Paroxysmal nocturnal haemoglobinuria (PNH) is an acquired clonal disorder arising from somatic mutations in the X-linked PIGA gene in hematopoietic stem cells, resulting in deficiency of glycosylphosphatidylinositol (GPI)-anchored proteins.1 The absence of complement regulators CD55 (decay-accelerating factor) and CD59 renders PNH erythrocytes and platelets susceptible to uncontrolled complement activation, leading to chronic intravascular haemolysis and a 5- to 10-fold increased risk of life-threatening thrombosis2 often at unusual sites (e.g. hepatic, cerebral or dermal veins).3 The anti-C5 monoclonal antibody eculizumab, and its long-acting successor ravulizumab, have transformed PNH management by blocking terminal complement activation, producing marked reductions in intravascular haemolysis and substantially lowering thrombotic risk.4, 5 However, thromboses persist in a small proportion of treated patients, particularly during episodes of breakthrough haemolysis or periods of suboptimal complement blockade.5 This residual risk underscores the contribution of complementary prothrombotic pathways beyond direct membrane attack complex (MAC) formation. Recent genomic studies have identified MUC4 (transmembrane mucin-4) mutations as a potential novel risk modifier. In a Chinese cohort of patients with classic PNH, Chen et al. reported a significantly higher prevalence of MUC4 exon 2 mutations among individuals with thrombotic events compared with those without thrombosis, with multivariate analysis identifying MUC4 mutation as an independent risk factor.6 Cumulative thrombosis incidence was markedly increased among mutation carriers. This unexpected finding implicates a transmembrane mucin in PNH thrombogenesis and challenges us to reconsider how genetic modifiers shape phenotypic severity. Thrombosis in PNH is not a singular consequence of haemolysis but rather the result of a complex, self-amplifying network involving complement activation, coagulation, inflammation, and endothelial dysfunction. Loss of CD55 and CD59 permits unregulated C3b deposition and MAC formation on blood cells. On platelets, sublytic MAC triggers activation, microparticle release, and surface phosphatidylserine exposure, creating pro-thrombotic particles.3 Concurrently, C5a generation activates neutrophils and monocytes via C5aR1, promoting tissue factor (TF) expression and neutrophil extracellular trap (NET) formation.3 NETs serve as a scaffold for platelet aggregation, while thrombin generated via TF activation can further cleave C5, establishing a complement-coagulation feedback loop.3 Intravascular haemolysis liberates cell-free haemoglobin and haem, which act as danger-associated molecular patterns. Haemoglobin scavenges nitric oxide, causing vasoconstriction and endothelial activation.7 Haem activates endothelial cells via TLR4, prompting Weibel–Palade body exocytosis and release of ultra-large von Willebrand factor multimers.7 Additionally, deficiency of GPI-anchored proteins on endothelial cells may impair surface localization of tissue factor pathway inhibitor (TFPIβ), reducing anticoagulant capacity, while loss of urokinase plasminogen activator receptor (uPAR) on PNH neutrophils impairs fibrinolysis.3 The link between MUC4 and PNH thrombosis derives primarily from Chen et al.’s 2023 study of 83 classic PNH patients.6 Targeted sequencing revealed that 22 of 32 patients with thrombotic events harboured MUC4 exon 2 mutations, compared to 5 of 49 without thrombosis. Multivariate analysis identified MUC4 mutation as an independent risk factor (OR 20.8, 95% CI 5.2–83.1). The cumulative incidence of thrombosis was 78.6% in MUC4-mutant patients versus 16.1% in wild-type patients.6 A subsequent study by Chen et al. in 2026 demonstrated increased deposition of C5b-9 on the surface of MUC4 knockdown cells, suggesting this as the basis for the increased thrombotic tendency.6, 8 These findings derive predominantly from single-centre Chinese cohorts and require independent validation. Importantly, these observations raise the possibility of genetically informed risk stratification. Patients with large clone size > 50%, prior thrombosis, or MUC4 mutation may represent a higher-risk subgroup in whom closer monitoring, optimization of complement inhibition, or consideration of adjunctive thromboprophylaxis could be justified. Prospective studies are required before routine incorporation of MUC4 testing into clinical algorithms. While MUC4 sequencing is not yet standard-of-care, targeted next-generation sequencing panels for PNH could feasibly include MUC4 exon 2. Proximal complement inhibitors offer promise by blocking complement activation upstream of C5, potentially mitigating both intravascular and extravascular haemolysis. Pegcetacoplan demonstrated superior haemolysis control compared with eculizumab in the Phase III PEGASUS trial,9 while iptacopan has shown efficacy as oral monotherapy in both complement-inhibitor–naïve and previously treated patients.10 Whether intensified or combined complement blockade reduces residual thrombotic risk, particularly in genetically defined high-risk subsets, remains to be determined. The identification of MUC4 mutations as a potent thrombosis risk amplifier in PNH provides a compelling genetic basis to assess residual thrombotic risk. While C5 inhibitors have dramatically improved outcomes, the ‘MUC4-complement-thrombosis axis’ highlights that cell-surface microenvironment modifiers can profoundly influence disease severity. Independent validation of MUC4's role across diverse cohorts and functional mechanistic studies are urgent priorities. If confirmed, integrating MUC4 profiling into clinical practice could enable precision thromboprophylaxis, guiding the use of intensified complement inhibition or adjunctive anticoagulation to eliminate thrombosis as a cause of mortality in PNH. Eng Soo Yap, Edwin Wei Sheng Thong and Cheryl Xiu Qi Lim wrote and reviewed the manuscript. The authors have nothing to report. The authors declare no conflicts of interest. The authors have nothing to report. Not applicable for invited letter.
Yap et al. (2026) studied this question.