The objective of this study was to determine whether folate receptor β (FRβ) performs an immune-related function besides folate transport in monocytes and macrophages. To test this hypothesis, we generated FRβ knockout (KO) C57BL/6 mice and compared their immunologic functions with those of wild type (WT) mice. We characterized the FRβ KO mice, analyzed immune-related gene and protein expression, evaluated responses to inflammatory stimuli, and examined tumor growth in both WT and KO mice.

Sample sizes were determined based on previous experience and pilot experiments to achieve sufficient statistical power. For animal studies, at least 3–7 mice per group were used, as indicated in the figure legends. Mice were randomly assigned to experimental groups. Investigators were not blinded during data collection and analysis. All animal studies were approved by the Purdue University Institutional Animal Care and Use Committee.

For human cancer patient survival analysis, publicly available data from The Cancer Genome Atlas (TCGA) were used. No statistical method was used to predetermine sample size for the human data analysis.

Experiments were performed at least three times independently unless otherwise noted. No data were excluded from the analyses.

TRAMP C2 cells (Murine prostate adenocarcinoma) were purchased from the ATCC and cultured in DMEM medium supplemented with 10% FBS, 4 mM L-glutamine, 5 μg/mL insulin, 10 nM dehydroisoandrosterone and 1% penicillin/streptomycin. MC38 cells (murine colon adenocarcinoma) were purchased from Kerafast and cultured in DMEM medium with 10% FBS, 2mM L-glutamine, 0.1 mM nonessential amino acids, 1 mM sodium pyruvate, 10 mM HEPES, 50ug/mL gentamycin sulfate, and 1% penicillin/streptomycin. For all experiments, cells were used within 10 passages from purchased stocks.

Six- to eight-week-old male or female wildtype C57BL/6 mice were purchased from Charles River. FRβ and FRδ knockout mouse breeders were generated by Richard Finnell at Baylor College of Medicine and bred at the Purdue University animal facility (9). Mice were housed in accordance with protocols approved by Purdue University Animal Care and Use Committee. Water, regular rodent diet (Envigo, #2018S/2018SC) or folate-deficient chow (Envigo, #TD.00434) were freely available. For all studies on tumor-bearing mice, mice were fed folate-deficient chow for two weeks before tumor implantation and maintained on the chow for the duration of the studies. Briefly, commercially available mouse chow contains megadoses of folic acid that raise the serum folate concentrations of the mice to ~700 nM. Physiological folate concentrations are however only 20 nM. Therefore, to reduce the folate levels in the mice to normal concentrations, we have maintained the mice on a low folate diet. These physiological folate concentrations are necessary to observe physiologic interactions of folate and folate-linked fluorescent dyes with endogenous folate receptors.

FRβ knockout mice were created by insertional mutagenesis in the Richard Finnell lab as described earlier (9). These mice were then backcrossed onto the C57BL/6J strain (Charles River) using inbred SWV background Folr2 ^(+/-) heterozygous male mice (from the Finnell lab) with C57BL/6 females in the Amaya Puig-Kröger lab to generate the final FRβ knockout (Folr2 ^-/-) mice. FRδ knockout mice were generated in collaboration with Purdue University’s Transgenic and Genome Editing Facility using the purchased embryos (B6.129P2(Cg)-Izumo1r^tm1.1Salb/Mmnc, Item# 037093-UNC-EMBRYO, MMRRC) (17). Tail tips (2-5mm) were collected from 2–3 weeks old FRβ KO pups in accordance with NIH Animal Research Advisory Committee (ARAC) Guidelines for Genotyping Mice & Rats. Samples were then processed as previously described (18). After PCR, samples were separated by electrophoresis on 2.5% agarose gels at 150 volts for 30 minutes, and DNA was imaged using a Blue View Transilluminator (Vernier Bio-Technology).

Whole spleens were harvested, fixed in formalin (10%), and H&E stained slides were prepared by following the standard protocol (19). For magnification control, hearts and livers were also stained with the spleens.

Male and female WT or FRβ KO or FRδ KO mice were inoculated subcutaneously with TRAMP-C2 cells (2×10⁶ cells/mouse in FRβ KO, 0.5×10⁶ cells/mouse in FRδ KO) or MC38 cells (4×10⁵ cells/mouse). Tumor sizes were measured 2–3 times per week and volumes were calculated as (L*W*W)/2, where L is the length, W is the width of the tumor. When tumors reached 1000-1500mm³, mice were sacrificed and tumors were dissociated with tumor dissociation kit (Miltenyi, Cat#130-096-730). Cells were filtered through a 70-μm cell strainer and treated with RBC lysis buffer (Biolegend, #420301) to deplete erythrocytes. After washing 2x with cold PBS, the resulting single cell suspensions were resuspended in FACS buffer, stained with Zombie Violet dye (BioLegend, Cat#423114), and incubated with anti-mouse TruStain FcX™ (BioLegend, Cat#101319) for 5–10 minutes on ice. Cells were then washed 2x with PBS and then stained with antibodies listed in Supplementary Table S1 . For antibody staining, cells were incubated with antibodies for 30min on ice. For small molecule staining, cells were incubated with FA-Cy5 (10nM) in the presence or absence of FA-glucosamine (1uM) for 1h. After washing twice with cold PBS, cells were analyzed by flow cytometry for the desired phenotypic markers: Monocyte-like (M)-MDSCs were identified as CD45+CD11b+Ly6C+, while granulocyte-like (G)-MDSCs were identified as CD45+CD11b+Ly6G+. Tumor associated macrophages (TAMs) were identified as CD45+CD11b+Ly6G-Ly6C-F4/80+. Additional analysis of MDSCs and TAMs included staining with anti-PD-L1 antibody and folate-Cy5 in the presence or absence of folate-glucosamine (competition). T cells were identified as CD45+CD3+. CD69 was employed as T cell activation marker, while PD1 was used as T cell suppression marker. All flow cytometry analyses were conducted on an Attune™ NxT Acoustic Focusing Cytometer (Invitrogen). Data were analyzed with Attune™ NxT Software.

Differentiation and polarization of murine bone marrow derived macrophages and bone marrow derived MDSCs.

Murine bone marrow cells were isolated from tibias and femurs of WT or FRβ KO male C57BL/6 mice as previously described (20). Cells were differentiated to BMDMs by culturing for 7 days in complete growth medium (regular or folate deficient RPMI 1640 medium supplemented with 10% FBS and 1% Penicillin/streptomycin) containing 20 ng/mL recombinant mouse M-CSF (Biolegend). The resulting macrophages were polarized into M1-like macrophages by treatment for 24 hours in complete growth medium containing recombinant mouse IFNγ (20 ng/mL; Biolegend) plus 1 ng/mL lipopolysaccharide (Biolegend), or M2-like macrophages by incubating for 48 hours with 20ng/mL IL-4 and 20ng/mL IL-13. IL-6 was analyzed in cell culture supernatants by ELISA and cells were harvested for RNA sequencing.

BM-MDSCs were obtained by culturing the isolated bone marrow cells for 7 days in complete growth medium containing 40ng/mL recombinant mouse GM-CSF (Biolegend) and 40ng/mL IL-6 (Biolegend). Mouse IL-6 and IL-10 in cell culture supernatants were analyzed using mouse IL-6 (Biolegend, # 431304) and IL-10 (Biolegend, # 431414) ELISA kits, respectively, according to the manuals.

Eight-week-old WT or FRβ KO male mice were instilled with 0.75mg/kg bleomycin, as previously described (20), and body weights were monitored throughout the study. Mice were imaged by micro-CT on days 14 and 21 post bleomycin instillation as described previously (21) to quantitate lung fibrosis. Bronchoalveolar lavage (BAL) fluids were harvested on day 21 and pelleted the cells by centrifugation for qPCR analysis of macrophage phenotypic markers.

For RNA isolation from mouse tissues, liver, kidney, and lung samples were freshly collected following euthanasia and homogenized in cold DNA/RNA Shield buffer (Zymo Research, #R1100-50). RNA was then extracted using the Direct-zol RNA Miniprep Kit (Zymo Research, #R2051), according to the manufacturer’s instructions.

For RNA isolation from mouse macrophages, the cells were lysed in cold DNA/RNA Shield buffer (Zymo Research, #R1100-50) followed by RNA extraction using the Quick-RNA™ MicroPrep kit (Zymo Research, #R1051) and reverse-transcribed to cDNA using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems™, #4374966) according to the manufacturer’s instructions as previously described (20). qPCR was performed using the iTaq™ Universal SYBR Green SuperMix (Bio-Rad Laboratories, #1725121), iCycler thermocycler, and iCycleriQ 3.0 software to examine the expression of folate transporters and M2-macrophages markers. Primer sequences for qPCR are shown in Supplementary Table S2 .

Total RNA was extracted from three mice per group, purified from mouse BMDM cells, pooled in each group, and sent to Novogene for QC and RNAseq analyses (Novoseq PE150). RNA samples were sequenced using an Illumina sequencer using a paired-end protocol, targeted read-length of 150 bp, and >40 million total reads per sample. Data quality control (adapter trimming and removal of reads when bases with quality scores <5 constituted >50% of the reads) was performed at Novogene. Reads were aligned to mouse (mm10) reference genome using HISAT2 aligner (22). The reads mapped to each gene were calculated at Novogene. The EdgeR Bioconductor package (23) was used to perform the differential expression analysis. Genes with cutoffs (pvalue < 0.05 and |log2FoldChange| > 1) were denoted as Differentially Expressed Genes (DEGs). Heatmaps and bar plots were created using the R-package pheatmap (24) and ggplot2 (25), respectively. GO enrichment analysis was performed by Novogene.

Macrophages were obtained by harvesting bone marrow from healthy WT and FRβ KO C57BL/6 female mice followed by differentiation in the presence of mouse M-CSF (20 ng/mL) for 7 days. Cells were then polarized to M2-like macrophages by culturing 48 hours in IL-4 (20 ng/mL) and IL-13 (20 ng/mL) containing media (folate deficient RPMI 1640 medium supplemented with 10% FBS and 1% Penicillin/Streptomycin). T cells were obtained separately starting with harvesting spleen from healthy female WT C57BL/6 mice followed by dissociation into single cell suspension through triturating, filtration with a 70μm cell strainer, and washing the strainer with 2% FBS/PBS containing 1mM EDTA. Next, RBCs in the splenocytes were lysed using RBC Lysis Buffer (BioLegend, #420302) and T cells were isolated using Dynabeads™ Untouched™ Mouse T Cells Kit (Invitrogen™, #11413D) followed by activation in IL-2 (50 U/mL) and CD3/CD28 beads (2:1 bead-to-cell ratio) for 48 hours. The T cells were then added to the above M2-macrophages and incubated for 18 hours under hypoxic conditions in presence of IL-2, IL-4, IL-13, and CD3/CD28 beads as described above and in the presence or absence of 5 μg/mL mouse anti-FRβ antibody (26). Finally, the cells were stained as described in the “Analysis of tumor growth and immune cell phenotype in WT and FRβ KO mice” section and the % of CD69+ T cells was quantitated as a measure of T cell activation. The % suppression was calculated as (1 - activation with macrophages/activation without macrophages)×100.

TCGA PanCancer RNA-seq datasets for Lung squamous cell carcinoma and kidney renal clear cell carcinoma were explored in cBioPortal (27). Then RNA-seq z-score plots were generated for FOLR2 gene, performed a quartile comparison of all the patients and plotted survival curves for FOLR2 high versus low patients. The z-scores -0.064-0.08 and 0.7-2.42 were considered as FOLR2 low and high, respectively for Lung squamous cell carcinoma and z-scores -0.56-0.09 and 0.66-2.53 were considered as FOLR2 low and high, respectively for kidney renal clear cell carcinoma.

Macrophages were obtained from healthy WT and FR-β KO C57BL/6 mice, seeded at 50,000 cells/well in 8-well chamber slides, and polarized into M2-like macrophages, as described above. After fixation, permeabilization, and blocking as described above, cells were treated with primary anti-RFC1 (reduced folate carrier 1, abcam, Catalog # ab193559, 1/1000 dilution) or anti-PCFT (proton coupled folate transporter, abcam, Catalog # ab25134, 10 µg/mL) antibody at 4°C overnight. The cells were then stained with AF488 conjugated Goat anti-Rabbit IgG secondary antibody (10µg/mL) and Hoechst 33342 nuclear dye (1µg/mL) and then imaged using Nikon A1R-MP multiphoton confocal microscope as described above.

Macrophages were obtained from healthy WT and FR-β KO C57BL/6 mice, and polarized into M2-like macrophages, as described above. The cells were washed with PBS and stained with Zombie Violet dye and FcX as described in the “Analysis of tumor growth and immune cell phenotype in WT and FRβ KO mice” section. Then the cells were treated with primary anti-RFC1 (reduced folate carrier 1, abcam, Catalog # ab193559, 1/1000 dilution) or anti-PCFT (proton coupled folate transporter, abcam, Catalog # ab25134, 5 µg/mL) antibody along with the other surface antibodies (Anti-mouse CD45, CD11b, and F4/80) at 4 °C for 1 hour in dark. The cells were then washed and stained with AF488 conjugated Goat anti-Rabbit IgG secondary antibody (10µg/mL) for 1 hour at 4°C in dark. Followed by washing the cells, flow cytometry analyses were performed as described above. (26, 28, 29)].

Statistical analyses were performed using GraphPad Prism software (V10). Data are presented as mean ± standard error of the mean (SEM). Unpaired two-tailed Student’s t-test was used to compare means between two groups. One-way ANOVA with appropriate post-hoc test was used to compare means among three or more groups. Two-way ANOVA was used to compare means with two independent variables. Log-rank (Mantel-Cox) test was used to compare survival curves. P values less than 0.05 were considered statistically significant. The specific statistical tests used for each experiment are indicated in the figure legends. No statistical methods were used to predetermine sample sizes.

FOLR2 is a myeloid-specific gene for folate receptor beta (FRβ) that is expressed on a few subsets of monocytes and macrophages, but absent from essentially all other cells of the body (30–34). Because most copies of FRβ on myeloid cells in healthy individuals do not bind folic acid ( Supplementary Figure S1 ) (10), and since the majority of FRβ on MDSCs and TAMs in tumor tissues is also nonfunctional, we hypothesized that FRβ might perform a function unrelated to folate transport, perhaps one involving immune regulation. To test this hypothesis, we generated an FRβ knockout (KO) mouse and compared its immunologic functions with those of wild type (WT) mice (9). As shown in Figures 1A-C and Supplementary Figures S2, Figures S3 , insertional mutagenesis of the FRβ (Folr2) gene yielded mice in which the WT FRβ could not be detected, but in which FRα and both major folate transporters (RFC and PCFT) were transcribed ( Figures 1D-F ) and expressed ( Figures 1G–J , Supplementary Figure S3 ) at normal levels. Because RFC and PCFT are known to supply most, if not all, folate needs of mature animals (5), it was not surprising that FRβ knockout mice were viable, fertile, and capable of growing to normal size. In fact, it was not until the KO mice were monitored for longer periods of time that it was observed that a substantial fraction (~20%) displayed spontaneous hair loss ( Figure 2A ), enlarged spleens ( Figure 2C ), and an unresolved dermatitis ( Figure 2B ) that were generally more severe in females than males and characterized by an elevated accumulation of macrophages in the affected tissues ( Supplementary Figure S4 ). Because these symptoms were independent of the folate content in the diet, and since the KO mice had normal levels of the folate transporters that mediate uptake of folate into essentially all cells of the body (3–5), we conclude that inadequate folate is not the cause of these abnormalities. Instead, since the same symptoms can be logically attributed to an immune dysregulation (35), we hypothesized that deletion of FRβ might somehow alter the immunosuppressive functions of the macrophages. The studies below are designed to test this hypothesis.

To determine whether FRβ deletion might alter the immunologic properties of the FRβ KO macrophages, we collected bone marrow from both WT and KO mice and differentiated the myeloid cells into MDSCs or M0-, M1-, or M2-like macrophages (see Methods). As seen in Figure 3A , MDSCs from WT mice secreted high levels of IL-10, whereas IL-10 was undetectable in the culture medium from MDSCs of KO mice. In contrast, IL-6 was more highly produced by the KO mice than WT mice ( Figure 3B ). Since the KO macrophages appeared otherwise normal, and because IL-10 is an immunosuppressive cytokine whereas IL-6 is an immunostimulatory cytokine, the data suggest that there is a shift towards a more inflammatory phenotype in the KO mice ( Figures 3A, B ).

To obtain a more detailed assessment of the immune-related changes in gene expression that might be affected by deletion of FRβ, we next conducted RNA-seq analysis of bone marrow derived M1-like macrophages from KO and WT mice ( Supplementary Figures S5A, B ). As shown in Figure 3C , deletion of FRβ almost uniformly enhanced expression of pro-inflammatory genes (e.g. Cd70, Osmr, Ifnk, Prtn3, Cxcl11, Cd247, Cx3cr1, Il1b, Il6, Il9r, Il12a, Il1r2, etc.), while downregulating expression of both anti-inflammatory genes (Ffar2, Btnl2, Adamdec1, Ccl7, and Lck) and genes involved in tissue repair (Has1, Hbegf, Pou3f1, Hmcn1, Xk, Mrc2). This shift towards a more inflammatory phenotype was further evidenced by increases in the transcription of genes involved in leucocyte migration and motility, i.e. processes required for infiltration of immune cells into inflamed lesions.

Next, to determine whether this global repolarization towards a more inflammatory phenotype might affect the KO mouse’s response to an inflammatory stress, we induced inflammation in the lungs of both WT and KO mice by instillation of bleomycin and then compared their abilities to suppress the consequent emergence of inflammatory symptoms. As shown in Figure 4A , KO mice lost weight more rapidly following bleomycin administration, suggesting they were less capable of handling the inflammatory stress, i.e. consistent with their downregulation of anti-inflammatory markers (i.e., Arg1, Cd206, and Mmp9) ( Figures 4B-D ) and upregulation of a pro-inflammatory marker, iNOS ( Figure 4E ) in their bronchoalveolar lavage macrophages (CD45+F4/80+). Interestingly, the KO mice also exhibited less pulmonary fibrosis than the WT mice ( Figure 4F ), congruent with their diminished tissue repair capacity seen in the gene expression panel ( Figure 3C ). Taken together, these data demonstrate that FRβ plays a central role in regulating gene expression programs associated with macrophage polarization, and that its absence compromises macrophage-mediated immunosuppressive functions critical for controlling inflammation.

Because of the relevance of immune function to tumor growth (36), we next investigated whether syngeneic tumors might grow at different rates in KO and WT mice. As shown in Figure 5A , TRAMP C2 prostate tumors grew more slowly in KO than WT C57BL/6 mice, resulting in an overall survival that was predictably better in the KO than WT mice ( Figure 5B ). Although total white cells (CD45+), myeloid cells (CD45+CD11b+), macrophages (CD45+F4/80+), Tregs (CD45+FoxP3CD4+CD25+) and T cells (CD45+CD3+) did not differ between WT and KO tumors ( Supplementary Figure S6A ), the myeloid cells were FRβ+ in the WT mice but FRβ- in the KO mice was then confirmed by flow cytometry (FA-Cy5) ( Figure 5C ). Moreover, to establish that this inhibition of tumor growth was not an aberrant property of the TRAMP C2 tumor model, we repeated the study in C57BL/6 mice implanted with MC38 colon cancer cells. As shown in Figure 5D , MC38 tumors also grew slower in KO than WT mice. Taken together, these data demonstrate that FRβ is critical to the immunosuppressive function of TAMs and MDSCs.

Next, to determine whether a similar dependence of tumor growth on FRβ expression might also occur in humans, we used the TCGA database to examine whether a correlation might exist between FRβ expression and the probability of survival in human cancer patients. As shown in panels E & F, a negative relationship does indeed exist between FRβ expression and patient survival, arguing that the murine data have relevance to human immuno-oncology. Specifically, in both lung squamous cell carcinoma ( Figure 5E ) and kidney renal clear cell carcinoma ( Figure 5F ), patients with high FOLR2 (FRβ) expression had significantly shorter overall survival compared to those with low expression. For lung squamous cell carcinoma, the Log-rank P value was 0.0076 and the hazard ratio (HR) was 1.685, indicating a 68.5% higher risk of death in the high FOLR2 group. In kidney renal clear cell carcinoma, the Log-rank P value was 0.0192 and the HR was 1.696, again demonstrating significantly worse survival in the high FOLR2 group. These results support the conclusion that elevated FRβ expression correlates with poor prognosis and reinforce the clinical relevance of the findings in our murine models.

Next, to obtain a qualitative indication of the mechanism by which FRβ might influence the tumor microenvironment, we dissociated cells from TRAMP C2 tumors and examined the activation markers of their component T cells by flow cytometry. As shown in Figure 5G , CD69+ T cells were elevated in tumors from the KO mice, suggesting that T cell activation was increased in the absence of FRβ. Congruent with this observation, KO T cells also expressed lower levels of the checkpoint receptor PD-1, and TAMs from the same tumors displayed lower amounts of PD-L1( Figure 5H , Supplementary Figure S6 ).

Next, to explore whether the above changes in T cell phenotype might be mechanistically linked to the aforementioned changes in myeloid cell function, we generated M2-like macrophages from WT and KO mouse bone marrows and compared their capacities to suppress T cell activation in vitro. As shown in Figures 6A-C , M2 macrophages from KO mice were less effective in inhibiting CD69 expression (a T cell activation marker) in CD8+, CD4+ and total T cell populations than macrophages from WT mice, i.e. confirming that FRβ impacts the ability of macrophages to regulate T cell activation. That this intercellular communication could involve physical contact between FRβ on the macrophage and a cognate receptor on the T cell was then demonstrated by showing that blockade of FRβ with an anti-FRβ monoclonal antibody reduced the suppressive function of WT macrophages to the level observed with KO macrophages ( Figures 6A-C , Supplementary Figure S7 ).

Finally, noting that a different isoform of the folate receptor, namely folate receptor delta (FRδ), constitutes a marker for highly immunosuppressive regulatory T cells (Tregs) (37), and recognizing that most FRδ also does not bind folic acid (38), we reasoned that FRδ might similarly perform an immunosuppressive function in Tregs, i.e. much like FRβ in TAMs/MDSCs. To obtain an initial indication of whether this hypothesis might be correct, we generated FRδ knockout mice and compared the growth rates of Tramp C2 tumors in FRδ KO and wild type C57BL/6 mice. As shown in Figure 7A , Tramp C2 tumors indeed grew slower in FRδ KO mice. Moreover, as revealed in Figure 7B , some of these KO mice also developed dermatitis. Taken together, these data suggest that FRδ may also perform an immunoregulatory function.