LHP588

LHP588 is a Phase 2 small molecule from Lighthouse Pharmaceuticals being tested in 300 mild-to-moderate Alzheimer's patients who test positive for Porphyromonas gingivalis - the bacterium behind gum disease [1]. NCT06847321 is recruiting now, with the primary cognitive endpoint (ADAS-Cog11) read out at 40-48 weeks. The trial follows directly from a hypothesis that crashed hard four years ago: Cortexyme's atuzaginstat (COR388) hit the same biological target, received an FDA partial clinical hold for hepatotoxicity in April 2021 [3], continued under a modified protocol, and then announced negative Phase 2/3 GAIN trial results in November 2021 [2]. LHP588 is the second clinical attempt to prove P. gingivalis matters in Alzheimer's, this time with a prospectively biomarker-selected population (only patients confirmed P. gingivalis-positive at enrollment) and presumably a cleaner chemical structure. If it works, it would be the first AD drug targeting an infectious cause and a multibillion-dollar franchise opportunity. If it fails, the bacterial hypothesis in AD is probably retired for the rest of the decade.

LHP588 is a novel small molecule that has never been approved anywhere. Phase 2, currently recruiting per ClinicalTrials.gov, with an enrollment target of 300 patients [1]. No public FDA designations - no breakthrough therapy, no fast track, no orphan that I can verify in current FDA databases. Lighthouse Pharmaceuticals is private, so financial disclosures are limited and there is no S-1 or 10-K to read; their funding stage, total capital raised, and runway are not publicly disclosed in our enrichment data - this is a material unknown for a 300-patient biomarker-enriched AD trial that typically runs $30-60M+. The SEC filings tied to the broader gingipain program in our enrichment data are all from Quince Therapeutics (formerly Cortexyme), which suggests either a continued IP relationship, a licensing arrangement, or simply that Quince still files quarterly because it inherited assets from the original atuzaginstat program when it rebranded in 2022 [4]. Lighthouse's exact corporate relationship to those original Cortexyme assets could not be confirmed without further verification. Expected readout: the primary endpoint is measured at week 40-48, so if dosing began in 2025 results would emerge in late 2026 or early 2027 (Q4 2026 - Q1 2027 estimate). No formal readout guidance has been issued publicly by Lighthouse. Quince has continued to file 10-Qs through Q1 2026, suggesting the corporate entity is still operating but no longer running its own clinical AD program.

The hypothesis: P. gingivalis, the bacterium that causes severe periodontitis (advanced gum disease), travels from inflamed gums to the brain and contributes to Alzheimer's pathology. The proposed CNS (central nervous system) entry route is bacteremia - bacteria leaking from inflamed periodontal pockets into the bloodstream and crossing a compromised blood-brain barrier - with monocyte hitchhiking and cranial-nerve tract invasion proposed as alternative or complementary routes [5]. Dominy et al. published the founding paper in Science Advances in 2019, showing P. gingivalis DNA in postmortem AD brains and demonstrating that the bacterium's proteases - called gingipains - degrade tau protein and trigger neurodegeneration in mice [5]. In plain English: gingipains are enzymes that chew up proteins, including tau - one of the two hallmark proteins (alongside amyloid-beta) that accumulate as toxic tangles in Alzheimer's brains - potentially seeding the very aggregates the disease is defined by. Block the gingipains, slow the disease. That was Cortexyme's entire thesis with atuzaginstat (COR388). Lighthouse's specific mechanism for LHP588 could not be confirmed. The audit flags that the database has no gene targets listed and the mechanism class is the generic 'investigational' bucket. If LHP588 is also a gingipain inhibitor, the biology is identical to atuzaginstat and so are the chemical liabilities. If it works through a different mode (broad antimicrobial, gingipain-independent neuroprotection), the supporting evidence is thinner because Dominy's data specifically supported gingipain inhibition. The broader bacterial-cause hypothesis remains a minority view in AD research. Most of the field focuses on amyloid (Leqembi, Kisunla), tau, and neuroinflammation. The infectious angle gained credibility from the 2019 paper and lost most of it again when atuzaginstat failed two years later. LHP588 is essentially the rematch.

NCT06847321 enrolls 300 patients with mild-to-moderate Alzheimer's who have shown progressive cognitive decline in the past year AND test positive for P. gingivalis infection [1]. The biomarker enrichment is the most important design feature. This is the lesson Cortexyme learned the hard way in GAIN: the topline missed its co-primary cognitive and functional endpoints in the full intent-to-treat population, and only a post-hoc subgroup analysis hinted at a possible signal in the P. gingivalis-positive subset - and that came after the FDA had already imposed the April 2021 hepatotoxicity hold and the protocol had been modified mid-trial [2][3]. The failure was therefore both a missed primary endpoint and a safety-driven program unwind, not one or the other. Lighthouse is running the biomarker-positive population prospectively, which is the right call. Primary endpoint: change from baseline in ADAS-Cog11 (Alzheimer's Disease Assessment Scale - Cognitive subscale, 11-item version) at week 40-48. ADAS-Cog11 is the standard cognitive endpoint for AD trials and what Leqembi and Kisunla used variants of in their key studies. Secondary endpoints are not fully detailed in our enrichment of the registry record and likely include functional measures (ADCS-ADL or similar) plus safety; the specifics need verification against the live ClinicalTrials.gov listing. The 40-48 week window is acceptable for a disease-modifying signal but on the shorter end - most modern AD trials run 18-24 months to detect slope changes in cognitive decline. Comparator is not explicitly stated in our enrichment but Phase 2 designs in this space are almost universally placebo-controlled. The bigger execution concern: biomarker-selected AD enrollment is slow because you have to screen many patients to find the P. gingivalis-positive subset. 300 patients for a small private company is ambitious without a partner.

Our model estimates this drug has a 5% chance of eventually being approved. It starts from a historical baseline of about 24% for drugs at this stage in this area, then adjusts based on ten facts about the trial and sponsor. The estimate is helped by larger-than-typical enrollment, but pulled down by heavier-than-usual blinding, a weak sponsor approval record, and limited earlier-phase results. The remaining factors are close to average and have little net effect on the final number.

Efficacy risk is the dominant failure mode. Atuzaginstat failed GAIN on the same hypothesis with a similar trial design [2]. The P. gingivalis-positive subgroup signal in GAIN was a post-hoc finding (meaning the subgroup was identified after the trial had already missed its primary endpoint, not pre-specified before unblinding - a much weaker form of evidence), and post-hoc subgroup signals in failed trials replicate poorly across drug development as a whole. Safety risk is the second-biggest unknown. Atuzaginstat triggered an FDA partial clinical hold for hepatotoxicity (liver damage, detected as elevated liver enzymes on blood tests) in April 2021 [3]. If LHP588 shares the same chemical scaffold or hits a different liver liability, the program stops on safety alone. No public Phase 1 LHP588 liver data is present in our enrichment, and we cannot confirm from public sources whether a Phase 1 has been completed or whether tox data has been disclosed at conferences. Execution risk: biomarker-enriched enrollment is slow because every patient must be screened for active P. gingivalis infection before randomization. Lighthouse is small and private; without a pharma partner, recruitment for 300 patients could stretch the program timeline by years. Unknown: Lighthouse's funding stage and runway are not publicly disclosed - a $30-60M+ trial is a significant capital ask for an unbacked private biotech and whether they have committed financing materially changes the execution risk. IP/patent risk: Quince (formerly Cortexyme) likely holds patents on gingipain inhibitors as a chemical and mechanistic class; whether those patents block or burden Lighthouse's freedom to operate, or require licensing/royalty arrangements, is unresolved from public sources and is a question any acquirer would diligence first. The Quince Therapeutics 10-Q filing activity in our SEC enrichment hints at some ongoing entity relationship that complicates the corporate picture [4]. Commercial risk: even with a positive trial, the addressable population is only P. gingivalis-positive AD patients - Dominy 2019 detected P. gingivalis DNA in a majority of AD brains examined [5], but clinical-grade seroprevalence in living mild-to-moderate AD patients remains poorly characterized and the addressable fraction has not been validated at scale. Payer coverage would require a validated companion diagnostic that does not currently exist at clinical scale. And the comparator is no longer pure standard-of-care: Leqembi and Kisunla now anchor the market.

Worth watching, not worth betting on. The signal that would matter: an interim safety readout showing no hepatotoxicity at 12 or 24 weeks, or any Phase 1 PK/PD data (pharmacokinetics/pharmacodynamics - how the drug is absorbed, distributed, and whether it reaches and engages its biological target) demonstrating gingipain target engagement in cerebrospinal fluid (CSF - the fluid surrounding the brain and spinal cord, the standard sampling site for proving a drug actually reaches the CNS). Without one of those, this is a small private company chasing a hypothesis that already failed once in a well-powered Phase 2/3 trial. Check back in Q4 2026 / Q1 2027 for the primary readout per the 40-48 week endpoint window. If Lighthouse files an S-1, raises a Series B with a name-brand biotech lead, or signs a development deal with a large pharma, that is the real signal of internal confidence - none of those events are public as of today. The Quince Therapeutics 10-Q cadence is worth tracking on the side; Quince still likely owns IP from the original Cortexyme program and may have economic exposure to LHP588's outcome through licensing, royalty, or residual ownership. Commercial framing: this is a binary bet on a contested hypothesis. Positive outcome - Lighthouse becomes acquisition bait at meaningful multiples, because any disease-modifying AD therapy commands big premiums (Leqembi and Kisunla pricing prove the willingness-to-pay). Negative outcome - the bacterial-Alzheimer's hypothesis is effectively retired and the field stays focused on amyloid, tau, and neuroinflammation. Most likely outcome: another negative readout, modest scientific learning.