Primary disorders of polyubiquitination: Dual roles in autoinflammation and immunodeficiency

Polyubiquitination is a posttranslational modification (PTM) that plays an essential role in dynamically regulating most intracellular signaling systems. By controlling proteasomal degradation of substrates, impacting their activity, and enabling the interaction of substrates with other proteins, polyubiquitination regulates substrate stability and intracellular signaling (Ciechanover et al., 1980; Hershko et al., 1980; Wilkinson et al., 1980). Consequently, polyubiquitination is not only omnipresent but also elemental to diverse dynamic cellular processes, including inflammatory and immune signaling pathways. As expected, since the identification of the ubiquitin pathway in the 1980s, and given its essential role in all cells, defects of polyubiquitination underlie various diseases in humans (Rape, 2018). In the last decade, specific defects in the polyubiquitination pathways have been mechanistically linked to an ever-growing number of primary disorders that underlie autoinflammation and/or immunodeficiency (Bousfiha et al., 2022; Tangye et al., 2022). The term immunodeficiency is used for disorders of the immune system that predispose to infectious diseases, and autoinflammation is the term used for disorders of hyperactivation of the immune system in the absence of autoimmunity. Primary disorders of the polyubiquitination pathways are individually rare, but their study at the molecular, cellular, tissue, and whole-organismal levels provides important insights into human biology.

Ubiquitin is a small (8.6 kDa) protein expressed in all eukaryotic cells. Ubiquitin is covalently conjugated to substrates in a tightly regulated, reversible, three-step cascade (Fig. 1) (Komander and Rape, 2012). First, a ubiquitin-activating (E1) enzyme activates ubiquitin by converting ATP and by the formation of a thioester (Schulman and Harper, 2009). Second, the activated ubiquitin is transferred to a cysteine residue in the active site of a ubiquitin-conjugating (E2) carrier enzyme by a transthiolation reaction. Third, the ubiquitin-loaded E2 carrier enzyme and one of the ubiquitin-ligating (E3) enzymes catalyze the ubiquitin transfer to a specific substrate protein (Metzger et al., 2014). During ubiquitination, substrate specificity is largely dictated by the E3 ligase. Depending on this E3 ligase, conjugation of ubiquitin to its substrate involves transthiolation of the E3 ligase as an intermediate step or occurs directly from the E2 enzyme. During this last step of ubiquitin conjugation to its substrate, an isopeptide bond is formed between the carboxyl group of ubiquitin’s C-terminal glycine and the ε-amino group of a substrate’s lysine. Ubiquitin can also be bound to nonlysine residues in a protein or to nonproteinaceous substrates such as polysaccharides and lipids (Otten et al., 2021). Ubiquitin is encoded by four different genes as polypeptides of a single ubiquitin fused to ribosomal subunits or as multiple head-to-tail ubiquitin repeats. The ubiquitin conjugation cascade is organized hierarchically and compasses a diverse interaction repertoire between 2 E1 enzyme-encoding genes, around 40 E2 enzyme-encoding genes, and about 600 E3 ligase-encoding genes in humans.

Ubiquitin can be conjugated to a substrate as a monomer or as a polymeric ubiquitin (polyubiquitin) chain. Polyubiquitin chains are organized by the conjugation of the C-terminal carboxyl group of one ubiquitin molecule to one the seven lysine residues (K6, K11, K27, K29, K33, K48, or K63) or to the N-terminal methionine (M1) of another ubiquitin molecule (Fig. 1) (French et al., 2021; Iwai, 2012; Pickart, 1997). In humans, mono- or polyubiquitination of substrates is modified by around 100 deubiquitinases (DUBs) with distinct specificity toward substrates and/or ubiquitin conjugation types (Clague et al., 2019). These DUBs catalyze the deconjugation of ubiquitin from substrates or the trimming of polyubiquitin chains, and provide an additional layer of regulation. The concerted action of E1, E2, and E3 enzymes and DUBs generates the architecture of mono- and polyubiquitin modifications of substrates (French et al., 2021). A diverse and highly complex mixture of homotypic and heterotypic polyubiquitin decoration is the result. The different types of polyubiquitination are recognized by receptors harboring ubiquitin-binding domains, relaying the specific polyubiquitin linkages into cellular responses (Husnjak and Dikic, 2012). Depending on the linkage type, polyubiquitin chains can exert diverse cellular functions. For example, K48-linked polyubiquitin chains mediate recognition and degradation of substrates by the proteasome, while K63- and M1-linked polyubiquitin chains regulate signaling complexes during immune responses (French et al., 2021; Iwai, 2012; Pickart, 1997). However, given the complexity and cell-type specificity of polyubiquitination, the coordinated consequences of the different polyubiquitination pathways on cellular processes can be perplexing and, at times, difficult to disentangle. The study of human disorders of the polyubiquitin pathways provides opportunities to comprehend apparent paradoxes.

Here, we review the known primary disorders of the polyubiquitination pathways in humans that lead to autoinflammation and/or immunodeficiency. We describe the molecular pathogenesis of these disorders, and their role in inflammation and infection. We focus on data from human genetic diseases and ancillary evidence from relevant model organisms.

In humans, only a few primary disorders of the polyubiquitin pathway involving E1 and E2 enzymes are known to date. This scarcity aligns with the limited number of genes responsible for mediating thousands of ubiquitination events on a plethora of substrates.

Somatic mutations in UBA1, encoded by the X chromosome, have been identified in patients with autoinflammation (Beck et al., 2020). The disorder, known as VEXAS (Vacuoles, E1-enzyme, X-linked, Autoinflammation, Somatic) syndrome, is caused mostly by missense mutations, where methionine-41 is substituted by threonine, leucine, or valine. These mutations disrupt expression of UBA1b, one of the three isoforms of UBA1. VEXAS syndrome predominantly affects males (Echerbault et al., 2024). The absence of UBA1b expression in affected cells causes systemic inflammation, impacting the skin, joints, and blood vessels. The pathophysiology of VEXAS syndrome is primarily driven by clonal expansion of myeloid cells (Gutierrez-Rodrigues et al., 2023). At the molecular level, UBA1b is critical for loading three E2 enzymes (UBE2J1, UBE2J2, and UBE2G2), which are essential components of the endoplasmic reticulum–associated degradation (ERAD) pathway (Magaziner et al., 2024). The ERAD pathway employs chaperone proteins to target misfolded proteins of the endoplasmic reticulum for polyubiquitination and subsequent degradation by the proteasome (Brodsky, 2012; Smith et al., 2011; Vembar and Brodsky, 2008). The UBA1b-mediated ERAD dysfunction in VEXAS syndrome results in increased cellular stress, apoptosis, and, consequently, autoinflammation. Emerging therapeutic options include JAK inhibitors (Bourbon et al., 2021), but cure is limited to hematopoietic stem cell transplantation (HSCT) (Beck et al., 2020).

Among the 40 human E2 enzymes, only UBE2T has been associated with a primary disorder. A compound heterozygous mutation in UBE2T has been reported in a single patient with features of Fanconi anemia, a rare bone marrow failure, and cancer predisposition syndrome (Rickman et al., 2015). UBE2T is the E2 enzyme responsible for transferring ubiquitin to FANCL, the E3 enzyme of the Fanconi anemia core complex. FANCL is essential for the monoubiquitination of FANCD2 and FANCI, which play a critical role in DNA repair. The patient’s mutations, a paternal deletion and a maternal duplication in UBE2T, were initially missed by whole-exome sequencing. In the patient’s dermal fibroblasts, the mutations lead to a complete loss of UBE2T expression and a loss of ubiquitination of FANCD2 and FANCI. Reversion of the maternal allele restored UBE2T expression in EBV-transformed B-cell lines and likely underlies somatic mosaicism in the patient’s hematopoietic compartment.

The number of known primary disorders of the (poly)ubiquitin pathway involving E1 and E2 enzymes is limited. As demonstrated by VEXAS syndrome, however, the number of patients suffering from these disorders can be significant (Beck et al., 2020; Echerbault et al., 2024). Somatic variants in genes in close functional proximity to UBA1 are to be expected. The involvement of E1 and E2 enzymes hierarchically high in the polyubiquitination pathways suggests that variants with functional impact may exert pleiotropic effects across tissues. Given the role of polyubiquitination in regulating inflammation, somatic mutations in specific cell lineages make the genes encoding the remaining E1 and E2 enzymes premier candidates for investigation in patients with unexplained autoinflammatory syndromes. No germline mutations have been yet reported for UBE2J1, UBE2J2, or UBE2G2. Given the severe autoinflammatory phenotype of VEXAS syndrome, these genes are strong candidates to underlie autoinflammatory disorders, provided their functions are not redundant.

M1 polyubiquitin chains are functionally important, nondegradative PTMs that govern inflammatory signaling (Fiil and Gyrd-Hansen, 2021; Iwai, 2012). M1 polyubiquitination is regulated by a dedicated set of enzymes (Fig. 2). M1 polyubiquitin chains are assembled by the E3 ligase linear-ubiquitin assembly complex (LUBAC), which consists of three protein subunits: HOIP, SHARPIN, and HOIL1 (Gerlach et al., 2011; Ikeda et al., 2011; Kirisako et al., 2006; Tokunaga et al., 2011). By promoting autoubiquitination and stabilizing the complex, activity of LUBAC is intrinsically regulated by HOIL1 (Fuseya et al., 2020; Peltzer et al., 2018). M1 polyubiquitin chains are hydrolyzed by the DUBs OTULIN and CYLD (Fig. 2). OTULIN is exclusively active toward M1 polyubiquitin, but CYLD also hydrolyzes K63 polyubiquitin chains (Hrdinka et al., 2016; Keusekotten et al., 2013; Komander et al., 2008; Ritorto et al., 2014). The activity and specificity of OTULIN and CYLD are steered by their phosphorylation (Elliott et al., 2014, 2021; Schaeffer et al., 2014). OTULIN and CYLD hydrolyze polyubiquitin chains but also interact with LUBAC in a mutually exclusive manner (Draber et al., 2015). OTULIN binds HOIP via its N-terminal PIM domain, while CYLD binds HOIP via the adapter molecule SPATA2 (Elliott et al., 2016; Kupka et al., 2016; Schaeffer et al., 2014; Schlicher et al., 2016; Wagner et al., 2016). OTULIN and CYLD are negative regulators of NF-κB signaling, but OTULIN also regulates LUBAC activity by preventing its autoubiquitination (Heger et al., 2018). M1 polyubiquitin chains are recognized by proteins that initiate inflammatory and cell death signaling, such as the NF-κB essential modulator (NEMO) (Fiil and Gyrd-Hansen, 2021; Iwai, 2012; Madiraju et al., 2022). Deficiencies of M1 polyubiquitination underlie a growing number of autoinflammation and/or immunodeficiency syndromes in humans.

The LUBAC components HOIL1, HOIP, and SHARPIN are encoded by RBCK1, RNF31, and SHARPIN, respectively. The first human LUBAC deficiencies were reported soon after the discovery of the M1 polyubiquitination machinery.

Complete autosomal recessive HOIL1 deficiency manifests as an early-onset lethal syndrome with concomitant features of both autoinflammation and immunodeficiency (Boisson et al., 2012; Krenn et al., 2018). Patients present severe systemic inflammation and recurrent pyogenic infections. M1 polyubiquitin is undetectable in the patients’ cells. In the patients, HOIL1 deficiency paradoxically results in an impaired response of dermal fibroblasts but hyperresponsiveness of monocytic cells to inflammatory stimuli (Boisson et al., 2012). A shift in NF-κB signaling from prosurvival to prodeath pathways likely underlies the sensitivity of LUBAC-deficient cells toward TNF-induced cell death. Although the patients’ inability to produce anti-polysaccharide antibodies partially contributes to immunodeficiency (Boisson et al., 2012), the cell type–specific consequence of HOIL1 deficiency offers an alternative, molecular, explanation for the apparently paradoxical clinical phenotype of autoinflammation and immunodeficiency in the patients. Muscular amylopectinosis, a metabolic abnormality seen in other autoinflammation syndromes, underlies the neuromuscular and cardiac sequelae of HOIL1 deficiency that is also observed in other case series (Nilsson et al., 2013; Wang et al., 2013).

Partial autosomal recessive HOIP deficiency has been reported in two unrelated patients (Boisson et al., 2015; Oda et al., 2019). Both patients display autoinflammation and immunodeficiency with pyogenic infections. One patient also manifests amylopectinosis and lymphangiectasia (Oda et al., 2019). At the cellular level in both patients, HOIP expression is impaired. LUBAC is destabilized, and M1 polyubiquitin is undetectable. Like in HOIL1 deficiency, the response to inflammatory stimuli in the patients’ dermal fibroblasts is impaired, while monocytic cells are hyperresponsive (Boisson et al., 2015; Oda et al., 2019). Both patients have low numbers of memory B cells, and their antibody production is impaired. The amino acid residue affected in one of the patients (Boisson et al., 2015) is essential for binding of HOIP to OTULIN and—via SPATA2—to CYLD. Although not biochemically investigated in this patient, an impaired regulation of LUBAC by the lack of binding of these DUBs might partially compensate for the LUBAC deficiency. The Hoip-null phenotype in mice is embryonically lethal (Peltzer et al., 2018). The clinical phenotype of complete HOIP deficiency in humans, if compatible with life, remains elusive.

Autosomal recessive SHARPIN deficiency was only recently described in two unrelated patients (Oda et al., 2024). The disorder manifests as a concomitant combination of autoinflammation and relatively mild immunodeficiency, in conjunction with a glycogen storage defect. The inflammatory features in SHARPIN deficiency are distinct from those seen in HOIP and HOIL1 deficiency. In contrast to SHARPIN-deficient mice, which develop keratinocyte-driven dermatitis (Gerlach et al., 2011; HogenEsch et al., 1993; Kumari et al., 2014; Rickard et al., 2014; Tokunaga et al., 2011), the patients have no dermatological manifestations (Oda et al., 2024). Like in the other LUBAC deficiencies, the response to inflammatory stimuli in the patients’ dermal fibroblasts is impaired, but these cells have an increased propensity for cell death. In contrast to HOIP and HOIL1 deficiency, monocytes of SHARPIN-deficient patients are normally responsive to inflammatory stimuli. The patients’ lymph nodes show a reduction in secondary lymphoid germinal center B-cell development, a finding also seen in SHARPIN-deficient mice. The molecular mechanisms of the species-specific phenotypic consequences of human and mouse SHARPIN deficiency remain unclear.

Combined autoinflammation and immunodeficiency manifest in all three recessive LUBAC deficiencies. Recent studies also indicate a link between M1 polyubiquitination and autoimmunity. A HOIL1 missense allele with reduced HOIL1 ligase activity, leading to enhanced LUBAC-mediated immune signaling, is associated with an increased risk of systemic lupus erythematosus in humans (Fuseya et al., 2024). Moreover, human TNIP1 haploinsufficiency underlies a systemic autoimmune disorder (Medhavy et al., 2024). By binding to M1 polyubiquitin chains, TNIP1 suppresses M1 polyubiquitin–mediated signaling. Population genetics metrics indicate that clinically relevant phenotypes in other autosomal dominant LUBAC deficiencies can be expected. RNF31 (encoding HOIP) has a high probability of being loss-of-function (LOF)–intolerant (Lek et al., 2016). Based on its consensus-based measure of negative selection (Rapaport et al., 2021), RNF31 clusters together with other, dominantly inheriting genetic deficiencies that underlie primary disorders. Broad phenotypic classifiers—including those related to recurrent infections and abnormal circulating immunoglobulin concentrations—in patients with chromosomal deletions affecting RNF31 (Firth et al., 2009) suggest that immunodeficiency may be the hallmark feature of HOIP haploinsufficiency in humans. Future studies will have to evaluate this hypothesis.

OTULIN is encoded by a gene on chromosome 5p. In the last 10 years, a spectrum of three, molecularly distinct, OTULIN deficiencies has been reported in humans.

Autosomal recessive OTULIN deficiency manifests early in life as the potentially lethal OTULIN-related autoinflammation syndrome (ORAS) (Damgaard et al., 2016, 2019; Zhou et al., 2016b; Zinngrebe et al., 2022). Biallelic OTULIN deficiency results in the accumulation of M1 polyubiquitin in all patients’ cells, and a compensatory loss of LUBAC expression in nonmyeloid cells (Damgaard et al., 2016, 2019; Zhou et al., 2016b). In contrast, LUBAC expression is stable in myeloid cells. Because of this cell type–dependent compensation, autoinflammation in ORAS patients is largely driven by a defective downregulation of NF-κB–dependent signaling in myeloid cells (Damgaard et al., 2016; Zhou et al., 2016b). The role of myeloid cells in driving autoinflammation is confirmed in biallelic OTULIN-deficient mice (Damgaard et al., 2016). HSCT significantly ameliorates, but not completely resolves, ORAS symptoms (Damgaard et al., 2019). Possibly, residual hyperinflammation is driven by nonhematopoietic cells. In addition to autoinflammation, some ORAS patients appear to exhibit an increased susceptibility to infections, suggesting the co-occurrence of immunodeficiency (Zhou et al., 2016b; Zinngrebe et al., 2022). All ORAS patients carry amorphic or severely hypomorphic variants. The clinical phenotype in patients carrying biallelic, mildly hypomorphic alleles remains to be established.

OTULIN haploinsufficiency is an autosomal dominant disorder predisposing to necrosis triggered by traumas and/or infections (Arts et al., 2023; Batlle-Masó et al., 2024; Spaan et al., 2022). The infectious trigger is typically Staphylococcus aureus (Spaan et al., 2022). Blood leukocyte subsets are unaffected. LUBAC expression is stable, and NF-κB signaling remains intact. Instead, the disorder underlies susceptibility of nonhematopoietic cells to the staphylococcal virulence factor α-toxin (Spaan et al., 2022). Through crosstalk with CYLD, OTULIN haploinsufficiency causes the accumulation of caveolin-1 modified with K63 polyubiquitin chains. Caveolin-1 accumulation at the plasma membrane of the patients’ dermal fibroblasts, but not leukocytes, enhances the cytotoxicity of α-toxin. Hypersensitivity of nonhematopoietic cells to α-toxin is not seen in OTULIN haploinsufficient mice. The patients’ alleles are severely hypomorphic or amorphic, and the disorder is phenocopied in patients with the 5p chromosomal deletion syndrome (Spaan et al., 2022). The clinical penetrance of OTULIN haploinsufficiency is incomplete. Indeed, some parents of ORAS patients present OTULIN haploinsufficiency. The late-onset immunodeficiency in patients with OTULIN haploinsufficiency may be overshadowed by the early-onset autoinflammation in ORAS.

Negative dominant OTULIN deficiency is an autosomal dominant disorder with partial phenotypic overlaps with ORAS and OTULIN haploinsufficiency (Davidson et al., 2024; Takeda et al., 2024). The patients present autoinflammation but also display mild features of immunodeficiency. The mutations reported occurred de novo in the probands. M1 polyubiquitin accumulates in the patients’ cells, and the cells are—like cells from ORAS patients but in contrast to cells from OTULIN haploinsufficient patients—susceptible to TNF-induced cell death, although NF-κB signaling is minimally affected (Davidson et al., 2024; Takeda et al., 2024). Negative dominance of the alleles is observed in cells, but not in cell-free conditions. Negative dominance promotes autoubiquitination of LUBAC, and dysregulation of LUBAC probably underlies the inflammatory sequelae observed in the patients. The alleles have normal capacity to bind HOIP, and negative dominance is likely exerted at the interface between OTULIN and LUBAC (Davidson et al., 2024). The alleles are catalytically impaired when tested in isolation, raising the possibility that the patients are haploinsufficient for the OTULIN-pool that is not in complex with LUBAC.

Depending on the molecular mechanism of the deficiency, patients suffering from biallelic OTULIN deficiency, OTULIN haploinsufficiency, and negative dominant OTULIN deficiency predominantly present autoinflammation, immunodeficiency, or both. Treatment with TNF or IL-1 inhibitors reduces autoinflammation in ORAS patients (Damgaard et al., 2016, 2019; Zhou et al., 2016b; Zinngrebe et al., 2022). Prophylactic or targeted antibiotics, sometimes in conjunction with immunomodulatory therapies, are indicated for immunodeficiency (Arts et al., 2023; Spaan et al., 2022). Like in the LUBAC deficiencies, the OTULIN deficiencies further demonstrate the cell type–dependent implications of disorders of M1 polyubiquitination. Moreover, the study of the OTULIN deficiencies has revealed unexpected crosstalk with other DUBs and ubiquitin pathways, like CYLD and K63 polyubiquitination. The so-far reported mutations associated with the OTULIN deficiencies all cluster in the catalytic DUB domain of the protein. It will be interesting to study the potential clinical implications of mutations in the noncatalytic PIM domain, responsible for binding of OTULIN to HOIP (Draber et al., 2015; Elliott et al., 2014; Keusekotten et al., 2013; Schaeffer et al., 2014), in natura in humans.

Significant molecular and functional crosstalk exists between the M1 and K63 polyubiquitination pathways (Fig. 2) (Emmerich et al., 2013). Like M1 polyubiquitin chains, K63 polyubiquitinations regulate signaling pathways important in inflammation (Iwai, 2012; Madiraju et al., 2022). Among others, K63 polyubiquitin chains are recognized by sensors that initiate NF-κB–dependent inflammatory and cell death signaling (Fiil and Gyrd-Hansen, 2021; Iwai, 2012; Madiraju et al., 2022). Several E3 enzymes can assemble K63 polyubiquitin chains. Examples relevant in human disease include the multifunctional cell death regulator X-linked inhibitor of apoptosis (XIAP) (Jost and Vucic, 2020) and Casitas B-lineage lymphoma (CBL), which attenuates immune and proliferative signaling (Swaminathan and Tsygankov, 2006). K63 polyubiquitin chains are hydrolyzed by the DUBs CYLD, which also hydrolyzes M1 polyubiquitin chains (Hrdinka et al., 2016; Keusekotten et al., 2013), and A20 (Wertz et al., 2004). A20 not only hydrolyzes K63 polyubiquitin chains but also decorates substrates with degradative K48 polyubiquitin chains. Moreover, A20 serves as a sensor for M1 polyubiquitin chains. K63 polyubiquitin chains themselves serve as a substrate for M1 polyubiquitin assembly by LUBAC (Fig. 2) (Emmerich et al., 2013). The known deficiencies of K63 polyubiquitination underlying autoinflammation and/or immunodeficiency in humans illustrate the complexities in the crosstalk between the M1 and K63 polyubiquitination pathways.

CYLD and A20 (encoded by TNFAIP3) share hydrolytic activity toward K63 polyubiquitin chains, but the clinical phenotypes of their deficiencies in humans are distinct.

Autosomal dominant CYLD deficiency underlies CYLD cutaneous syndrome (Bignell et al., 2000), a spectrum of clinically overlapping, tumorous skin disorders also known as the Brooke–Spiegler syndrome. The patients’ alleles are LOF, and skin tumors arise from somatic loss of heterozygosity (LOH) (Blake and Toro, 2009; Leonard et al., 2001). Penetrance is incomplete, with a poorly understood predominance for penetrance in females. CYLD is a tumor-suppressor gene that governs survival, proliferation, and inflammation by regulating the NF-κB, Wnt, and TGF-β signaling pathways through deubiquitination (Marín-Rubio et al., 2023). CYLD is broadly expressed in all cell types and tissues (Uhlén et al., 2015), but human CYLD deficiency is clinically restricted to susceptibility for the development of skin tumors, and CYLD-deficient patients do not present autoinflammation or immunodeficiency (Rajan et al., 2009). Complete, biallelic Cyld knockout mice are viable, but those with truncating alleles succumb in utero (Massoumi et al., 2006; Trompouki et al., 2009). Population genetics metrics indicate that CYLD is under strong negative selection pressure (Lek et al., 2016; Rapaport et al., 2021), suggesting that somatic LOH in humans is only tolerated in the skin.

Haploinsufficiency of A20 (HA20) is an autosomal dominant immune dysregulation syndrome encompassing Behçet-like features combined with autoimmunity, lymphoproliferation, and mild immunodeficiency (Aeschlimann et al., 2018; Zhou et al., 2016a). The clinical spectrum of HA20 is heterogeneous, and the disease is variably penetrant and expressive. HA20 can manifest not only in childhood with systemic autoinflammation, but also later in life with local disease or immunodeficiency (Karri et al., 2024). Humoral deficiencies include hypogammaglobulinemia. Studies in transgenic mice have failed to attribute any of the specific A20 activities to the phenotypes seen in patients (Priem et al., 2020). Given the K48 and K63 polyubiquitin–modulating activity of A20 and its function as a M1 polyubiquitin sensor, the combined consequences of HA20 on K48, K63, and M1 polyubiquitin signaling likely result in an upregulation of NF-κB signaling and a propensity for cell death in the patients’ cells (Aeschlimann et al., 2018; Zhou et al., 2016a). The patients’ alleles are LOF, but correlations between genotype and the severity of the phenotype remain obscure (Karri et al., 2024). Nonetheless, TNF inhibitors and colchicine are often effective in the management of HA20 (Karri et al., 2024).

CYLD deficiency and HA20 demonstrate the intricate crosstalk between the M1 and K63 polyubiquitin pathways in regulating inflammation in humans. For reasons of simplicity, polyubiquitin chains are typically considered as homotypic, but increasing evidence points to the existence of heterotypic chains with mixed polyubiquitin configurations (Rodriguez Carvajal et al., 2021). For CYLD deficiency and HA20, it remains to be seen whether phenotypic characteristics can be attributed to molecular defects specific to either M1 or K63 polyubiquitination. The consequences of Spata2 deficiency in mice have not been extensively studied in isolation (Griewahn et al., 2023; Wei et al., 2017), but their viability appears normal, and these mice do not display an overt immunological phenotype. The clinical phenotype of human SPATA2 deficiency is unknown. SPATA2 is subject to strong negative selection pressure (Rapaport et al., 2021), and—like CYLD and TNFAIP3—predicted to be intolerant to LOF (Lek et al., 2016). Since CYLD relies on SPATA2 to bind HOIP, human SPATA2 deficiency might underlie autoinflammation and/or immunodeficiency as well.

XIAP and CBL are both E3 enzymes that conjugate K63 polyubiquitin chains. Their deficiencies manifest with different phenotypes in humans.

XIAP deficiency is an X-linked recessive immune dysregulation syndrome. In males, the disorder manifests with a wide spectrum of inflammatory phenotypes including immunodeficiency and autoinflammation, cytopenia, hypogammaglobulinemia, and a propensity to develop hemophagocytic lymphohistiocytosis (Mudde et al., 2021). The age of onset and severity of disease are variable, and some carriers remain asymptomatic throughout life. XIAP affects various signaling pathways. The different functions of XIAP are dictated by its modular structure. XIAP exerts antiapoptotic effects by its capacity to bind and inhibit caspases (Mudde et al., 2021). By binding directly to K63 polyubiquitin chains on NEMO, XIAP activates NF-κB signaling (Gyrd-Hansen et al., 2008). By mediating K63 polyubiquitination of RIPK1 and RIPK2, XIAP prevents TNF-mediated cell death (Lawlor et al., 2015) and promotes NOD2-dependent inflammatory signaling by recruiting LUBAC (Damgaard et al., 2012), respectively. The contributions of the different functions of XIAP to specific disease manifestations remain unknown because the correlation between the genotype and the phenotype is poor (Mudde et al., 2021). The only curative treatment option for XIAP deficiency is HSCT (Mudde et al., 2021).

Specific germline-encoded, monoallelic CBL variants underlie a disorder resembling Noonan syndrome. This syndrome is characterized by developmental defects and a risk of myeloid neoplasms (Martinelli et al., 2010; Niemeyer et al., 2010). The causative CBL alleles, LOF for their ligase activity but retaining substrate binding capacity, also cause juvenile monomyelocytic leukemia (JMML) by means of somatic LOH via uniparental isodisomy (Loh et al., 2009; Stieglitz et al., 2015). About 20% of patients with JMML due to somatic LOH for CBL develop autoinflammation, mostly manifesting with vasculitis (Niemeyer et al., 2010). A recent study shows that autoinflammation in these patients is driven by acquired and retained homozygosity for the causative CBL variants in leukocytes (Bohlen et al., 2024). By conjugating K63 polyubiquitin chains to tyrosine kinases, CBL regulates their degradation and attenuates signal transduction (Swaminathan and Tsygankov, 2006). Peripheral blood monocytes of patients with CBL-LOH secrete large amounts of inflammatory cytokines in an ERK pathway–dependent manner (Bohlen et al., 2024). The constitutive activation status of their cells resembles the hyperinflammatory state seen in monocytes of patients with VEXAS syndrome (Beck et al., 2020). Autoinflammation is absent in patients with JMML who have undergone HSCT, consistent with a causal role of leukocytes (Niemeyer et al., 2010).

Recent studies demonstrate that ubiquitin can be conjugated to nonproteinaceous substrates. One example includes the polyubiquitination, among others with K63 linkage types, of bacterial lipopolysaccharide via the E3 enzyme RNF213 (Otten et al., 2021; Saavedra-Sanchez et al., 2024, Preprint). RNF213 is a known regulator of CYLD and SPATA2 (Bhardwaj et al., 2025). Heterozygous de novo missense variants in RNF213 underlie a spectrum of neurological conditions including early-onset stroke, Leigh syndrome (a progressive neurological disorder), and Moyamoya disease (ischemia originating proximal to the basal ganglia) (Brunet et al., 2024). Patients’ cells display upregulation of inflammatory pathways associated with NF-κB signaling (Liu et al., 2024). One distinct RNF213 variant, segregating in heterozygosity with disease in a large kindred, is associated with hypercytokinemia and urticarial lesions (Louvrier et al., 2022). Further studies are needed to clarify the pathophysiological mechanisms underlying RNF213-related conditions, and to elucidate the link between the E3 enzyme function of RNF213 and its regulation of CYLD/SPATA2 activity. Such studies will provide guidance for the identification of more patients with primary disorders due to variants in RNF213.

Cells continuously produce new proteins, and unwanted, damaged, or misfolded proteins need to be degraded. Degradation occurs primarily via a cytosolic complex with proteolytic activity known as the proteasome. The proteasome is formed by a functional core 20S complex, to which are added other components to form a 26S proteasome, a 30S proteasome, or an immunoproteasome (Fig. 3). The additional components include assembly chaperones and regulatory molecules. Each component of the complex participates in roles determining its composition, its selectivity for substrates, and the mechanisms of its processing of substrates. A central way in which proteasomes recognize targets for processing is by the specific ubiquitin linkages decorated on these targets, such as with K48 polyubiquitin chains. In addition, misfolded or disordered proteins are directly targeted for proteasomal degradation, even in the absence of specific ubiquitin linkages (Makaros et al., 2023; Myers et al., 2018). The expression of proteasomal subunits is cell type– and tissue-dependent, permitting cell type–specific functions as demonstrated by the immunoproteasomes (Fig. 3) (Murata et al., 2018). In considering the role of proteasomes in immunological diseases, the production of antigens is a critical determinant of responses to infection and the development of autoimmunity. To start with, however, primary disorders due to defects of the proteasome machinery manifest with autoinflammation.

Several proteasomal deficiencies in humans have been identified for core functional components and assembly chaperones.

Biallelic LOF variants in core proteasome components underlie diseases known as joint contractures, muscular atrophy, microcytic anemia, and panniculitis-induced lipodystrophy syndrome (Agarwal et al., 2010) and chronic atypical neutrophilic dermatosis, with lipodystrophy and elevated temperature (Liu et al., 2012). The acronyms go some way to describing the clinical spectrum associated with these disorders. Biallelic LOF variants have now been found in most of the functional components of the proteasome (encoded by PSMB8, PSMB4, PSMA3, PSMB9, and PSMB10) (Brehm et al., 2015). The group of severe autoinflammation conditions is commonly referred to as proteasome-associated autoinflammatory syndromes (PRAAS) (Fig. 3). Recent expansions of the genetic etiologies of PRAAS include autosomal recessive PSMA5 and PSMC5 deficiency (Papendorf et al., 2023), and deficiency of the earliest proteasome assembly chaperone PSMG2 (de Jesus et al., 2019). Given the potential for additive LOF variants across multiple proteasome subunits within an individual patient, the observation of any monoallelic LOF allele in a patient presenting severe autoinflammation warrants a thorough genetic investigation to find a second hit. Although the phenotypic spectrum of PRAAS is relatively broad, a clear genotype–phenotype correlation is lacking.

Clinically, autoinflammation also characterized the initial patient with PRAAS due to a heterozygous variant in a proteasomal chaperone encoded by POMP (Fig. 3) (Mégarbané et al., 2002). Subsequent investigations of autosomal dominant POMP deficiency uncovered more clinical complexity for this disorder, including immunodeficiency manifestations. Two individuals carrying heterozygous frameshift variants in POMP, their truncated alleles escaping nonsense-mediated decay and participating in a dominant-negative inhibition of proteasome function (Poli et al., 2018), present overt immunodeficiency with bacterial and viral infections. Their disease is associated with elevated CD4⁺/CD8⁺ T-cell ratios with predominantly naive T cells. Autosomal dominant POMP deficiency, thus, connects autoinflammation and immunodeficiency in PRAAS. Patients carrying monoallelic variants in PSMB9 (Kanazawa et al., 2021) and PSMB10 (van der Made et al., 2024) skew toward immunodeficiency as a more frequent, or even exclusive, manifestation. Patients with de novo heterozygous PSMB10 variants have T-B-NK ± severe combined immunodeficiency (van der Made et al., 2024). Possibly, dominant-negative PSMB10 deficiency prevents substitution with functionally compensating subunits and, thus, has a divergent phenotype from the recessive trait causing PRAAS. This could more severely affect the adaptive immune cells that depend on a functional immunoproteasome, and hence skew patients toward immunodeficiency.

Profiling of patients with PRAAS revealed increases in several inflammatory cytokines, in particular type I interferons, which were recapitulated in cell line models of disease (Brehm et al., 2015; Liu et al., 2012).

One trigger of inflammation by the proteasome is via the unfolded protein response (UPR). The UPR is a cellular stress response, and the sensors driving the UPR monitor ER stress. Indeed, PRAAS patient cells display upregulation of the UPR (Poli et al., 2018). Genetic deletion of the canonical UPR pathways (the IRE1, PERK, and ATF6 pathways), however, did not blunt the type I interferon signaling in cellular models of PRAAS (Davidson et al., 2022). A search through known innate immune sensors that regulate the type I interferon response identified protein kinase R (PKR) (Davidson et al., 2022). PKR was known as a sensor of nucleic acids, but, in PRAAS, it senses the buildup of proteins destined for proteasomal removal. In the patients’ cells, cytosol, specific accumulation of the PKR-interacting cytokine IL-24, activates PKR and the type I interferon response. It appears that the glycosylation status of IL-24 governs its proteasomal degradation and the accumulation in PRAAS (Davidson et al., 2022). IL-24 may, thus, serve as an elegant marker for dysfunctions associated with the proteasome. Based on the molecular characterization of the inflammatory pathways driving disease in PRAAS, tailored therapies have been employed in patients. Treatment with JAK inhibitors, blunting signaling downstream of type I interferons, results in a notable clinical improvement in patients with PRAAS (Boyadzhiev et al., 2019; Kim et al., 2018; Sanchez et al., 2018). Employment of JAK inhibitors often permits tapering or cessation of corticosteroids, and adverse events lead to discontinuation only infrequently. The consequences of proteasome deficiencies, however, may be broader than the excessive production of type I interferons. For patients with immunodeficiency in particular, HSCT may be considered.

Given the central role of proteasomes in cellular protein homeostasis, other disorders have the potential to negatively impact the function of the proteasome.

OTULIN-related autoinflammatory syndrome (ORAS) (Damgaard et al., 2016; Zhou et al., 2016b) shares some phenotypic overlap with PRAAS. As described elsewhere in this review, OTULIN is a linear DUB. The first patients with ORAS were initially considered to have PRAAS, but did not carry LOF variants in the proteasome subunits (Zhou et al., 2016b). The phenotypic overlap between PRAAS and ORAS suggests a partially shared mechanistic basis. Indeed, ORAS patients display a type I interferon signature. Biochemical evidence indicates that proteasome subunits themselves are substrates for OTULIN (Tao et al., 2021). A reduced proteasome activity possibly accounts for the observed type I interferon signature in ORAS cells. Whether accumulation of IL-24 and triggering of PKR are at play remains to be established. Other diseases primarily linked to proteasome deficiency are neurodevelopmental disorders where the genes impacted encode the proteasome subunits themselves, such as PSMB1, PSMC1, PSMC3, and PSMD12 (Cuinat et al., 2023). Unlike their autoinflammation counterparts, most of these neurodevelopmental disorders result from heterozygous de novo LOF alleles. Many patients show a type I interferon signature, but the therapeutic potential of JAK inhibitors appears poor given their limited effect on the neurological manifestations of the type I interferonopathy Aicardi–Goutières syndrome. Thus, effective interventions in this class of neurodevelopmental disorders remain an unmet clinical need (Frémond et al., 2023).

K48, K63, and M1 polyubiquitin chains are relatively well studied, but the role of other linkage types remains poorly understood. Because of this relative lack of knowledge, K6, K11, K27, K29, and K33 polyubiquitin chains are termed noncanonical linkages. K6, K11, and K29 polyubiquitin chains are linked to protein degradation pathways, sometimes in connection with K48 polyubiquitination (Tracz and Bialek, 2021). K6 polyubiquitin chains play a role in mitophagy (Akutsu et al., 2011), and K11 polyubiquitination is often associated with K48 polyubiquitination to modulate proteasomal degradation (Tracz and Bialek, 2021). K29 polyubiquitination is important for quality control and degradation of partially or incorrectly folded protein complexes (Tracz and Bialek, 2021). The remaining (K27 and K33) polyubiquitination linkages appear to be involved in more complex processes with differential effects depending on the studies and the E3 ligases involved (Tracz and Bialek, 2021), but both are involved in immunity and inflammation. Primary disorders have been reported in three E3 enzymes catalyzing K27 or K33 polyubiquitination.

Following in vitro experiments and studies in animal models, nascent evidence indicates that K27 polyubiquitin chains regulate innate antiviral immunity in humans.

The autocrine motility factor receptor (AMFR) is a transmembrane E3 enzyme that plays a role in regulating multiple biological pathways in humans. AMFR mediates the polyubiquitination of lysine and cysteine residues on substrates for proteasomal degradation (Fang et al., 2001; Joshi et al., 2017; Liang et al., 2003; Liu et al., 2014; Pabarcus et al., 2009; Shimizu et al., 1999; Wang et al., 2017). In complex with other proteins, AMFR also participates in the ERAD pathway (Jo et al., 2011, 2013; Song et al., 2005). In addition, and together with the insulin-induced gene 1 (INSIG1), AMFR forms an E3 enzyme complex that decorates stimulator of interferon response CGAMP interactor 1 (STING) with K27 polyubiquitin chains. This polyubiquitination of STING serves as an anchoring platform TANK-binding kinase 1 (TBK1) and facilitates its translocation to the perinuclear microsomes. Indeed, depletion of AMFR impairs STING-mediated antiviral transcriptional responses (Wang et al., 2014). Moreover, AMFR directly interacts with TAK1-binding protein 3 (TAB3), a molecule involved in the canonical NF-κB and MAPK pathways. By decorating TAB3 with K27 polyubiquitination chains, AMFR promotes TAK1 activation (Sun et al., 2023).

Autosomal recessive AMFR deficiency has been reported in several patients with hereditary spastic paraplegias, a rare, inherited neurodegenerative or neurodevelopmental disorder (Deng et al., 2023). The clinical phenotype appears to be mostly related to dysfunctions of the ERAD pathway and of lipid homeostasis, and no infection or inflammatory phenotype has been reported in this series. The absence of an inflammatory or infectious phenotype differs from the report of one pediatric patient, presenting a severe varicella zoster infection with subsequent life-threatening hyperinflammation, who carried a heterozygous LOF variant in AMFR (Thomsen et al., 2024). Molecular studies indicate that the patient’s allele most likely acts by negative dominance, impairing cGAS-STING–mediated signaling and compromising the control of viral replication. The study suggests a direct role of K27 polyubiquitination in controlling viral infections in humans, but further research is needed to fully establish its significance in the context of inflammation and viral immunity.

Studies in humans and mice suggest that defects of K33 polyubiquitination of the T-cell receptor (TCR) underlie autoimmunity.

Casitas B lymphoma-b (CBLB) encodes a protein containing a tyrosine kinase–binding domain at its N terminus, a RING domain and a proline-rich domain at the center, and a leucine zipper domain at its C terminus. CBLB acts as a negative regulator of signal transduction pathways for the TCR, B-cell receptor (BCR), and high-affinity immunoglobulin epsilon receptor (FCER1) (Tang et al., 2019). In naive T cells, CBLB inhibits vav guanine nucleotide exchange factor 1 (VAV1) activation upon TCR engagement, imposing a requirement for CD28 costimulation to enable proliferation and IL-2 production. In mice and in complex with Itchy E3 Ubiquitin Protein Ligase (Itch), CBLB conjugates K33 polyubiquitin chains on the CD3-ζ chain (Huang et al., 2010). K-33–linked polyubiquitination of the CD3-ζ chain facilitates its interaction with zeta chain of T-cell receptor–associated protein kinase 70 (Zap70), thereby regulating T-cell signaling and governing autoimmunity in mice (Huang et al., 2010).

Recently, three unrelated patients with early-onset autoimmunity were reported to carry biallelic mutations in CBLB (Janssen et al., 2022). Their T cells exhibited hyperproliferation in response to anti-CD3 cross-linking. The ubiquitination status of CD3 in these patients was not assessed, leaving open the question whether K33 polyubiquitination of CD3 in humans is indeed CBLB-dependent. Patients with autosomal recessive ITCH deficiency also present inflammatory conditions, but their inflammatory phenotype is associated with other, multisystem syndromic features (Brittain et al., 2019; Lohr et al., 2010). ITCH conjugates multiple polyubiquitin configurations, including K27-, K29-, K33-, and K63 polyubiquitin chains (Chastagner et al., 2006; Huang et al., 2010; Meijer et al., 2013; Peng et al., 2011; Tao et al., 2009). This versatile activity of ITCH potentially explains the broader clinical phenotype of ITCH deficiency when compared to CBLB deficiency. The current lack of molecular investigations of the TCR chain in the patients prevents a definitive conclusion regarding the role of ITCH and CBLB in K33 polyubiquitination and the stabilization of the CD3-ζ chain in humans.

Over 550 primary disorders, driven by germline or somatic mutations or by autoimmune phenocopies, are currently known to underlie susceptibility to infections, autoimmunity, autoinflammation, allergy, bone marrow failure, and/or malignancy (Bousfiha et al., 2022; Tangye et al., 2022). The human genetic defects reviewed here, underlying deficiencies of polyubiquitination pathways, all manifest with two seemingly contradictory clinical phenotypes: autoinflammation, immunodeficiency, or both. Depending on the gene affected, immunodeficiency and autoinflammation can manifest synchronously or asynchronously in the same patient, or in isolation as a function of the mechanism of the genetic defect. For disorders of the proteasome degradation pathway, phenotypic skewing toward autoinflammation or immunodeficiency is likely driven by the proteasome subunits affected, and the cell types expressing the gene product (Fig. 4). Given the extensive crosstalk between the M1 and K63 polyubiquitin pathways, the molecular mechanisms underlying phenotypic expressivity of autoinflammation versus immunodeficiency in disorders of nondegradative ubiquitination may be more complex and cell type–dependent. As illustrated by the different OTULIN deficiencies, distinct alleles (LOF, negative dominant) and zygosity states (monoallelic, biallelic) can underlie disparate clinical phenotypes (Fig. 4). The LUBAC deficiencies further demonstrate that autoinflammation and immunodeficiency are not mutually exclusive phenotypes. Considering the infectious and noninfectious triggers that induce inflammation, we propose that the nondegradative polyubiquitination pathways govern the inflammatory response threshold, uniting autoinflammation and immunodeficiency as continuous clinical phenotypes (Fig. 4). Although case series will be small given the rarity of these disorders, future studies might reveal whether immunomodulatory regimens are beneficial for the resolution of both autoinflammation and immunodeficiency.

Several enzymes in the polyubiquitination pathways have multifunctional properties, spanning different polyubiquitin linkage types and affecting diverse signaling cascades. The resulting crosstalk between the respective polyubiquitin pathways is only partially understood. The study of human deficiencies of the polyubiquitin pathways also points to an indirect biochemical crosstalk at multiple levels, one example being the apparent interplay between OTULIN and CYLD. In addition to crosstalk by enzymes involved in conjugation and deubiquitination, another layer of intricacy originates in the heterotypic configuration of branched polyubiquitin chains. Homotypic polyubiquitin chains are organized by the serial conjugation of ubiquitin molecules to one specific lysine residue or to the N-terminal methionine. The other lysine residues on ubiquitin, however, can form branches themselves with additional ubiquitin molecules. Moreover, serine, threonine, or cysteine residues can also serve as substrates for ubiquitination. Together, these additional branches enable heterotypic polyubiquitination configurations. These heterotypic configurations increase the complexity by which polyubiquitination regulates cellular physiology (Akizuki et al., 2024; Tracz and Bialek, 2021). The mechanisms governing the diverse polyubiquitination processes remain understudied (Wanka et al., 2024). The complexity is vast, considering that (1) ubiquitin itself contains 7 lysine residues and an N-terminal methionine that can be ubiquitinated, (2) over 60,000 lysine residues across >9,000 proteins serve as potential ubiquitination sites (Rose et al., 2016), and (3) DUBs and other factors can further modify polyubiquitin chain length and branching configurations (Akizuki et al., 2024; Ikeda, 2023; Iwai, 2014; Lutz et al., 2020; Tracz and Bialek, 2021). Our understanding of the contributions of specific defects in this biochemical complexity to inflammation and immunity in humans is only nascent.

About a quarter of all monogenic primary disorders of the polyubiquitination pathways were initially reported in single patients (Bousfiha et al., 2022; Tangye et al., 2022). With more patients described over time for most disorders, the current state of knowledge demonstrates that the penetrance and expressivity of these disorders are, typically, very variable. Moreover, and while autoinflammation and/or immunodeficiency are the global clinical denominators, expansion of the number of patients reported indicates that the clinical manifestations of disease (e.g., the organs affected) may be diverse. The phenotypic spectrum of the various deficiencies, encompassing autoinflammation, immunodeficiency, or both, implies that some patients will only manifest disease when exposed to specific pathogens, or when subjected to certain cellular stressors. The molecular basis of incomplete penetrance, variable expressivity, and clinical heterogeneity is incompletely understood. In addition to the genetic mechanisms proposed above, clinical phenotypes may be influenced by a cell type– or cell clone–specific, somatic commitment to allele-specific expression (Stewart et al., 2025). It will be interesting to learn whether allele-specific expression also accounts for incomplete penetrance of monogenic primary disorders of the polyubiquitination pathways. All currently known genetic defects of the polyubiquitination pathways are monogenic, but lessons learned from PRAAS-related conditions indicate that some may be modified by mutations in other genes or perhaps be even truly digenic. As we begin to understand the mechanisms leading to phenotypic complexity and heterogeneity, we will be able to solve more cases and discover even more candidate genes in polyubiquitin-related pathways resulting in primary disorders.

The advent of next-generation sequencing has brought an increase in the number of primary disorders related to the polyubiquitination pathways (Bousfiha et al., 2022; Tangye et al., 2022). The discovery of somatic variants in UBA1 demonstrates a scope for further expansion beyond germline-encoded disorders. UBA1 being an E1 enzyme, somatic variants of downstream E2 enzymes resulting in autoinflammation and/or immunodeficiency are to be expected. Moreover, monogenic disorders for most E3 enzymes, many of which regulate critical immune pathways, are yet to be discovered. The genes, deficiencies of which are currently known to underlie monogenic disorders of the polyubiquitination pathways, are subject to negative selection pressure (Karczewski et al., 2020; Lek et al., 2016; Rapaport et al., 2021). Given the severity of the clinical phenotypes of these known deficiencies, it is expected that genes underlying to be discovered disorders are subject to negative selection pressure as well. Guidelines for genetic studies in single patients, allowing for the establishment of a causal relationship between the candidate genotype and the clinical phenotype, are particularly useful for fully penetrant traits (Casanova et al., 2014). Given that yet-to-be-discovered monogenic disorders of the polyubiquitination pathways may come from smaller case series with complex phenotypes and incomplete penetrance, functional validations will become increasingly important (Box 1). Because of intrinsic physiological differences between man and mice, animal models can serve as ancillary lines of evidence. The human specificity of some triggers of disease, such as pathogens adapted to the human host, further attests to a patient-centered approach in the discovery of new genetic etiologies.

A.N. Spaan has received funding relevant to this work from the Dutch Research Council (NWO) under the Talent Program (09150172110006), and the European Research Council (ERC) under the European Union’s Horizon Europe program (101161713). The Laboratory of Human Genetics of Infectious Diseases is supported by the Howard Hughes Medical Institute, the Rockefeller University, the St. Giles Foundation, the National Center for Advancing Translational Sciences (NCATS), the National Institutes of Health (NIH) (R21AI159728), NIH Clinical and Translational Science Award (CTSA) program (UL1TR001866), the French National Research Agency (ANR) under the “Investments for the Future” program (ANR-10-IAHU-01), the Integrative Biology of Emerging Infectious Diseases Laboratory of Excellence (ANR-10-LABX-62-IBEID), ANR AIDIRAK (ANR-23-CE15-0011), the Square Foundation, Grandir - Fonds de solidarité pour l’enfance, William E. Ford, General Atlantic’s Chairman and Chief Executive Officer, Gabriel Caillaux, General Atlantic’s Co-President, Managing Director and Head of Business in EMEA, and the General Atlantic Foundation, Institut National de la Santé et de la Recherche Médicale (INSERM), and the University of Paris Cité.

Author contributions: A.N. Spaan: conceptualization, funding acquisition, investigation, methodology, visualization, and writing—original draft, review, and editing. B. Boisson: conceptualization, funding acquisition, and writing—original draft, review, and editing. S.L. Masters: conceptualization and writing—original draft, review, and editing.