Targeted therapy of pyrrolo[2,3-d]pyrimidine antifolates in a syngeneic mouse model of high grade serous ovarian cancer and the impact on the tumor microenvironment

Novel therapies are urgently needed for epithelial ovarian cancer (EOC), the most lethal gynecologic malignancy. In addition, therapies that target unique vulnerabilities in the tumor microenvironment (TME) of EOC have largely been unrealized. One strategy to achieve selective drug delivery for EOC therapy involves use of targeted antifolates via their uptake by folate receptor (FR) proteins, resulting in inhibition of essential one-carbon (C1) metabolic pathways. FRα is highly expressed in EOCs, along with the proton-coupled folate transporter (PCFT); FRβ is expressed on activated macrophages, a major infiltrating immune population in EOC. Thus, there is great potential for targeting both the tumor and the TME with agents delivered via selective transport by FRs and PCFT. In this report, we investigated the therapeutic potential of a novel cytosolic C1 6-substituted pyrrolo[2,3-d]pyrimidine inhibitor AGF94, with selectivity for uptake by FRs and PCFT and inhibition of de novo purine nucleotide biosynthesis, against a syngeneic model of ovarian cancer (BR-Luc) which recapitulates high-grade serous ovarian cancer in patients. In vitro activity of AGF94 was extended in vivo against orthotopic BR-Luc tumors. With late-stage subcutaneous BR-Luc xenografts, AGF94 treatment resulted in substantial anti-tumor efficacy, accompanied by significantly decreased M2-like FRβ-expressing macrophages and increased CD3+ T cells, whereas CD4+ and CD8+ T cells were unaffected. Our studies demonstrate potent anti-tumor efficacy of AGF94 in the therapy of EOC in the context of an intact immune system, and provide a framework for targeting the immunosuppressive TME as an essential component of therapy.

The tumor microenvironment (TME) has emerged as a key component in disease progression and therapy resistance in high grade serous ovarian cancer (HGSOC), the most common subtype of EOC 14 . The peritoneal spread of HGSOC creates a unique TME that promotes a complex interplay between the malignant ascites and surrounding tissues. This allows for tumor progression and immune evasion 14 . The predominant innate immune cellular component in ovarian cancer-associated ascites is the tumor-associated macrophage (TAM) population, which contributes to an immunosuppressive environment 15,16 . TAMs also play an important role in metastasis and angiogenesis by releasing proangiogenic factors (e.g., vascular endothelial growth factor, matrix metalloproteinase) 17,18 . Thus, inhibiting TAMs could in principle suppress tumor progression. Interestingly, ovarian cancer-associated TAMs express FRβ, affording an opportunity to inhibit TAMs via the selective uptake of FR-targeted therapeutics 19,20 .
De novo purine biosynthesis is a critical pathway in tumor cells as purine depletion limits ATP and GTP for DNA synthesis and repair, and for cellular energetics. Our laboratory previously described novel 6-substituted pyrrolo [2,3-d]pyrimidine antifolate compounds that inhibit one-carbon (C1) metabolism in de novo purine biosynthesis, resulting in potent in vitro anti-tumor efficacy in tumor models including EOC 10,21,22 . While substantial in vivo efficacy was reported in these studies with EOC xenografts in immune-compromised mice 10,21 , anti-tumor activity in the presence of an intact immune system was not explored. Our lead analogs (i.e., AGF94, AGF278, and AGF283; Fig. 1) inhibited glycinamide ribonucleotide (GAR) formyltransferase (GARFTase), the first folate-dependent step in the 10 reactions comprising de novo purine biosynthesis, resulting in perturbations in purine precursors and decreased pools of purine nucleotides 10,[21][22][23] . Anti-tumor activity of this series of compounds including EOC reflects their endocytosis by FRα and/or their facilitative uptake by PCFT 10,21 .
Reflecting the urgent need for new therapies to target this complex and highly aggressive tumor, in this report we investigated the impact of our novel FR-and PCFT-targeted pyrrolo [2,3-d]pyrimidine agents on both the tumor and the TME in HGSOC. To directly examine the effects of an intact immune system on the anti-tumor efficacy of our agents, we used a novel syngeneic mouse model of HGSOC (BR-Luc) 24,25 . The BR-Luc model was developed from the BR-5 EOC subline and both models are characterized by knockout of BRCA1 and p53, although BR-Luc cells also express a luciferase reporter 24,25 . Importantly, the BR-Luc EOC model recapitulates human HGSOC histology and patterns of metastasis, as well as responses to therapy 25 . We used the BR-Luc syngeneic model to investigate the relation between anti-tumor efficacy and the presence of immune infiltrates, including the impact on FRβ-expressing TAMs and CD3+, CD4+ and CD8+ T cells, accompanying treatment with the pyrrolo[2,3-d]pyrimidine antifolate inhibitor AGF94. Our results are impactful in that they further establish the translational potential of this novel series of compounds against a clinically relevant syngeneic model of HGSOC. They also provide proof-of-concept that targeting M2-like macrophages in the TME of HGSOC via FRβ likely contributes to in vivo efficacy of FR-targeted inhibitors of this class. In vitro efficacy of C1 inhibitors for BR-5 and BR-Luc murine HGSOC cells. We previously established an extensive structure-activity-relationship (SAR) profile for 6-substituted pyrrolo [2,3-d]pyrimidine compounds with modifications in the side chain, including the bridge length and the nature of the aromatic moiety, as selective transport substrates for FRs and PCFT, and as inhibitors of de novo purine nucleotide biosynthesis 13,27,29 . To further characterize the murine HGSOC models BR-5 and BR-Luc, we tested the in vitro antiproliferative activities of our lead pyrrolo [2,3-d]pyrimidine C1 inhibitors AGF94, AGF278 and AGF283 21,22 (Fig. 1). Results with BR-5 and BR-Luc cells were compared to those for IGROV1 human EOC cells (Table 1) in the absence and presence of excess folic acid (200 nM), which selectively blocks FR-mediated drug uptake without effects on RFC or PCFT 10 . BR-5 and BR-Luc cells showed potent inhibition by all the compounds with IC 50 values ranging from ~ 6 nM for AGF94 to ~ 75 nM for AGF283 (both with BR-Luc cells) (Table 1). Further, inhibitions of the murine EOCs by all compounds approximated those for IGROV1 human EOC cells. For both murine and human EOC cells, inhibitions were significantly decreased by excess folic acid, establishing their cellular uptake by FRα with secondary uptake by PCFT 10,23 . In vivo analysis of efficacy for AGF94 with intraperitoneal BR-Luc EOC. An  www.nature.com/scientificreports/ subcutaneous (SC) human EOC xenografts in an immune-compromised SCID mouse model 10,22 to evaluation of a syngeneic FVB mouse model of HGSOC. We selected the lead compound AGF94 and chose the BR-Luc EOC model for further studies based on its in vitro sensitivity to AGF94 which paralleled that for IGROV1 human EOC cells (Table 1). Further, the BR-Luc tumor can be imaged by luminescence detection. As proof-of-concept of antitumor efficacy of AGF94 in mice with intact immune function and to recapitulate the clinical-pathologic features of HGSOC, we initially used intraperitoneal (IP) engraftment of BR-Luc tumors in female FVB mice. IP presentation of the BR-Luc tumor allows for dissemination throughout the peritoneum, as occurs in patients.
For all in vivo studies, mice were maintained on a low-folate diet to reduce highly elevated serum folate concentrations (from the standard folate-replete diet) to levels approximating those in humans 10,21,22 . BR-Luc cells (5 × 10 6 /mouse) were injected IP on day 0, and the mice were non-selectively distributed to each arm (control and AGF94-treated (n = 4), with separate matching cohorts for imaging (n = 3)). AGF94 was administered IV (Q4dx4 at 32 mg/kg/injection) beginning on day 4, for a total dose of 128 mg/kg (Fig. 4B). AGF94-treated mice sustained a median 5.5% body weight loss nadir on days 18 and 26 with full recovery by day 28. Disease progression and anti-tumor efficacy were monitored by weighing mice daily, palpating and measuring IP tumor masses, observing overt symptoms and luminescence imaging. Mice were euthanized at the disease end point, characterized by abdominal distension and onset of labored breathing due to ascites accumulation (> 1-2 mL), and/or the presence of palpable tumor masses > 5% of body weight (i.e., up to 1 g of cumulative solid tumor burden). Upon necropsy, accumulation of hemorrhagic ascites (1-3 mL) was observed, accompanied by disseminated metastatic disease, involving numerous small tumor nodes which localized to the ovaries, mesenteric lymph nodes and adipose tissue associated with the pancreas and GI space. and AICARFTase (B) were measured using cDNAs from primary specimens including normal ovary (n = 8) and HGSOC (n = 39) (OriGene) and results were compared to those for EOC cell lines including IGROV1, SKOV3, A2780 and A2780 E-80. Transcript levels were normalized to β-actin transcripts. Statistical analyses were performed between normal samples/tissues and tumor samples/tissues using the Wilcoxon rank-sum test. (C,D) Transcript levels of cytosolic C1 metabolic targets (GARFTase and AICARFTase) were determined in the BR-5 and BR-Luc syngeneic mouse models of HGSOC by real-time RT-PCR. Transcripts were normalized to β-actin transcripts and results are shown relative to levels in mouse liver (assigned a value of 1). Results are presented as mean values ± standard errors from at least three experiments. The p values are as follows: ****p < 0.0001. See Supplementary Table S1  www.nature.com/scientificreports/ AGF94 was active against the BR-Luc orthotopic IP model (Fig. 4A,C). Following 4 treatments with AGF94, tumors were undetectable by luminescence imaging (Fig. 4C shows imaging of the matched control and AGF94treated cohorts including 3 mice/group), although the tumors regrew following cessation of therapy. The median days to death in the control mice were 20.25 days (range 17.5-22 days) compared to 34.25 days (range 30.5-40.5 days) for mice treated with AGF94. This produced a 69% increased lifespan (%ILS) and 2.1 logs of gross cell kill based on a 2-day doubling time for IP implanted cells. Results are presented as mean values (± standard errors) from at least three experiments. Table 1. IC 50 values (in nM) for anti-proliferative activities of novel compounds on HGSOC mouse models BR-Luc and BR-5 compared to cisplatin. In vitro proliferation assays were performed for BR-5 and BR-Luc mouse HGSOC tumor cells (both express RFC, FRα, and PCFT). Anti-proliferative studies were performed in folate-free RPMI1640 media with 10% dialyzed FBS and 25 nM leucovorin. To measure FR-mediated drug delivery, experiments were performed in the presence or absence of 200 nM folic acid. Results are presented as IC 50 values, corresponding to the concentrations that inhibit growth by 50% relative to cells incubated without drug. The data are presented as mean values from at least 3 experiments (± standard errors). The pyrrolo[2,3-d] pyrimidine antifolates were previously described 21,22 . ND not done. www.nature.com/scientificreports/

In vivo anti-tumor efficacy analysis of AGF94 against advanced-stage subcutaneous BR-Luc and impact on the infiltrating immune microenvironment. We extended our initial in vivo studies
with AGF94 to advanced-stage SC implanted BR-Luc tumors in FVB mice which permits careful monitoring of the relationships between anti-tumor efficacy and the impact of drug treatment on the infiltrating immune microenvironment without considerations of metastatic spread. FVB mice were maintained on a folate-deficient diet and were implanted bilaterally SC with BR-Luc tumor fragments by trocar on day 0 (Fig. 5B). The tumors were measured on day 7 and the mice were then non-selectively distributed into treatment and control groups (6 mice/group) for determinations of antitumor efficacy, with parallel cohorts to accommodate imaging (n = 4, 2 each for control and AGF94-treated groups) and analysis of tumor immune infiltration (6-9 each for control and AGF94-treated groups). The median tumor burden for BR-Luc advanced stage disease was 388 mg (range 356-412 mg; day 7). Matching control and treated cohorts (n = 4 mice per group) maintained on a standard (folate replete) mouse diet had a median tumor burden of 218 mg (range 213-223 mg; day 7). The mice were dosed with AGF94 (32 mg/kg IV on a Q4dx4 schedule), starting on day 8. Figure 5B summarizes the treatment scheme and Fig. 5A shows the luminescence detection of BR-Luc tumors on day 7 before initiation of AGF94 treatment. All mice were weighed daily for the duration of study, as well as observed and assessed prior to each treatment or imaging. Tumor growth was monitored two-to-three times weekly until the tumor burden endpoint (5-10% of body weight) was reached. For mice maintained on the folate-deficient diet, anti-tumor activity 3 days post treatment with AGF94 (day 11) was reflected in a %T/C value of 29.5% (median) ( Fig. 5D; further determinations of %T/C were not possible due to euthanization of controls on or after day 11, reflecting a doubling time (Td) of 1.1 days for the SC BR-Luc tumor). For quantitative assessments of anti-tumor efficacy, tumor growth delay (T-C) was used (defined as the www.nature.com/scientificreports/ median time for the tumor to reach 1 g for treated (T) and control (C) groups). AGF94-treated mice yielded a median 11.5 day tumor growth delay and 3.15 logs of gross cell kill (defined as T-C/3.32 × Td) (Fig. 5C,D). Again, AGF94 treatment was well tolerated with the only dose-limiting symptom being weight loss (nadir of 6.4% on day 18 and full recovery by day 26). For the mice maintained on the standard (folate-replete) diet, AGF94 was inactive (%T/C > 100%, Fig. 5D; i.e., there was no weight loss or other adverse symptoms observed). An important goal of our study was to explore the impact of 2, 3 or 4 doses of AGF94 upon the tumor microenvironment, including infiltrating lymphocytes accompanying its anti-tumor effects. Further, since AGF94 is transported in part by FRs, including FRβ 22 , we measured its effects on the FRβ-expressing TAMs based on published reports that targeting macrophages via FRβ may have therapeutic potential for treating inflammatory diseases and cancer 19,20,30 .
To assess the effects of drug treatment upon the infiltrating immune population, tumors and spleens (as a control) were harvested from parallel cohorts of AGF94-treated mice after 2, 3 and 4 injections, as well as from untreated control mice. The impact of drug treatment was initially determined on the total macrophage population, defined as dual expressing CD11b+ and F4/80+ cells from live CD45+ cells including the FRβ-expressing population (Fig. 6A,B). We measured the FRβ-expressing CD11b+/F4/80+ cells with FRβ-specific antibody (Supplementary Materials, Figure S1). After treatment with AGF94, there were statistically significant decreases after 2 and 3 doses on the CD11b+/F4/80+/FRβ+ population which appeared to diminish with subsequent dosing compared to untreated mice.
We also examined the impact of AGF94 dosing on the infiltrating T cell populations. Treatment with AGF94 was accompanied by a statistically significant increase (~ 50-100%) in the CD3+ T cells, as defined by the percentage of live CD45+ cells (Fig. 7A). There was no impact on the proportions of CD4+ and CD8+ T cells in the www.nature.com/scientificreports/ tumors (Fig. 7B,C) and spleens (not shown) (measured as a percentage of CD3+ T cells) after 2, 3, or 4 doses of AGF94.
As validation of our flow cytometry results and to assess the spatial landscape of the immune infiltrate, we performed immunofluorescence (IF) staining on tumor sections from control and drug-treated mice following 4 doses of AGF94. Both CD3+ and CD8+ T cells were detected in sections from AGF94-treated mice, albeit to variable extents, consistent with the flow cytometry results (Supplementary Materials, Figure S2). Importantly, T cells were disseminated throughout the tumor bed regardless of treatment status, substantiating the quantitative flow cytometry results which suggest that these cells are not diminished by treatment with AGF94. Dual staining was performed with FRβ antibody and pan-macrophage F4/80+ antibody. As expected, FRß positive cells were most easily detected in the sections from untreated mice, with substantially fewer FRß positive cells in the AGF94-treated sections (Fig. 8).

Discussion
The tumor microenvironment contains a host of infiltrating immune cells, including TAMs and T-lymphocytes, with TAMs considered the principal immune cellular component which results in an immunosuppressive environment 15,16 . FRβ is expressed on IL-10-producing M2-like macrophages (CD163+, CD68+, CD14+ IL-10), corresponding to the anti-inflammatory/pro-tumor TAM subtype, prompting substantial interest in depleting TAMs by exploiting FRβ on the surface of macrophages 19 . A BIM (BCL-2-interacting mediator of cell death) plasmid encapsulated in a folate "lipoplex" was developed to target the tumor microenvironment in lung cancer 30 and an anti-mouse FRβ monoclonal antibody conjugated to Pseudomonas exotoxin A depleted TAMs and reduced tumor growth a C6 rat glioma model 33 . Further, a folate-conjugated TLR7 agonist showed in vivo activity in assorted tumor models and reversed expression of a high M2-like to M1-like macrophage ratio and increased the infiltration of cytotoxic CD8 T cells 20 .
The present study explored the potential of the novel pyrrolo[2,3-d]pyrimidine antifolate AGF94 for dual targeting HGSOC directly, as well as indirectly via its effects on the tumor microenvironment. AGF94 is a prototype FR-and PCFT-targeted pyrrolo[2,3-d]pyrimidine antifolate previously reported to show broad-ranging  www.nature.com/scientificreports/ anti-tumor efficacy (including human EOC xenograft models) 9,10,12,21 . AGF94 is potent inhibitor of GARFTase 22,34 , the first folate-dependent step in de novo purine biosynthesis, a critical anabolic pathway in malignant cells 35 . Inhibition of purine nucleotide biosynthesis kills tumors independent of the wild-type/mutant p53 status 36,37 , shows tumor selectivity based on impaired purine salvage 38,39 , and results in suppression of mTOR signaling 40,41 . We used a syngeneic immunocompetent FVB mouse model of orthotopic HGSOC that accurately recapitulates the histology and progression of human HGSOC 24,25 and found that AGF94 effected potent in vivo anti-tumor efficacy. Our results with the orthotopic IP BR-Luc EOC model were extended to an advanced stage SC model of BR-Luc so as to better study the changes in the TME immune populations accompanying treatments with AGF94. Anti-tumor activity was accompanied by a direct impact on the TME including significantly decreased FRβ-expressing TAMs, and no evidence of CD3+ T cell depletion or impact on the relative proportions of CD4+ and CD8+ T cells after drug treatment.
Based on these results, we suggest that an FRβ-transported C1 inhibitor such as AGF94 represents an exciting new approach for therapy of HGSOC through its ability to directly target the tumor via uptake by FRα and PCFT, and its effects on the TME, particularly FRβ-expressing TAMs. Of course, our results also raise important new questions that will be the focus of future studies. For instance, what are the relative contributions in targeting tumors with AGF94 via FRα and PCFT versus FRβ-expressing TAMs to the net anti-tumor response observed, and are the CD3+ T-cells functional after treatment? As PCFT was reported to be expressed with FRβ in M2-like macrophages 42 , is there an additional role for PCFT in the uptake of our targeted agents in the pro-tumor M-2 like subtype of TAMs. While FRβ-targeting may be important for decreasing TAMs in response to AGF94 treatment, the role of C1 metabolism in the inflammatory response is still evolving. Indeed, C1 metabolism could represent in a unique vulnerability for TAMs as serine metabolism was associated with an inflammatory response in macrophages via its impact on glutathione synthesis 43 and S-adenosyl methionine depletion was reported to inhibit inflammatory macrophages 44 . Of course, the impact of purine inhibitors such as AGF94 may have a broader impact, via effects on adenosine pools in the tumor microenvironment from tumor cells which act as an immune modulator resulting in macrophage polarization 45 and inhibition of cytotoxic effector functions of NK and CD8+ T cells [46][47][48] . Future studies will explore the general impact of treating HGSOG with AGF94 and related agents on C1 metabolism in relation to direct targeting tumor cells vis á vis inflammatory macrophages and T-cells in therapeutic response.

Materials and methods
All methods were performed in accordance with the relevant guidelines and regulations.  [2,3-d]pyrimidine antifolates AGF94, AGF278 and AGF283 were synthesized as previously described 21,22 . Additional chemicals were purchased from commercial sources in the highest available purities.

EOC cell lines and antiproliferative experiments. BR-5 and BR-Luc murine EOC cells were gener-
ous gifts from Dr. Sandra Orsulic (UCLA) 25,49 . KB nasopharyngeal carcinoma and SKOV3 cells were obtained from the American Type Culture Collection (Manassas, VA). IGROV1 (NCI-IGROV1) (passage 5) clear cell carcinoma cells were obtained from the Division of Cancer Treatment and Diagnosis, National Cancer Institute (Frederick, MD). A2780 and A2780-E80 cells were generous gifts from Dr. Gen-Sheng Wu (Karmanos Cancer Institute, Detroit, MI). All the cell lines were cultured at 37 °C under 5% CO 2 in complete folate-free RPMI 1640 supplemented with 10% fetal bovine serum (FBS) (Sigma-Aldrich; St. Louis, MO) and 100 units/ mL penicillin/100 μg/mL streptomycin, and 2 mM L-glutamine. All cell lines were authenticated by STR analysis by Genetica DNA Laboratories (Burlington, NC) and tested for Mycoplasma by PCR using a Mycloplasma testing kit (Venor™ GeM Mycoplasma Detection Kit, Sigma). Frozen stocks were generated from authenticated mycoplasma-free cultures.
Cell proliferation assays were performed as described 10,26 . Cells were plated in 96-well dishes at densities ranging from 2500 to 5000 cells/well in 200 μL media and treated with a range of inhibitors spanning concentrations from 0 to 1000 nM. Experiments used BR-5 and BR-Luc cells and folate-free RPMI 1640 media with 10% dialyzed FBS and 100 units/mL penicillin/100 μg/mL streptomycin, supplemented with 25 nM leucovorin and 2 mM L-glutamine. FR-mediated drug uptake was assessed in parallel incubations including 200 nM folic acid. Cells were treated over a 96 h period at 37 °C with 5% CO 2 . Relative cell numbers were quantified using the CellTiter-blue cell viability assay (Promega, Madison, WI) and a fluorescence plate reader 26 . Raw data were exported to Excel for analysis and the results plotted using Graphpad Prism 6.0. Determinations of IC 50 s were made corresponding to the drug concentrations that resulted in 50% loss of cell growth.

Real-time RT-PCR of transcripts for the major folate transporters and C1 metabolic targets.
RNAs were isolated from the murine (BR-5, BR-Luc) and human (IGROV1, SKOV3, A2780 E-80, and A2780) EOC cell lines using TRIzol reagent (Life Technologies). cDNAs were synthesized with random hexamers and MuLV reverse transcriptase (including RNase inhibitor; Applied Biosystems), and purified using a QIAquick PCR Purification Kit (Qiagen).
Human cDNAs were purchased from Origene (HORT502) containing 48 lyophilized cDNAs from EOC patient specimens (8 stage I, 9 stage II, 17 stage III, and 6 stage IV) and 8 cDNAs from normal ovaries. Patient pathology characteristics are in Supplemental Table S1. www.nature.com/scientificreports/ Quantitative real-time RT-PCR was performed using a Roche LightCycler 480 real-time PCR (Roche Diagnostics) with gene-specific primers for mouse PCFT, RFC, GARFTase, AICARFTase, and FRα, or human GARFTase and AICARFTase, as appropriate (Supplemental Table S2), and FastStart DNA Master SYBR Green I Reaction Mix (Roche Diagnostics, Indianapolis, IN). Transcript levels were normalized to β-actin transcripts. For the murine transcripts, levels were normalized to transcript levels in mouse liver. FRα levels were measured for the KB,   IGROV1, SKOV3, BR-5, and BR-Luc cells with a functional readout involving measuring [ 3 H]folic acid binding to surface FRs 10 . Cells were plated at a density of 1-2 × 10 6 cells in complete folate-free RPMI1640 (10% FBS) media. Cells were allowed to adhere to the plates for 24 h. The following day, the cells were washed (3x) with 4 °C Dulbecco's phosphate-buffered saline (DPBS). Cells were washed with acetate buffer (10 mM sodium acetate, 150 mM NaCl, pH 3.5) (3 ×) at 4 °C to release FRα-bound folates, then neutralized with Hepes-buffered saline (HBS) (20 mM Hepes, 140 mM NaCl, 5 mM KCl, 2 nM MgCl 2 , and 5 mM glucose, pH 7.4) at 4 °C. The cells were incubated at 0 °C for 15 min with [ 3 H]folic acid (50 nM, specific activity 0.5 Ci/mmol), in the presence or absence of non-radioactive folic acid as a competitor (10 µmol/L). After a 15 min incubation, the cells were washed with ice-cold HBS and proteins were solubilized with 0.5 N NaOH. Cell homogenates were assayed for radioactivity with a liquid scintillation counter and protein concentrations were measured using the Folinphenol reagent 50  PCFT transport assays. PCFT transport assays were performed with IGROV1, BR-5 and BR-Luc cells.

[ 3 H]Folic acid binding as a measure of surface FRs. Total
Cells were plated in 60 mm dishes containing complete folate-free RPMI 1640 with 10% FBS, including 2 mM L-glutamine, and antibiotics; cultures were used when they were 80-90% confluent. Uptake of [ 3 H]MTX (at 0.5 µM) was measured over 5 min at 37° C in MES-buffered saline (20 mM MES, 140 mM NaCl, 5 mM KCl, 2 mM MgCl 2 , and 5 mM glucose; pH 5.5). Under these conditions, uptake by FRs is negligible. To quench transport fluxes, dishes were washed 3 times with ice-cold DPBS. Cells were solubilized in 0.5 N NaOH and radioactive contents and protein concentrations 50 were determined. Uptake was expressed as pmol [ 3 H]MTX per mg protein. To confirm PCFT-mediated transport activity, 10 µM nonradioactive AGF94 was added to the transport incubations to block PCFT uptake.
In vivo studies. The mouse studies were approved by the Wayne State University Institutional Animal Care and Use Committee (IACUC). Syngeneic female FVB mice were purchased from Charles River Labs (Wilmington, MA) or Envigo (Indianapolis, IN). In vivo tumor maintenance and drug efficacy studies for SC BR-Luc tumors are completely analogous to those previously described 9,10,21,22 .
For the in vivo therapeutic trials, study mice were maintained on either a folate-deficient diet from Har-lanTeklad (Envigo, Indianapolis, IN) (catalog # TD.00434) or a folate-replete diet from Lab Diet (catalog # 5021; autoclavable mouse breeder diet) (for both SC and intraperitoneal (IP) trials) starting 9-14 days before tumor implant (depending on tumor staging). Mice were supplied with food and water ad libitum. Serum folate levels were determined prior to tumor implant and post study via Lactobacillus casei bioassays 51 .
In vivo studies using the BR-Luc tumor were performed using both IP and SC presentations depending upon the desired study objective. The BR-Luc tumor was first established subcutaneously from cultured cells with implanted donor mice used to set up the efficacy studies. BR-Luc tumors from donor mice were aseptically harvested, mechanically dissociated into single cell suspensions, centrifuged, and suspended in sterile chilled saline and injected IP into study mice (5 × 10 6 /0.2 mL per mouse) on day 0. The mice were unselectively distributed into control and AGF94-treated groups (4 mice/group for efficacy; a separate cohort of 3 mice/group was used for imaging). Treatment (IV tail vein; 0.2 mL volume) with AGF94 was initiated 4 days post-tumor implantation at 32 mg/kg injection on a Q4dx4 schedule (total dose of 128 mg/kg). The mice were weighed and observed daily for symptoms from drug treatment and disease onset (abdominal distention, palpable internal masses), with periodic monitoring of internal progression by bioluminescent imaging. To evaluate the qualitive efficacy of AGF94 treatment, mice were imaged 24 h after receiving 2, 3, and 4 doses of drug. Imaging was performed with a Bruker CareStream in vivo Xtreme (Billerica, MA). Mice were injected IP with 150 mg/kg of XenoLight D-luciferin bioluminescent substrate (Perkin Elmer; catalog #122799). Imaging was initiated 10 min after D-luciferin injection after anesthetization with isoflurane with 2% oxygen (Fluriso; VetOne) (3% for induction and 1.5% for maintenance). For the IP study, mice were euthanized at defined disease endpoints (> 2 mL ascites, internal cumulative tumor mass up to 1 g or at onset of lethargy or respiratory impairment).
For the SC efficacy trial, mice were implanted bilaterally along the flanks via sterile 12-gauge trocar with 30 mg tumor fragments aseptically harvested from tumor donor mice on day 0. Tumors were measured every 3-4 days. On day 7, when tumor burdens approached 350-450 mg (by caliper; representative of advanced stage disease), the mice were non-selectively distributed into control and treatment groups (n = 6/group), with parallel cohorts included for imaging (n = 2/group) and analysis of immune infiltration (6-9 mice/group). Luminescent imaging was performed on day 7 to establish a baseline for advanced disease. Treatment with AGF94 was initiated on day 8 on a Q4dx4 schedule at 32 mg/kg IV injection (days 8,12,16,20). Mice were weighed and observed daily and tumors measured by caliper 2-3 times weekly, with weekly luminescent imaging for parallel cohorts. All mice were euthanized at harvest or study endpoint when the tumor burden reached 5-10% of the body weight. Tumor volumes (in mg) were calculated using the formula, length × width 2 /2 (ascertained from caliper measurements in mm, where mg = mm 3 ). Total tumor burden per mouse (in mg) was determined by addition of tumor volumes on the right and left flanks. Median tumor burden was determined on each measurement day for each group. These values were used in the efficacy analysis (below). www.nature.com/scientificreports/ Quantitative endpoints for the IP study include tumor growth delay (T-C) using the median survival time (median time of sacrifice) in days for treated (T) and control (C) groups, and the percent increase in lifespan (% ILS) using the formula T-C/C × 100. Log cell kill calculations were performed as described for the SC study below. Tumor volume doubling times (Td) were determined by best fit straight line from log-linear growth plot of control group tumors in log growth phase (100-800 mg for SC tumors) or by the difference in median survival times in days of titered no treatment controls (i.e., 10 6 and 10 4 cells, using the formula Td = median survival in days for 10 6 − median survival time for 10 4 /3.32 × 2).
Quantitative endpoints for the SC study include: (1) tumor growth delay [T-C, where T is the median time in days required for the treatment group tumors to reach a predetermined size (e.g., 1000 mg), and C is the median time in days for the control group tumors to reach the same size; tumor-free survivors are excluded from these calculations]; (2) gross log 10 cell kill (LCK), determined by the formula LCK = (T-C; tumor growth delay in days)/3.32 × Td (tumor doubling time in days determined by growth plot); and (3) T/C values (in percent), corresponding to periodic caliper measurements using the median tumor burden for treatment (T) and control (C) groups when control tumors are still in exponential growth phase (i.e., 500-1250 mg).
Tumor and spleen dissociation and flow cytometry analysis of immune populations. After 2,3, and 4 treatments of AGF94 for the SC treatment trial, mice from the respective study arms were euthanized and tumors and spleens were dissociated. Tumors were dissociated using a GentleMACS dissociator (Miltenyi Biotec, Bergisch Gladbach, Germany) and the manufacturer's protocol for mouse tumor dissociation. Following dissociation, tumors and associated immune infiltrate cells were resuspended with 1.5% FBS in 1 × DPBS for flow cytometry.

Statistical analyses.
All data reflect at least three biological replicates unless noted otherwise. For in vitro cell-based assays, expression levels were assessed using unpaired t-test after log 2 transformation to meet the normality assumption. For transcript analysis, a nonparametric Wilcoxon rank sum test was performed. For survival curves between control and AGF94-treated mice with IP BR-Luc tumors, a log-rank test was performed. For flow cytometry, a nonparametric Wilcoxon rank-sum test was performed. Statistical analyses were performed using Excel and Graphpad Prism 6.0.