Checkpoint immunotherapy was supposed to be the answer. Block PD-1, release the T cells, cure the cancer. For some patients, it works beautifully. For most, the tumor shrugs it off.
The problem runs deeper than PD-1. Tumors have multiple layers of immune suppression, and antibodies that block a single receptor on the cell surface can only do so much. What if the real brakes are inside the cell?
PTPN2 and PTPN1 are protein tyrosine phosphatases, intracellular enzymes that quietly dephosphorylate JAK and STAT proteins, dampening the very signaling cascades that make immune cells dangerous to tumors. Lose these phosphatases genetically, and T cells go haywire in the best possible way. The catch: phosphatase active sites have been called undruggable for decades.
This paper says otherwise. Meet AC484, a small molecule that fits inside that "undruggable" pocket with a biochemical IC50 of 1.8 nM against PTPN2 and 2.5 nM against PTPN1, and you can swallow it as a pill.
Getting a drug into a phosphatase active site is genuinely hard. The pocket is highly polar, which means inhibitors need to be polar too, and polar molecules tend to have terrible cell permeability. Previous PTPN1 inhibitors hit sub-10 nM biochemical potency but fell apart in cells, with cellular activity above 10 µM. That gap makes them useless as drugs.
The team used structure-based drug design to thread the needle. The crystal structure of AC484 bound to PTPN2 shows the thiadiazolidinone dioxide moiety forming nine hydrogen bonds with active-site residues including Cys216, Arg222, and Asp182. An isopentyl-amine group reaches further to engage Asp50 and Met256 hydrophobically.
AC484 is a zwitterion: the thiadiazolidinone dioxide NH has a pKa of 0.9, the amine NH has a pKa of 10. This unusual charge distribution is what drives the combination of 86% unbound fraction in mouse plasma and low clearance. It clears renally and biliarily, not hepatically, which is rare and pharmacokinetically favorable.
The result: a cellular EC50 of 176 nM for pSTAT1 induction in B16 melanoma cells, a 15-fold improvement over the precursor compound A-650 (2,651 nM). Oral dosing in mice shows linear, dose-proportional exposure from 3 to 100 mg/kg. The drug gets in, hits the target, and gets out cleanly.
AC484 hits cancer from two directions at once, and that dual mechanism is what makes it different from anything in the clinic.
On the tumor side: AC484 dose-dependently enhances IFNγ-driven growth arrest in B16 cells, phenocopying the effect of genetically deleting both Ptpn2 and Ptpn1. Transcriptomic profiling shows that AC484 plus IFNγ drives the same interferon-stimulated gene (ISG) signatures as Ptpn2/n1-null cells. Antigen presentation goes up too: in B16-OVA cells, AC484 at 0.1 µM increases surface SIINFEKL-H2-Kb MFI after IFNγ stimulation, and pretreated tumor cells are killed more efficiently by OT-I T cells in co-culture. Critically, AC484 alone without IFNγ does not spontaneously activate ISGs, so it amplifies an existing immune signal rather than creating noise.
On the immune side: primary mouse T cells activated with anti-CD3/CD28 show increased CD69+ and CD25+ frequencies, plus higher IFNγ and TNF production, in a dose-dependent manner. The drug also increases LCK and FYN phosphorylation, lowering the TCR activation threshold. In human whole blood from both healthy donors and cancer patients, AC484 dose-dependently increases pSTAT5 after IL-2 stimulation and boosts CD69 expression and cytokine production after TCR stimulation. The cancer patient blood responds just like healthy donor blood. That's the clinical signal you want to see.
AC484 works on both the tumor (sensitizing it to immune attack) and the immune cells (making them more aggressive). Neither effect alone explains the full story.
Four tumor models. Four wins. In B16 melanoma, KPC pancreatic adenocarcinoma, 4T1 mammary carcinoma, and EMT-6 breast carcinoma, oral AC484 induces tumor regression that matches or beats anti-PD-1 therapy. The 4T1 and EMT-6 models are the ones that matter most here: both are resistant to PD-1 blockade, with the anti-PD-1 curve sitting right on top of the untreated control. AC484 breaks through both.
The metastasis data is striking. In a B16 pulmonary metastasis model, untreated and anti-PD-1-treated mice develop detectable lung tumors by day 10. AC484-treated mice develop no detectable disease and hit a 100% survival rate. Anti-PD-1 does not achieve this.
The mechanism is immune-dependent: AC484 controls KPC tumors in wild-type C57BL/6J mice but has no effect in NSG mice, which lack functional T and NK cells. Rechallenge experiments in MC38-cured mice show durable immune memory, with previously treated mice rejecting tumor regrowth that kills naive animals.
Which immune cells matter depends on the tumor. In MHC class I-proficient KPC tumors, CD8+ T cell depletion abolishes AC484 efficacy while NK depletion does not. Flip to Jak1-deficient KPC tumors (which express low MHC-I), and the dependency reverses: NK cells carry the response. One drug, two immune effector arms, context-dependent deployment.
AC484 overcomes PD-1 resistance by engaging both CD8+ T cells and NK cells. The dominant effector depends on whether the tumor can present antigen via MHC-I, giving AC484 a broader coverage profile than checkpoint antibodies.
Single-cell RNA sequencing of 68,060 CD45+ tumor-infiltrating cells from B16 and KPC tumors tells the full story of what AC484 does to the tumor microenvironment. The lymphoid-to-myeloid ratio shifts upward, driven by expansion of CD8+ T cells and NK cells. The M1-to-M2 macrophage ratio increases. MDSCs drop while a new population of ISG-high monocytes appears. The chemokines Ccl22 and Ccl17, which recruit regulatory T cells, fall. Proinflammatory signals including Ifng, Tnf, Cxcl9, and Cxcl10 rise. This is a textbook inflamed tumor microenvironment, built from scratch by a pill.
The T cell story goes deeper. AC484 induces a distinct effector CD8+ T cell population expressing high levels of Gzmb, Prf1, and Ifng that is simply absent in untreated and anti-PD-1-treated tumors. Bulk ATAC-seq and RNA-seq on sorted SLAMF6+ (progenitor) and TIM-3+ (terminally exhausted) T cells show that AC484 opens chromatin at memory-associated loci including Il7r, Sell, and Tcf7, while closing it at exhaustion-associated loci including Tox, Itgav, and Epas1.
The transcription factor motifs enriched in AC484-treated TIM-3+ cells include BATF, JUNB, STAT1, STAT3, and STAT5. Nine of the top ten most enriched motifs match those previously reported in T cells treated with the combination of IL-2 and anti-PD-L1. A gene set derived from that combination therapy is highly enriched in AC484-treated TIM-3+ T cells, while the anti-PD-L1-alone gene set enriches in anti-PD-1-treated cells. AC484 is doing what two drugs do, in one molecule, through JAK-STAT amplification.
Flow cytometry confirms it at the protein level: TIM-3 and TOX expression drop in tumor-infiltrating CD8+ T cells from AC484-treated mice, and pSTAT5 MFI is elevated ex vivo in those same cells. Anti-PD-1 treatment does not increase pSTAT5. The metabolic data closes the loop: AC484 increases both maximal oxygen consumption rate and extracellular acidification rate in activated primary T cells, with higher total mitochondrial content by MitoTracker staining. T cells treated with AC484 for four days in vitro, then washed and transferred into tumor-bearing mice, suppress EL4-OVA tumor growth with four of ten mice achieving complete responses. The drug leaves an epigenetic mark that persists after it is gone.
The phosphatase field spent years being told these targets were undruggable. The active sites were too polar, the inhibitors too impermeable, the cellular activity too weak to matter in vivo. AC484 breaks all three of those assumptions simultaneously, with a cellular EC50 of 176 nM and linear pharmacokinetics down to 3 mg/kg in mice.
From a practical standpoint, the reversibility of a small molecule matters. Immune infiltrates in rat toxicology studies at 300 mg/kg resolved within 28 days of stopping treatment. That kind of on/off control is not available with antibody-based therapies, where the drug persists for weeks after the last dose. For managing immune-related adverse events, that difference is clinically meaningful.
The breadth of the immune response is what sets AC484 apart from PD-1 blockade. By amplifying JAK-STAT signaling across T cells, NK cells, macrophages, and dendritic cells simultaneously, it can engage whichever effector arm is available in a given tumor context. Tumors that have lost MHC-I expression to escape T cell killing are still vulnerable to NK surveillance. Tumors with JAK1 mutations that block IFNγ sensing are still controlled. The drug routes around the resistance mechanisms that defeat checkpoint inhibitors.
AC484 is currently in Phase I clinical trials as monotherapy and in combination with anti-PD-1 in patients with advanced solid tumors (NCT04777994). The preclinical data suggests the combination arm is worth watching: in the CT26 model, AC484 plus anti-PD-1 outperforms either agent alone. If the human data holds, this could represent a genuinely new class of cancer immunotherapy built on a target that was written off for a generation.