BXCL701

Talabostat (BXCL701) is an oral small-molecule inhibitor of two intracellular enzymes called DPP8 and DPP9, plus a tumor-stromal enzyme called FAP. It is non-selective and also hits the diabetes-drug target DPP4 and DPP2, which matters for the toxicity profile. BioXcel Therapeutics is running it in combination with Merck's pembrolizumab across cancers where PD-1 blockade alone fails, with current activity in metastatic castration-resistant prostate cancer (including the small cell neuroendocrine variant, called SCNC), metastatic pancreatic adenocarcinoma, and relapsed or refractory AML and MDS [1][2][5][6][7]. The drug has a long history: Point Therapeutics took it into two randomized placebo-controlled Phase 3 trials in second-line non-small cell lung cancer starting in October 2005 (talabostat plus docetaxel, and talabostat plus pemetrexed). Both were halted at interim analysis and the FDA placed the program on clinical hold [10]. BioXcel later in-licensed the asset and relaunched it as an immuno-oncology partner for checkpoint inhibitors [4][10]. The database flags max phase as 3 based on that legacy NSCLC program; every currently enrolling study is Phase 1 or Phase 2, so the regulatory-practical phase is 2. The commercial case rests on whether BXCL701 can turn immunologically cold tumors (prostate, pancreatic, AML) into pembrolizumab responders. The published Phase 2 SCNC data from Aggarwal et al. in J Immunother Cancer, May 2026, show a 25 percent composite response rate (95 percent CI 8.3 to 41 percent, n=28 evaluable) with median overall survival of 13.6 months and 12-month OS of 56.5 percent [1]. BioXcel (ticker BTAI) is a micro-cap in serious financial distress; its June 30, 2026 10-Q discloses $13.8M cash against $104.5M of current debt and includes a going-concern warning [12]. This program is effectively the company's last oncology shot, and the realistic path forward is a partnership, divestiture, or restructuring.

Not a novel compound. Talabostat (then PT-100, Val-boroPro) was advanced into two randomized Phase 3 NSCLC trials by Point Therapeutics in 2005 to 2007, both halted at interim analysis with an FDA clinical hold [10]. Earlier Phase 2 studies in melanoma plus cisplatin and NSCLC plus docetaxel in 2009 were also disappointing [3][4]. BioXcel in-licensed the asset and relaunched it as a pembrolizumab combination partner on the thesis that DPP8/9 inhibition triggers inflammasome-driven immune activation that was missed by the earlier chemo pairings. Current active trials: NCT05703542, a Phase 1 dose-finding study in relapsed or refractory AML and MDS at Dana-Farber (n=12, active not recruiting) [5]; NCT05558982, a Phase 2 at Georgetown combining BXCL701 with pembrolizumab in metastatic pancreatic ductal adenocarcinoma with progression-free survival at 18 weeks as the primary endpoint (n=22) [6]; and NCT03910660, the Phase 1/2 in mCRPC (both adenocarcinoma and small cell neuroendocrine phenotypes) now reported in long-term follow-up in the SCNC subset [1][7]. The Phase 2 basket paper in advanced solid tumors by Ahmed et al. was published in Cancer in 2025 and showed a composite response rate of roughly 11 percent across tumor types with no clean single-indication signal to anchor a registration path [2]. No publicly disclosed FDA breakthrough, fast track (the earlier Point-era fast track designation applied to a prior indication), or orphan designation is currently active for BXCL701. BioXcel 10-Q and 8-K filings through 2026 reference pipeline updates and strategic review but do not announce a Phase 3 initiation [8][12]. The DB phase=3 label reflects historical maximum phase from the halted Point Therapeutics NSCLC program, not the current practical regulatory stage, which is Phase 2.

Three primary targets, one logic: wake up a tumor's immune environment so that PD-1 blockade can finish the job. DPP8 and DPP9 are enzymes that live inside immune cells and clip small proteins. When you block them, you accidentally trip a safety sensor called the NLRP1/CARD8 inflammasome, which forces the cell to die by pyroptosis, a kind of inflammatory suicide that spills danger signals and recruits T cells. In AML blasts and myeloid-lineage cells this pathway is especially active, which is why the AML/MDS trial exists [5]. FAP (fibroblast activation protein alpha) is a cell-surface protease on cancer-associated fibroblasts, the connective-tissue cells that build the stiff, immunosuppressive wall around solid tumors [9]. Blocking FAP is meant to thin out that wall so T cells and antibodies can reach the cancer. Pancreatic and prostate cancer are two tumor types notorious for FAP-rich stroma that keeps checkpoint inhibitors from working. Talabostat is non-selective: in addition to DPP8, DPP9, and FAP it also inhibits DPP4 and DPP2. DPP4 is the target of marketed diabetes drugs like sitagliptin, so DPP4 inhibition by itself is well tolerated, but the combined polypharmacology complicates the dose-response picture. The hallmark on-target toxicities of talabostat (hypotension, edema, cytokine release) are generally attributed to DPP8/9-mediated inflammasome activation rather than DPP4 inhibition. The mechanism is biologically credible but not clinically validated at the pivotal level. FAP has been chased for twenty years with mixed results across antibodies, radioligands, and CAR-T. DPP8/9 inhibition as an oncology strategy rests more on preclinical pyroptosis data than human validation. Talabostat hits all these targets without selectivity, which is both the pitch and the toxicity problem.

The lead-indication study to watch has now reported. Aggarwal et al. in J Immunother Cancer, May 2026, reported long-term follow-up of the Phase 2 BXCL701 plus pembrolizumab arm in small cell neuroendocrine prostate cancer (SCNC) [1]. SCNC is a transdifferentiated form of prostate cancer, meaning the tumor cells have rewired themselves to no longer rely on testosterone signaling, which is what makes it lethal and standard-of-care-resistant after androgen-receptor blockade fails. In 28 evaluable patients, the composite response rate (combining RECIST - a standardized imaging-based measure of tumor shrinkage - with PSA, the prostate-specific antigen blood marker, and circulating tumor cell conversion) was 25 percent (95 percent CI 8.3 to 41 percent). RECIST response rate alone was 20 percent in 25 evaluable patients. Median overall survival was 13.6 months (95 percent CI 10.9 to not reached), and 12-month overall survival was 56.5 percent. Composite median duration of response was 6+ months (range 1.3 to 17.4). In a population with essentially no standard of care, these are meaningful but modest numbers. NCT05558982 at Georgetown pairs BXCL701 with pembrolizumab in metastatic pancreatic adenocarcinoma, n=22, with progression-free survival (PFS) at 18 weeks as the primary endpoint, active not recruiting [6]. The pancreatic study is single-arm with no comparator, so interpretation will depend entirely on how the PFS rate stacks against the roughly 10 to 15 percent historical benchmark for second-line pancreatic cancer. NCT05703542 is the Dana-Farber Phase 1 in relapsed or refractory AML/MDS, n=12, dose-finding with maximum tolerated dose (MTD, the highest dose that does not cause unacceptable side effects) as the primary endpoint [5]. The broader solid-tumor basket data in Cancer 2025 established tolerability and showed a composite response rate of roughly 11 percent across tumor types but no clean single-indication signal to anchor a registration path [2]. No randomized Phase 3 is currently enrolling. No specific calendar readout dates for the pancreatic or AML studies have been publicly disclosed.

Our model estimates a 30% chance this drug is eventually approved. It starts from the historical base rate for Phase 3 drugs in this area (about 57%), then adjusts using ten facts about the trial and sponsor. What moves the number most: it is helped by a non-randomized design; it is held back by smaller-than-typical enrollment for this phase, the sponsor's thin or weak approval record, and weak or limited earlier-phase results. The other facts land near average for this stage, so they leave the estimate roughly where the base rate put it.

Safety: talabostat has a known on-target toxicity profile including hypotension and peripheral edema driven by DPP8/9 inhibition and inflammasome activation. The Phase 3 NSCLC program under Point Therapeutics was halted at interim analysis in part because the therapeutic window was narrow when combined with chemotherapy [10]. Adding pembrolizumab layers immune-related adverse events (colitis, pneumonitis, hepatitis) on top. The DPP4 off-target activity is a theoretical concern for glucose homeostasis but has not been the dose-limiting issue in clinical reports. Efficacy: single-arm Phase 2 studies with composite endpoints and small n (12 to 28 patients) are easy to spin positively and hard to confirm. The 25 percent composite response rate in SCNC [1] is encouraging but the lower CI bound is 8.3 percent, which does not rule out a null effect. Pancreatic cancer in particular has killed many combination-IO programs that looked encouraging in Phase 2. Without a biomarker, every tumor type is a separate bet. Execution: BioXcel is in acute financial distress. The June 30, 2026 10-Q reports $13.8M cash against $104.5M of current debt and a Q2 2026 net loss of $14.7M on $182K of product revenue [12]. First-half 2026 operating cash use was $17.9M (improved from $24.6M a year prior), implying roughly $9M/quarter burn, which gives on the order of 1 to 2 quarters of runway absent new financing. Management has disclosed substantial doubt about going-concern and is pursuing a sale, merger, financing, or other strategic transaction including potential debtor-in-possession financing. A pivotal Phase 3 would cost well over $100M that BioXcel does not have, so the realistic path forward is a partnership, divestiture, or restructuring. Commercial: even a win in SCNC prostate cancer is a small market (a few thousand US patients annually). Pancreatic or AML would be bigger, but payer scrutiny on PD-1 combinations without a biomarker is already tight. The current composite response rate and OS data support an accelerated-approval conversation but not blockbuster pricing power.

Worth watching, with low expectations. The published Phase 2 SCNC data from Aggarwal et al. 2026 give a 25 percent composite response rate (95 percent CI 8.3 to 41 percent), median OS of 13.6 months, and 12-month OS of 56.5 percent in 28 evaluable patients, with composite durations of response of 6+ months [1]. In a population with essentially no standard of care, this is a plausible accelerated-approval story. It is not a blowout, and the wide confidence interval leaves room for the signal to attenuate in a confirmatory study. The pancreatic PFS-at-18-weeks readout from NCT05558982 is the second checkpoint; a rate above 25 percent would be meaningful, below 15 percent would not be [6]. The dominant risk is not clinical, it is corporate. The June 30, 2026 10-Q and going-concern language make clear that oncology development inside BioXcel is unlikely to be funded without a strategic transaction [12]. The most informative near-term event is therefore a partnership or acquisition announcement rather than a clinical readout. If a well-capitalized partner takes BXCL701 and runs a confirmatory Phase 3 in SCNC, this becomes a real program. If no partner emerges, the asset likely stalls regardless of the data. Independently of BXCL701's own merits, the broader FAP space (radioligands, CAR-T, bispecifics) is heating up at larger pharmas, so the mechanism itself may get validated by someone else first.