Sequencing the exomes of a million people and looking for those who, by genetic accident, appear protected: that is the strategy that brought FNIP1 to light. The rare carriers of a variant that truncates this gene show a lowered triglyceride/HDL ratio, a more favourable body composition and markedly lower odds of cardiometabolic disease. Loss of function, here, is not a disease but an advantage — and a potential therapeutic target, not yet a drug.

Source: nature.com

In plain terms

Our cells have a brake that limits energy expenditure and favours fat storage. The FNIP1 gene is part of that brake. A team screened the exomes of more than a million people and identified a handful of individuals (about one in seven thousand) in whom this gene is naturally "broken" on one copy. The result: a healthier lipid panel, less deep abdominal fat, a greater share of lean mass, a lower average blood-sugar level — and about 60% fewer cardiometabolic diseases among carriers, a figure expressed in odds (detailed below) and carrying a wide margin of uncertainty.

What follows is more nuanced than the summary suggests. In mice, switching off FNIP1 alone in the liver changes nothing: a cousin gene, FNIP2, takes over. Both must be cut, or their partner FLCN, to recover protection against diet-induced obesity. This redundancy exists in mice and apparently not in humans — the authors themselves explain it as a species difference. It is a serious lead for future treatments, but the safety caveat is twofold. Cutting FNIP1 on both sides causes, in other patients, an absence of B lymphocytes and a disease of the heart muscle. And cutting FLCN, the partner whose inactivation is sufficient on its own in mice, exposes one in turn to pulmonary cysts and a marked excess risk of kidney cancer. The whole challenge will be to release the brake, in the right place, without breaking something else.

Discovery

Technical explanation

1. The chosen biomarker is no accident — TG:HDL as a window onto energy state. Rather than screening a hundred phenotypes, the team starts from a single composite indicator, the ratio $\mathrm{TG{:}HDL} = [\text{triglycerides}]/[\text{HDL-cholesterol}]$TG:HDL=[triglycerides]/[HDL-cholesterol]. A high ratio accompanies fat accumulation, insulin resistance and the risk of type 2 diabetes, myocardial infarction and liver damage. The authors first verify this correspondence epidemiologically: that is what legitimises using its genetics as a probe of the energy balance. Method and proof: correlating the biomarker with clinical outcomes before the genetic analysis demonstrates that what is about to be mapped really does relate to a genuine risk, not to an isolated laboratory number.

2. Two scales of variation, two tools. The common variant (frequent, weak effect) and the rare variant (sometimes highly penetrant) are not detected in the same way. The multi-ancestry GWAS captures the first — 1,617 independent signals at 801 loci; the exome-wide association study at the gene level aggregates rare coding variants to reveal the second. Aggregating at the gene level, rather than variant by variant, is what provides the statistical power needed when each allele is too rare to stand out alone. This screen brings up 59 genes enriched in energy regulators expressed in the liver and adipose tissue, 23 of which — nearly two in five — encode targets of drugs already approved or in clinical development: a signal of biological plausibility. The authors specify that these gene–TG:HDL associations are robust and consistent across a series of sensitivity and subgroup analyses (adjustments for common variants, for ancestry, alternative phenotype definitions, subgroups by sex, ancestry and fasting state).

3. FNIP1: a brake downstream of AMPK. FNIP1 encodes a protein that associates with folliculin (FLCN). The FNIP1–FLCN complex represses the transcription factors TFEB and TFE3, conductors of mitochondrial and lysosomal biogenesis, and therefore of energy expenditure. When the cell runs short of energy, the kinase AMPK phosphorylates conserved serines of FNIP1, which deactivates the FNIP1–FLCN complex and lets TFEB reach the nucleus to switch the combustion machinery back on. Mechanically, a variant that truncates FNIP1 amounts to removing a part from the brake: TFEB and TFE3 are less repressed, energy expenditure and lipid breakdown increase — consistent with the lowered TG:HDL observed in carriers.

1. From the human variant to the models — and the mouse surprise. A human genetic association remains correlational; to test causality, one has to manipulate. This is where the study is more interesting, and more cautious, than its press summary. In cells, the team switches off FNIP1 by siRNA in primary human hepatocytes. In animals, it proceeds by targeted liver ablation, using guide RNAs delivered by an adeno-associated virus of serotype 8 (AAV8-gRNA), in mice on a high-fat and high-fructose diet. The central result is counter-intuitive: inhibition of Fnip1 alone, like that of Fnip2 alone, has no effect on weight gain. Only the double inhibition Fnip1 + Fnip2, or inhibition of their partner Flcn, protects against diet-induced obesity. The authors interpret this absence of phenotype as a species difference: in humans, none of the 155 carriers of truncating FNIP1 variants additionally carried a truncating FNIP2 variant, suggesting that the functional redundancy observed in mice does not operate in the same way. What this proves, exactly: the FLCN–FNIP pathway is causally involved in hepatic energy metabolism; it is not a demonstration that loss of FNIP1 alone produces the effect in an animal model. The nuance is decisive, and it is carried by the authors themselves. They further specify that the benefits observed in adipose tissue appear secondary to changes in hepatic energy metabolism, since their AAV8 approach did not touch the genes outside the liver.

2. What the 60% figure is really worth. The most widely repeated claim deserves to be read in its exact form. These are odds roughly 60% lower of cardiometabolic disease in heterozygous carriers, on a composite endpoint combining coronary artery disease, type 2 diabetes, metabolic steatotic liver disease (MASLD) and cirrhosis: $\mathrm{OR} = 0.39$OR=0.39, 95% confidence interval from 0.22 to 0.69, $p = 0.0011$p=0.0011. Three precautions follow. First, an odds ratio is not a relative risk, and the two coincide only for rare events. Second, the confidence interval is wide: its upper bound, 0.69, corresponds to a reduction of only about 31% — the effect is significant, but its magnitude remains poorly delimited. Third, it is a composite: the figure does not say what share belongs to each of the four outcomes.

Why It Worked

The strength of the study lies in its scale and in the chaining of its evidence. A million exomes provide the power to detect a variant present in about one person in seven thousand — invisible in a conventional cohort. The choice of an energy-state biomarker clinically validated beforehand steers the screen towards the intended biology rather than towards noise. Finally, the loop "human genetics → manipulation in hepatocytes and in mice" turns a correlation into a mechanistic demonstration of the pathway — provided one names precisely what was cut.

There remains the gap between the announcement and the proof, and it is twofold. First, the cardiometabolic effect is an association measured in humans, not an intervention trial: nobody has inhibited FNIP1 in an adult to observe a fall in events. Second, the causality established in models concerns the FLCN–FNIP pathway, not Fnip1 in isolation, whose hepatic ablation has no effect in mice. Presenting the murine model as a direct reproduction of the human phenotype would be inaccurate.

The most serious limitation is one of safety, and it concerns the whole pathway, not only FNIP1. On the FNIP1 side: complete, biallelic deficiency is associated in the clinical literature with an immunodeficiency — absent B lymphocytes with agammaglobulinaemia — and with hypertrophic cardiomyopathy. The picture in heterozygotes is more reassuring but not blank: the authors report that heterozygous FNIP1 variants show no statistically significant association with a composite immunodeficiency phenotype built from electronic health records, while explicitly ranking among the adverse effects to be avoided those observed "to a milder degree" in those same heterozygotes.

On the FLCN side the picture is heavier — and the point deserves emphasis, since it is that gene which the murine ablation cuts to obtain the metabolic protection. Heterozygous FLCN variants cause Birt–Hogg–Dubé syndrome, an autosomal dominant adult-onset disorder combining pulmonary cysts, pneumothorax and tumour predisposition, renal in particular. The authors quantify it in their own data: compared with non-carriers, heterozygous carriers of loss-of-function FLCN variants show an 18-fold higher risk of pneumothorax and a 9-fold higher risk of kidney cancer. The metabolic benefit of the FLCN–FNIP pathway is therefore paid for, in systemic loss of function, with a documented and serious risk profile.

This is precisely why their discussion proposes a workaround: selective inhibition in hepatocytes would make it possible to avoid the undesirable phenotypes of a systemic loss of function of this pathway — those of heterozygous FLCN carriers, those of FNIP1 homozygotes, and to a lesser degree those of FNIP1 heterozygotes. "Target validated by genetics" therefore does not yet mean "drug", and the route of administration is part of the problem rather than a detail.

Causal Chain

mega-scale exome sequencing (>1 M people, three continents) → choice of TG:HDL as a probe of energy state → GWAS (1,617 independent signals, 801 loci) + gene-level association (rare variants) → 59 energy genes of the liver and adipose tissue, 39% of them already targeted by drugs → focus on the 155 carriers of ultra-rare truncating FNIP1 variants → favourable lipid profile and body composition, lowered HbA1c and liver fat, cardiometabolic odds at 0.39 (association) → manipulation in primary human hepatocytes and targeted hepatic ablation in mice → finding that Fnip1 alone has no effect, Fnip2 compensating, and that protection requires Flcn or the double ablation → causal conclusion bearing on the FLCN–FNIP pathway, with an acknowledged species difference → proposal of a liver-selective inhibition to circumvent the risks of a systemic loss of function of the pathway — B-lymphocyte and cardiac on the FNIP1 side, pulmonary and renal on the FLCN side, whose heterozygous carriers show an 18-fold higher risk of pneumothorax and a 9-fold higher risk of kidney cancer → today: preclinical stage, questions of safety and targeting still open.

Anecdote

"Protective" genetics has a lineage. In 2006, Cohen and colleagues showed that rare loss-of-function mutations in PCSK9 durably lowered LDL and coronary risk — a human observation that directly inspired a class of cholesterol-lowering drugs. FNIP1 is being pursued in the same spirit: find the people whom nature has already "treated" with a switched-off gene, then trace back to the target. The method has proved itself; it does not erase the safety step, which remains specific to each gene — and, in the present case, it first runs up against an animal model that refuses to reproduce the human phenotype when only Fnip1 is cut.

Legacy and Current Data

Cardiometabolic diseases remain, collectively, the leading cause of death worldwide — which is what gives weight to any new actionable pathway. The study also illustrates a shift in method: million-scale exome sequencing, coupled with the analysis of rare coding variants, is becoming a systematic instrument for target identification — 23 of the 59 genes recovered (39%) already encoded targets of approved or clinical-stage drugs, and 31 (53%) known targets in the broad sense, which serves as a plausibility control.

What remains to be established is clear: the exact magnitude of the benefit, today bracketed by an interval running from −31% to −78% on the odds; the transposability of a mechanism that the murine model only accounts for at the cost of a double ablation; and above all the safety of pharmacological inhibition in adults with an intact gene, which can only come from dedicated trials.

The researcher's eye — open questions

(interpretation by the writer, not results of the study) Three experiments would settle what comes next. A formal Mendelian randomisation, using the variants as instruments, would clarify whether the fall in odds is indeed causal in humans and not due to a confounder. An inducible adult inhibition model — mice with an initially normal gene, then switched off, and if possible in a line where the redundancy of Fnip2 is lifted — would test the safety of late blockade, closer to a drug, while precisely monitoring the sentinel organs of the pathway: the B-lymphocyte compartment and the heart for FNIP1, the lung and the kidney for FLCN. Finally, a dose-dependence test comparing heterozygosity, partial inhibition and complete inhibition would say whether there is a window in which the metabolic benefit is gained before paying the immune, cardiac, pulmonary and renal cost — a question all the more pertinent since the authors are already proposing to circumvent it by hepatic targeting.

Sources

Primary source consulted in its full open-access text (Results, Discussion and Methods); metadata and DOIs verified during the fact-checking audit, on 7 August 2026.

• Hindy G. et al., "FNIP1 variants are associated with favourable metabolism in 1 million humans", Nature, online publication of 5 August 2026. Peer-reviewed article, open access (CC BY-NC-ND 4.0). DOI: 10.1038/s41586-026-10864-2
• "Switching off the FNIP1 gene protects against metabolic disease", Nature Research Briefing, 5 August 2026. DOI: 10.1038/d41586-026-02391-x
• "Rare Gene Variants Linked to Metabolic Benefits, Cardiometabolic Disease Protection", GenomeWeb, 5 August 2026 — statement by corresponding author L. Lotta on the robustness of the association. Article byline not verifiable at the time of consultation.

Background references

Reminders of mechanism and context (prior literature, distinct from the results of the present study).

• Malik N. et al., "Induction of lysosomal and mitochondrial biogenesis by AMPK phosphorylation of FNIP1", Science, 2023, 380, eabj5559 — the AMPK → FNIP1 → TFEB axis. DOI: 10.1126/science.abj5559
• Saettini F. et al., "Absent B cells, agammaglobulinemia, and hypertrophic cardiomyopathy in folliculin-interacting protein 1 deficiency", Blood, 2021, 137(4), 493–499 — phenotype of biallelic FNIP1 deficiency. DOI: 10.1182/blood.2020006441
• Cohen J. C. et al., "Sequence variations in PCSK9, low LDL, and protection against coronary heart disease", New England Journal of Medicine, 2006, 354, 1264–1272 — the "protective genetics → therapeutic target" precedent. DOI: 10.1056/NEJMoa054013

---

Transparency: the cardiometabolic effect is a human genetic association, expressed as an odds ratio with a wide confidence interval (0.22–0.69), and not an intervention trial. The causality demonstrated in models bears on the FLCN–FNIP pathway taken as a whole: in mice, hepatic ablation of Fnip1 alone has no effect. The feasibility and safety of therapeutic targeting remain open questions.