transcriptome-wide impact (TWI) scores⁶⁰ from KO-1 vs WT-1 iMGL samples confirmed that myelin induced a substantial transcriptomic shift, with KO-1 samples showing attenuated responses at both time points (231 DEGs in KO-1 vs 921 DEGs in WT-1 at 2 hours, 5648 DE genes in KO-1 vs 8370 genes in WT-1 at 24 hours, (Figure 7B and Supplemental Figure 12A). Despite fewer DEGs in KO-1 samples treated by myelin, many of the DEGs at 24 hours were shared with WT-1 iMGLs treated with myelin (4,729 shared DEGs, r = 0.89, based on genes DE at an FDR cutoff of 0.05 in at least one of the two comparisons) (Figure 7C; Supplemental Figure 12B). Overall, the loss of GPR34 attenuated the broader transcriptomic response induced by prolonged myelin exposure but did not abolish or profoundly alter the nature of the response, implying that additional myelin-responsive pathways act alongside GPR34. This contrasts with acute calcium signaling response to myelin, which appears to be largely mediated by GPR34 (Figure 4H-I).

Figure 7:

At 2 hours, myelin induced a significant transcriptional response in WT-1 iMGLs with roughly 1,500 DEGs, characterized by rapid upregulation of immediate-early (IEGs) and cytokine-responsive genes such as TNF, DUSP1, and FOS consistent with acute activation of microglia (Supplemental Figure 12C). Caution must always be used with IEGs since such genes are known to also be related to certain activation artifacts, as has been reported previously⁶¹, By 24 hours, the transcriptional response to myelin expanded to roughly 8,000 DEGs and shifted toward lipid-related gene sets. Specifically, myelin exposure triggered downregulation of cholesterol biosynthesis genes (HMGCS1, MVK, FDFT1, DHCR24), upregulation of LDLR negative regulator (MYLIP), downregulation of cholesterol uptake gene (LDLR), and upregulation of cholesterol export genes (ABCG1 and PLTP), collectively suggesting changes in handling and export of cholesterol (itself a component of myelin)^62–65 (Supplemental Figure 12C).

To address whether GPR34 deficiency alters the microglial response to myelin, we compared the transcriptional changes between WT-1 and KO-1 iMGLs. At 24 hours following myelin treatment, VPS37B (a core ESCRT-I component required for lysosomal degradation) and COASY (which catalyzes the final steps of coenzyme A biosynthesis) were both reduced in KO-1 iMGLs relative to WT-1 samples (Figure 7D). Relative to untreated controls, VPS37B was upregulated by myelin in both genotypes. In contrast, COASY was downregulated in WT-1s after myelin treatment but was not significantly different in KO-1 cells. These changes suggest impaired vesicle trafficking and altered metabolic-related programs in KO-1 iMGLs. Upregulation of DAM-related and lysosomal-related genes such as APOE, CCL3, CD68, CST7, GPNMB, and TREM2 was also blunted in KO-1 iMGLs after 24-hour myelin exposure (Figure 7E, F, Supplement Table 4)⁹.

At the pathway level, myelin exposure in WT-1 iMGLs at 2 hours led to enrichment among upregulated genes of pathways related to chemokine and cytokine activity, neuroinflammatory response, and proliferation (Figure 7G). In the KO-1 versus WT-1 comparison, many of these immune-related and proliferation pathways were instead downregulated, reflecting a blunted early transcriptional response of KO-1 iMGLs to myelin. By 24 hours, WT-1 samples showed enrichment of upregulated genes in DAM signatures, cholesterol metabolism, and maintenance of immune and inflammatory responses, alongside a concurrent decrease in proliferation. In the KO-1 versus WT-1 comparison, these responses were altered: DAM- and immune-related programs, as well as phagocytosis pathways, were downregulated, while proliferation pathways were upregulated in the KO-1, suggesting an attenuated or dampened transcriptional response relative to WT-1 controls (Figure 7G).

We observed that myelin treatment (2 hours and 24 hours) elicits similar gene-set-level transcriptional responses in WT-1 and KO-1 iMGLs (Supplemental Figure 12D; see also Figure 7B, C, G and Supplemental Figure 12B). Altogether, our data suggest that GPR34 is not required for microglial transcriptomic response to myelin stimulus but may modulate the magnitude of the response. We hypothesize that there are other receptors on microglia, besides GPR34, that can recognize myelin or the components of myelin.

Discussion

Although GPR34 has long been recognized as a microglia-enriched GPCR whose expression is mis-regulated in various brain disease conditions, its normal physiological functions in microglia and its relevance to human neurodegenerative diseases have remained unclear^{14,18,20,21,26,66,67}. In the mouse brain, loss of Gpr34 shifted microglia toward DAM and inflammatory-related transcriptional states (Figure 1). The transcriptional modules upregulated in Gpr34-deficient microglia overlapped with those upregulated in the 5xFAD model of AD. Gpr34 KO in 5xFAD mice further potentiated DAM-and inflammation-related features of microglia and exacerbated glial activation with no change in amyloid burden in this amyloidosis mouse model (Figures 2-3). Together these data suggest that GPR34 plays a role in preventing activation of microglial pathways that are associated with disease and inflammation. Our findings build on previous studies showing that Gpr34 deficiency increases immune and cytokine gene expression^18,20 and accelerates transition to DAM microglia²¹. Overall, these data support a model in which GPR34 helps maintain microglial homeostasis, consistent with its high expression in homeostatic microglia and suppression in disease-associated microglia, including in our reanalysis of a published DAM dataset9. The generation of the DAM and homeostatic gene sets is described in the Methods.

In 5xFAD brains, we found reduced Gpr34 expression, consistent with reports that Gpr34 levels are reduced in human brain tissue from neurodegenerative patient contexts¹⁶, although another recent paper reported the opposite trend²². Our RNA-seq and histopathological results from Gpr34 KO;5xFAD mice appear to contrast with recent reports that Gpr34 genetic disruption or inhibition with drugs alleviates neuroinflammation and promotes amyloid clearance in the APP/PS1 and 5xFAD models^17,21,22 and with a separate report in which Gpr34 agonist treatment promotes in vivo clearance of amyloid plaque in the APP/NLGF model¹⁶. These apparent discrepancies may reflect differences in model severity, experimental approach, disease stage, or timing of intervention.

Notably, our study focuses on an earlier timepoint (6 months) in 5xFAD mice compared to 9 months of age in Zhou et al., 2025 when inflammatory pathways in microglia may be differently or more strongly engaged²². Lin et al., 2023 studied Gpr34 knockdown in the hippocampus of APP/PS1 mice, introducing another layer of variability¹⁷. In addition, our study utilized a constitutive Gpr34 KO, introducing developmental effects on microglia prior to the appearance of amyloidosis, in contrast to a pharmacological perturbation ²² or genetic intervention after the onset of disease pathology^17,19. Future systematic studies that compare the temporal and mechanistic dimensions of GPR34 function will be essential to clarify whether, when, and how its functions change during development or disease progression.

GPR34-deficient iMGLs displayed blunted Ca²⁺ and pERK signaling in response to LysoPS-derived agonists and myelin (Figure 4). Our findings align with prior reports showing that GPR34 mediates LysoPS- and myelin-derived signaling and regulates microglial lipid homeostasis^19,21,22. Interestingly, while LysoPS-derived agonists of GPR34 such as Compound 4B activated robust signaling in iMGLs, LysoPS itself induced responses that were only partially dependent on GPR34. This suggests that there are additional receptors that can sense Lyso-PS and related lipids^{14,26,28,45,68}. Interestingly, we found that GPR34-deficient iMGLs were only deficient in phagocytosis of myelin with no impairment in β-amyloid or E. coli uptake. This is in concordance with the idea that myelin uptake involves LysoPS/GPR34 interactions, since synthetic amyloid fibrils do not contain Lyso-PS and bacteria are very low in Lyso-PS. This contrasts with prior studies reporting variable outcomes^16,19,20,22. Lin et al., 2024 found no effect of Gpr34 loss on myelin uptake¹⁹. Etani et al., 2024 reported increased β-amyloid phagocytosis with GPR34 activation without affecting myelin¹⁶, and Preissler 2015 described reduced uptake of zymosan and FBS-coated particles in Gpr34-KO mouse microglia²⁰. These discrepancies likely reflect differences in model systems, assay and other experimental factors. Two of these studies evaluated myelin uptake in primary mouse microglia^16,19. However, primary mouse microglia rapidly downregulate GPCR expression upon removal from the CNS and during extended culture, unlike physiologically maintained iPSC-derived microglia, in which GPR34 expression remains relatively stable¹³ (our unpublished data). Further, acute pharmacological activation to interrogate GPR34-dependent uptake¹⁶, may not corroborate with the impacts of long-term genetic loss in a knockout model²⁰.

GPR34 signaling can modulate microglial state and appears to favor a homeostatic rather than inflammatory function of microglia. Therefore, we hypothesize that pharmacologic approaches that activate or inhibit GPR34 activity could be used to alter microglial state in neuroinflammatory and neurodegenerative disorders^{16,41,45,69–71}. Whether stimulation or inhibition of GPR34 would be beneficial is likely to depend on the disease pathology, the stage of the disease, and the specific functions of microglia that are being targeted.

Our study is limited to a single disease model and relies primarily on genetic loss-of-function approaches in vivo. Although we used lysoPS and Compound 4B to prove GPR34 signaling pharmacologically in vitro, future studies should carefully test the impact of pharmacological modulation of GPR34 signaling in vivo through both agonist and antagonist approaches. Future research should also address sex-dependent variability and expand to additional disease contexts and examine how GPR34 interacts with other microglial receptors and lipid-sensing mechanisms^{12,13,32,72–76}. These comparisons will help clarify whether GPR34 acts in parallel or synergistically with pathways mediating microglial homeostasis, lipid metabolism, and the inflammatory response.

Data availability

Code used for this manuscript is available on GitHub (see https://github.com/seanken/GPR34_DAM). Raw sequencing data have been uploaded to the Gene Expression Omnibus (GEO) under accession numbers GSE327711 for the iMGL RNAseq data and GSE327428 for the mouse RNAseq datasets. The processed data will be put in the Broad Single Cell Portal.

Acknowledgements

We thank Kris Dickson and Berengere Sauvagnat for their contributions to project strategy and project management, respectively. We are grateful to the members of the Sheng and Stevens labs for their valuable input, and to Jen Pan and Oivin Guicherit for their project discussions. We also thank Leiby Vargas, Matthew Beck, Nathan Chan, and Tyler Caron for assistance with mouse colony management; the Broad Institute Genomics Platform for support with RNA sequencing; the Broad Proteomics Platform for assistance with proteomics analyses; and the Flow Cytometry Core for access to instrumentation and technical support. We are especially grateful to Laurence Marie Daheron and Caroline E. Becker from the Harvard iPS Core Facility for generating the GPR34 KO iPSCs. We thank Jeffrey Howard for administrative support, Elvira Dzhura and Abigail Desrochers for laboratory management, and Laney DiGiovanni for budgeting support.

Additional information

Author Contributions

M.S. conceived the project and supervised all aspects of the work. P.K. initiated and oversaw the iMGL project and supervised in vivo data analysis involving Gpr34 KO and 5xFAD experiments. A.G. designed and conducted iMGL experiments: phagocytosis, RNA-seq, and myelin RNA-seq. M.K. conducted in vivo experiments and contributed to data analysis. Y.Z. co-initiated the iMGL platform, and Y.Z. and Q.X. performed pharmacological validation. W.E.M. generated the GPR34 knockout clones, co-initiated and co-validated the iMGL platform, performed early iMGL RNA-seq experiment and contributed to in vivo experiments. S.S. analyzed RNA-seq data and contributed to experimental design under the supervision of J.L. J.D., C.D., and N.M. advised and contributed to iMGL experiments under the mentorship of B.S. S.N. performed in vivo immunohistochemistry and analysis under the supervision of H.B. E.Le. initiated in vivo experiments. X.A. conducted the in vivo RNA-seq experiments under the supervision of J.L. T.N. and M.D. contributed to iMGL preparation. K.J.S. and D.B. maintained and characterized the Gpr34 knockout animals. J.Z. contributed to iMGL-associated RNA-seq experiments. A. V. and D.V. performed mass spectrometry proteomics analysis under the supervision of H.K. and S.A.C. E.Lo. and C.B. developed pharmacological compounds under the supervision of D.Mc., S.G., and M.W. D.Mi. contributed to RNA-seq data analysis. M.J. advised on all aspects of projects involving iMGL work. P.S.K., A.G., S.S., Y.L.Z., and M.K. wrote the manuscript with assistance from M.S., and input from all authors

Funding

This work was supported by the Stanley Center for Psychiatric Research, Broad Institute.

Funding

Stanely Family Foundation

• Morgan Sheng

Additional files

Supplemental Figure Legends

Methods

Supplemental Table 1

Supplemental Table 2

Supplemental Table 3

Supplemental Table 4

Supplemental Figures