The genetic and molecular basis of a connexin-linked skin disease

The human 21-member connexin family of gap junction (GJ) channel-forming proteins, classified into α, β, γ, δ, and ε subgroups based on sequence homology, all share a conserved structural topology (Figure 1A–C). All connexins have a cytoplasmic amino-terminal domain (NTD), four helical transmembrane domains (M1–M4), two extracellular loops (E1 and E2), a cytoplasmic loop (CL), and a cytoplasmic carboxy-terminal domain (CTD) (Figure 1B,C). This topology has been confirmed for at least one α-, β-, γ-, and δ-connexin through high-resolution structural analysis of Cx26 (GJB2) [2,3], Cx32 (GJB1) [4] Cx31.3 (GJC3) [5], Cx36 (GJD2) [6], Cx46/Cx50 (GJA3 and GJA8, respectively) [7] and Cx43 (GJA1) [8,9] channels as revealed by x-ray crystallography and/or cryogenic electron microscopy. The CL and CTD vary the most between isoforms, while the NTD, transmembrane domains, and extracellular loops are highly conserved [10,11]. E1 and E2 are involved in GJ docking [12], with additional roles for E2 in determining heterotypic compatibility [12]. The NTD, M1, M2, E1, E2 and the CL contribute to forming the pore-lining region giving rise to a hydrophilic channel [2,13]. Meanwhile, M3 and M4 face the hydrophobic environment of the phospholipid bilayer [2]. Finally, the highly variable CL and CTD contain sites for various post-translational modifications and motifs for protein-protein interactions that regulate the fate of connexins in cells [14].

Connexins canonically form hemichannels (HCs) and GJ channels that directly couple cells with the extracellular milieu or adjoined cells, respectively (Figure 1D,E) [15]. Gap junctional intercellular communication (GJIC) conferred by connexins is a fundamental process that occurs in most cells of the body to maintain normal physiology and health [16–18]. In recent years, the physiological and pathophysiological roles of HCs in autocrine and paracrine signaling have also become apparent [15,17,19,20].

To fulfill these essential forms of intercellular communication, most cell types in the body express 1–3 connexin isoforms [15]. Interestingly, human epidermal keratinocytes differentially express at least eight so-called ‘keratinocyte connexin’ isoforms (namely Cx26, Cx30, Cx30.3, Cx31, Cx32, Cx40, Cx43 and Cx45) (Figure 2) during their short lifespan, making them uniquely enriched in connexins [21]. Granulosa cells are the only other human cell type known to express more connexin isoforms [15]. Keratinocyte connexins establish a vast communicative network through the avascular epidermis and are key regulators of keratinocyte proliferation, differentiation, and migration [22–24].

The crucial roles of these keratinocyte connexins within the epidermis are highlighted by gene ablation/attenuation studies and the association between connexin-encoding gene mutations and skin disease [25,26]. Variants in the genes encoding five of these keratinocyte connexins are clinically linked to at least seven cutaneous disorders with varying severity including some life-shortening conditions [25,26]. However, not all gene variants result in skin phenotypes, as Cx26 gene variants are the most common cause of sensorineural hearing loss that can occur with or without skin manifestations in a variant-dependent manner [26,27]. Similarly, the vast majority of Cx43 variants cause oculodentodigital dysplasia (ODDD), a rare congenital condition impacting various systems that rarely present with skin phenotypes [26,28,29].

Erythrokeratodermia variabilis (EKV) first described by da Costa in 1925 [30] is a rare hereditary skin disorder characterized by well-demarcated lesions consisting of fixed hyperkeratotic plaques and transient migratory erythematous patches that tend to progressively worsen with age [31]. This classical EKV phenotype is sometimes referred to as EKV of Mendes da Costa, while a similar clinical presentation with distinct circinate erythematous lesions is termed EKV Cram-Mevorah [32,33]. Another similar clinical presentation with static erythema is called progressive symmetric erythrokeratoderma (PSEK) [34]. In all cases, EKV and PSEK disease presentation is limited to the skin and occasionally its appendages (hair, nails), which enables clinical distinction between other conditions such as ODDD and keratoderma-hypotrichosis-leukonychia totalis syndrome [35]. Lesions typically present at birth or during infancy [36,37] and are subject to considerable variability even between affected relatives [38]. This variability has been proposed to be due to trauma (i.e. mechanical irritation, UV exposure), stress, genetics, epigenetics and other factors [37].

Clinicians face a significant challenge in correctly differentiating between different keratodermas and erythrokeratodermas, as these disorders share many clinical and histological features. This has led to considerable debate and confusion over appropriate diagnoses and disease nomenclature. Some of this confusion may be due to the clinical ambiguity of the term ‘keratoderma’, which has led to the inaccurate clustering of drastically different clinical forms of keratoderma into a PSEK category [34]. This is especially apparent in many diagnoses of severe keratodermas or erythrokeratodermas linked to non-connexin encoding genes in which the clinical phenotype is clearly different from connexin-linked erythrokeratoderma [39–44]. Indeed, loricrin gene variants cause a range of phenotypic presentations, including PSEK-like keratodermas, thus the field adopted the terminology of ‘loricrin keratoderma’ to distinguish this disorder from other similar presenting diseases with distinct molecular etiologies [45]. Furthermore, since as early as 1991, it has been proposed that EKV and PSEK may represent different clinical manifestations of the same condition [46], since the main clinical difference between EKV and PSEK is that EKV lesions are well-demarcated and migratory, while PSEK lesions are non-migratory and symmetric [34]. It has been argued that the available clinical evidence does not support the existence of a non-migratory PSEK phenotype, rather the migratory erythemas in EKV are sometimes static and increasing in size, resembling the PSEK phenotype [34]. Once evidence emerged that the same GJB4 variant (Cx30.3-G12D) may cause phenotypes consistent with a diagnosis of either EKV or PSEK in the same family, van Steensel and colleagues proposed adopting the new nomenclature of erythrokeratodermia variabilis et progressiva (EKVP) to encompass both the variable and progressive nature of the lesions [34,47]. For simplicity, we will use the term EKVP throughout this paper to collectively refer to these variable clinical manifestations, except in situations where the clinical phenotype demands distinguishing between EKVP and other erythrokeratodermas.

Here, we set out to review the known and putative underlying molecular causes of EKVP associated with connexin gene variants, what is known about the underlying pathogenic mechanisms, the landscape of treatment options, and the pressing questions that need to be addressed.

Clinically, EKVP is heterogenous and has been most often linked to variants in the genes encoding Cx30.3 (GJB4; OMIM #617524) (Table 1, Figure 3A), Cx31 (GJB3; OMIM #133200) (Table 2, Figure 3B), and Cx43 (GJA1; OMIM #617525) (Table 3, Figure 3C) [31]. Rare cases of erythrokeratodermas with similar clinical presentations to EKVP or PSEK have also been linked to variants in the KDSR (OMIM #617526) [39], KRT83 (OMIM #617756) [42], TRPM4 (OMIM #618531) [43], and PERP (OMIM #619209) [41] genes. Variants in the connexin-encoding genes represent an overwhelming majority of the EKVP cases reported in the literature where genetic analyses were performed, of which GJB3 and GJB4 variants are the most common. Notably, there are several families with a history of EKVP over several generations [38,48,56], although direct genetic evidence to match the clinical diagnoses are only available for a limited number of these individuals. Therefore, the numbers presented here only include patients with direct genetic evidence (i.e. sequencing, confirmation via restriction endonucleases) and a confirmed diagnosis of EKVP by one or more clinicians.

In 1998, Richard et al. [56] were the first to identify pathogenic variants in GJB3 that segregated with the EKVP phenotype in three families. However, since they were unable to identify GJB3 variants in eight families with EKVP the authors speculated that EKVP might be heterogenous and proposed functional partners of Cx31 as candidate EKVP-linked genes. This notion was confirmed two years later when a GJB4 gene variant was found to segregate in a family with EKVP for three generations [33]. Over time more variants in GJB4 and GJB3 were identified (summarized in Tables 1 and 2, respectively). Intriguingly, many of these variants impacted orthologous residues between the two isoforms [38]. While no clinical features could be identified to differentiate Cx31- from Cx30.3-linked EKVP, a more severe phenotype generally involving palmoplantar keratoderma (PPK) was associated with Cx31 variants. Furthermore, Cx30.3 variants sometimes cause erythema gyratum repens-like lesions [49]. Cases of EKVP have been reported in which patients carried mutations in both GJB3 and GJB4, though these cases involved a pathogenic variant in one gene along with a benign polymorphism in the other [55]. However, families with EKVP continued to be identified that lacked pathogenic variants in GJB3 and GJB4 [38,55,89,90], pointing to a possible involvement of other genes in the pathogenesis of EKVP and/or to genetic changes in the non-coding regions of GJB3 and GJB4 [91]. In 2015, GJA1 was identified as a third connexin-encoding gene associated with EKVP [75]. However, the pathogenicity of all Cx43 variants identified to date remain to be fully established as all variants were de novo and thus lacked evidence of segregation with disease over several generations [75,76]. Since then, four additional genes (KDSR, KRT83, TRPM4, and PERP), all of which are unrelated to connexins, have been identified in patients diagnosed with a more severe erythrokeratoderma than patients with connexin-linked EKVP [39–43]. This is consistent with the hypothesis that, despite the clinical similarity, the etiology of EKVP and other erythrokeratodermas might be distinct [91]. The heterogenous nature of EKVP could explain the range of phenotypic variability, even within relatives harboring the same genetic variant, pointing to the influence of individual genetic and non-genetic factors on disease presentation and severity [38].

In humans, Cx30.3 is most highly expressed in suprabasal epidermal keratinocytes and found at low abundance in the lungs, cornea, anus, scrotum, and granulosa cells [15]. Interestingly, evidence suggests Cx30.3 may also be expressed in the intercalated discs of hypertrophic and hypertensive hearts, despite the fact Cx30.3 is not known to be normally expressed in healthy hearts [92]. Interestingly, the homozygous Cx30.3-E204A variant may be linked to hypertrophic cardiomyopathy, though other putative causes were not ruled out [92]. To date, there are 14 known single nucleotide substitutions in GJB4, encoding 13 Cx30.3 missense variants, associated with EKVP (Table 1, Figure 3A). These sites of amino acid substitution localize within the predicted NTD, all four transmembrane domains (M1-M4), the CL, and E1 of Cx30.3 (Figure 2A). Interestingly, all these residues, other than P61, are orthologous or proximal to residues that line an intracellular water pocket in Cx26 that is thought to be important for channel gating [93]. Residues orthologous to the mutational sites in Cx30.3, other than P61, are associated with disease in one or two other β-subtype connexins. Three residues in Cx30.3 — G12, C86, and F137 — appear to be mutational hotspots in β-subtype connexins, as mutations to the orthologous residues in Cx31, Cx32, and Cx26 are associated with EKVP, demyelinating disease and hearing loss, respectively. Curiously, two of the mutational hotspots (G12 and F137) and two other variants (Cx30.3-T130M and Cx31-L135V) fall within calmodulin binding sites found in the NTD and CL of connexins [94–96]. Though these have not been directly investigated to date, it is possible these variants all alter the functionality of these regulatory motifs. Nevertheless, the exact roles of these residues remain unclear, though they are evidently important for normal connexin biology.

Patients with Cx30.3 variants generally present with the classical features of EKVP, including fixed and usually symmetric hyperkeratosis and recurrent episodes of transient erythema that changes over the course of hours to days (Table 1). Some patients with Cx30.3 variants presented with distinct erythema-gyratum repens-like lesions [33,38,53], though most cases of Cx30.3-linked EKVP lacked this feature (Table 1). Disease severity is generally mild to moderate and most frequently affects the limbs, trunk and less commonly the face and other body parts (Table 1). A few exceptions of moderate-to-severe disease presentation characterized by a wider distribution of lesions, including generalized hyperkeratosis, have been reported in the case of the Cx30.3-G12D, Cx30.3-T85P, Cx30.3-R98C, Cx30.3-F137L, and Cx30.3-F189Y variants (Table 1) [33,38,52]. Disease presentation can also be highly variable, as in the case of a family with the Cx30.3-T85P variant where striking age-related differences were noted particularly in the erythematous component, which was evident in childhood and largely absent in adulthood [38]. This variability highlights the influence of age, environmental factors and individual genetics/epigenetics on disease presentation. PPK was absent in most patients with Cx30.3 variants, except for two spontaneous cases of EKVP linked to the Cx30.3-V37M and Cx30.3-M190L variants [50,55].

Like Cx30.3, Cx31 is highly expressed in suprabasal epidermal keratinocytes. Cx31 is also found in sebaceous glands, hair follicles, lungs, tissues of the nose, cornea, scrotum, endocrine pancreas, and granulosa cells [15]. The clinical association between select Cx31 gene variants and hearing loss also points to its expression and role in the auditory system [71,97]. At the time of writing, 18 single nucleotide substitutions in GJB3 encoding 16 Cx31 variants (three different DNA changes have been found in patients encoding the same Cx31-F137L variant) have been identified in patients diagnosed with EKVP (Table 2, Figure 3B). The sites of amino acid substitution associated with EKVP occur in all topological domains of Cx31 except for the C-terminal domain (Figure 3B) and are distinct from those linked to hearing loss. Most cases of Cx31-associated EKVP are autosomal dominant, except for three autosomal recessive variants: Cx31-V30I, Cx31-L34P, and Cx31-E100K (Figure 3B). Residues orthologous to all mutational sites in Cx31 are associated with disease in one to three other β-subtype connexins.

Generally, patients with Cx31 variants have more severe EKVP, affecting the same or more widespread areas, than patients with Cx30.3 variants (Table 2). However, disease severity can vary from mild to severe (Table 2). The trunk and limbs are most frequently impacted, with lesions less commonly affecting the face, buttocks, and body folds. Patients generally have both classical components of EKVP, though in some cases the presence or absence of the erythematous component was not described [56]. Two unrelated cases of advanced EKVP caused by the Cx31-F137L variant were amongst the most severe with hystrix-like hyperkeratosis, that reached 2 cm thick and became malodorous in at least one case [69]. Both cases showed partial clearance following retinoid therapy. A third, unrelated, case in a 3-year old patient had similar extent and distribution of lesions that had not reached such an advanced stage [70]. PPK is more prevalent in patients with Cx31 variants than those with Cx30.3 variants, as this feature was noted in patients with 11 of the 16 Cx31 variants known to date. However, the presence or absence of PPK did not correlate with any specific molecular defects observed in variants. Nevertheless, this observation may indicate that Cx31 plays a more important role in the thick skin of the palms and soles than Cx30.3.

In humans, Cx43 is ubiquitously expressed in nearly all organs and cell types [15]. Given this range of tissue expression, it is not surprising that most of the ∼100 known disease-associated Cx43 variants cause multiorgan dysfunction as is the case for ODDD, which may on rare occasions present with disease phenotypes in the skin and its appendages [26,28,29]. Interestingly, select Cx43 variants involving non-ODDD associated amino acid residues are associated with at least five other disorders [26]. Of these variants, four dominant de novo Cx43 variants (Cx43-A44V, Cx43-E227D, Cx43-P283L, and Cx43-T290N) (Figure 3C, Table 3) have been suspected of being causal of EKVP suggesting these variants likely exhibit milder defects than ODDD-linked variants and/or manifest as pathologies only when found in specific cellular contexts [26,75,76]. Analysis of the cryogenic electron microscopy resolved structure of the C-terminal truncated Cx43 GJ channel (PDB ID 7F94) [8] revealed that these EKVP-associated residues occur in M1 (Cx43-A44V), M4 (Cx43-E227D), and the CTD (Cx43-P283L and Cx43-T290N) of the Cx43 polypeptide. As all four variants were de novo, there is no evidence of co-segregation with disease making it challenging to definitively state that these variants are causal of EKVP.

Clinical examination of patients with Cx43-linked EKVP revealed hyperkeratotic plaques and scaling most commonly impacting the limbs and frictional surfaces, with transient erythema or figurate erythema overlaying the hyperkeratotic lesions in most cases [75,76]. Disease onset began at 5–8 months of age in patients carrying the Cx43-A44V or Cx43-E227D variants, with lesions beginning at frictional surfaces and ultimately becoming generalized [75]. A patient carrying the Cx43-P283L variant had relatively mild disease presentation with symmetrical erythrokeratotic plaques on the hands, feet, wrists, and ankles with onset at 10 months of age [76]. Interestingly, an unrelated patient heterozygous for both the Cx43-P283L and Cx43-T290N variants had much more severe disease with earlier onset at 1 month of age. This patient presented with multiple hyperkeratotic plaques on the face, neck, elbows, wrists, hands, feet, limbs, and inguinal region [76]. This limited genotype-phenotype correlation could possibly suggest that the presence of two variant Cx43 alleles results in more severe disease than when a patient harbors only the Cx43-P283L variant, although this remains speculative. PPK and enlarged porcelain-white lunulae were also noted in all patients with the Cx43-E227D or Cx43-A44V variants [75] while PPK was also noted in the patient with Cx43-P283L/T290N [76].

Studies into 5 of the 13 known Cx30.3 variants have begun to elucidate the underlying molecular and cellular defects. The first Cx30.3 variant studied was Cx30.3-F137L, which showed impaired expression and GJ plaque formation when stably expressed in HeLa cells [32]. Functional dye microinjection studies revealed decreased GJIC in cells stably co-expressing wildtype (WT) Cx31 and Cx30.3-F137L, consistent with a trans-dominant disruption of Cx31 GJ formation [32]. On the other hand, the Cx30.3-V37M variant was expressed at low levels and dominantly disrupted WT Cx30.3 protein levels in patient skin biopsies, either through impaired expression or enhanced degradation [50]. Finally, we recently reported that three Cx30.3 variants (Cx30.3-G12D, Cx30.3-T85P, and Cx30.3-F189Y) when expressed without a WT Cx30.3 counterpart, were GJIC incompetent due to a trafficking impairment resulting in their enriched localization within the endoplasmic reticulum [77]. Connexin-deficient rat epidermal keratinocytes expressing these variants were also more permeable to propidium iodide, which could reflect a HC gain-of-function and/or compromised membrane integrity due to cell death [77]. All three variants could be rescued into GJs by co-expressed connexins, though the functional status of intermixed channels was not tested [77]. While there are some subtle differences, all five Cx30.3 variants tested to date caused either partial or complete trafficking impairment, suggesting this may be a common mechanism underlying Cx30.3-linked EKVP. In addition, four variants (Cx30.3-G12D, Cx30.3-T85P, Cx30.3-F137L, and Cx30.3-F189Y) had deleterious effects on HC function and/or cell death, the former being a mechanism frequently associated with connexin-linked skin disorders [25,26,98–100]. Paracrine signaling via the release of ATP and/or other factors could drive inflammatory signaling and subsequent hyperproliferation of keratinocytes [101] thereby underlying the two classical components of EKVP. The Cx30.3-G12D, Cx30.3-T85P, Cx30.3-F137L, and Cx30.3-F189Y variants are also well documented to co-segregate with disease suggesting they are most likely causal of EKVP, while Cx30.3-V37M lacks evidence of segregating with disease and its molecular characterization has been more limited. All other Cx30.3 variants either have limited or no evidence of segregation with disease and/or have yet to be characterized further in vitro. For these reasons, the pathogenicity of Cx30.3-V37M and the other Cx30.3 variants remains ambiguous and awaits the genotyping of more related individuals with EKVP harboring these variants.

The molecular mechanisms associated with nine Cx31 variants have been interrogated in vitro and/or in vivo. All variants studied to date (Cx31-G12R, Cx31-G12D, Cx31-V30I, Cx31-L34P, Cx31-R42P, Cx31-G45E, Cx31-C86S, Cx31-F137L, and Cx31-L209F) exhibit some degree of trafficking impairment likely resulting in a reduction or loss of GJIC [55,60,78–82,84,86]. In the case of Cx31-R42P, this trafficking impairment was also evident in the stratum granulosum in lesional biopsies [78], while a different study showed this variant also formed constitutively active HCs and led to necrotic cell death [84]. Interestingly, Cx30.3-G12D, Cx31-G12R, Cx31-G12D, and Cx32-G12S all displayed trafficking/assembly defects pointing to the importance of G12 in β-subtype connexins [77,78,102,103]. Many, but not all, Cx31 variants induced cell death either downstream of unfolded protein response activation [81,84] or through other unidentified mechanisms [77,86]. It has been speculated that variant-induced cell death may activate a wound-healing response in surviving cells leading to hyperproliferation [78]. In addition, cell death could activate inflammatory signaling that could underlie the erythematous component of EKVP. However, the transient nature of the erythematous lesions could reflect that connexin variants have varying effects on different keratinocyte populations [78]. In addition to in vitro models, one transgenic mouse model heterozygous for Cx31-F137L partially re-capitulated an EKVP-like phenotype characterized by hyperproliferative skin but lacked the erythematous component [83]. A similar EKVP-like phenotype was observed in a mouse model overexpressing Cx31 in the skin, possibly suggesting that excessive Cx31 expression may dysregulate epidermal physiology [82]. Dye transfer assays, that report on the level of GJIC, in embryonic stem cells from transgenic mice and wound healing experiments in the same genetically-modified mice suggested the Cx31-F137L variant may trans-dominantly disrupt Cx43 [83]. The Cx31-G45E variant also disrupted Cx43 leading to its entrapment within the endoplasmic reticulum possibly reflecting a direct interaction between these isoforms [86]. Overall, there is strong evidence for the pathogenicity of five Cx31 variants (Cx31-G12R, Cx31-G12D, Cx31-R42P, Cx31-G45E, and Cx31-C86S), based on their co-segregation with disease and molecular manifestation in vitro. Most notably all five variants induced cell death and impaired GJ plaque formation in cultured cells. Although Cx31-F137L has been identified in only three unrelated patients to date, not only was the disease presentation strikingly similar across the patient cohort but this variant also partially modeled EKVP in transgenic mice, collectively supporting a pathogenic classification for this variant. Two remaining variants, Cx31-L135V and Cx31-T202N, co-segregate with disease, and are therefore more likely to be pathogenic, but the molecular mechanisms associated with these variants remain unknown. All other Cx31 variants lack evidence of co-segregation with disease, and thus cannot be definitively classified as pathogenic until more familial patients are identified, and a detailed genotype characterization is performed.

In EKVP patient skin sections, Cx43 primarily localized to intracellular compartments [75,76]. However, Cx43 occasionally localized to the cell membrane in skin sections from patients with the Cx43-P283L and Cx43-P283L/Cx43-T290N variants [76]. Cx43-A44V and Cx43-E227D tagged with hemagglutinin were detected within intracellular compartments with a sub-population co-localizing with the cis-Golgi marker GM130 when expressed in HeLa cells [75]. Interestingly, the expression of untagged Cx43-A44V and Cx43-E227D in HeLa cells led to the formation of clearly identifiable GJs [85]. In line with this, both variants formed functional GJs with similar conductance and voltage gating as WT Cx43 in Xenopus laevis oocytes [85]. Furthermore, Xenopus oocytes expressing either Cx43-A44V or Cx43-E227D exhibited pathologically augmented HC activity [85], a phenomenon linked to inflammatory diseases and connexin-linked skin disorders [19,99,104]. This HC gain-of-function possibly suggests a role for Cx43 HCs in nail physiology, as prominent white lunulae were observed in patients with the Cx43-A44V and Cx43-E227D variants.

In another study, we found GFP-tagged Cx43-P283L and Cx43-T290N, expressed alone or in various combinations, formed functional GJs in Cx43-ablated REKs and otherwise appeared benign aside from a modest change in turnover kinetics for the Cx43-P283L variant [88]. A similar delay in the Cx43 lifecycle due to phosphorylation of Cx43 on S368 contributes to the dysregulated connexin expression that occurs in psoriasis [105]. Under all tested conditions, no robust defects were detected for the Cx43-T290N variant except when co-expressed with Cx43-P283L, indicating Cx43-T290N may be a benign variant. Overall, augmented Cx43 HC function seems to correlate with moderate-to-severe disease phenotype induced by Cx43 variants, while variants that do not alter HC function may cause less severe disease. Collectively, in vitro data points to the pathogenicity of the Cx43-A44V and Cx43-E227D variants, while the effect of the Cx43-P283L variant remains ambiguous, and Cx43-T290N is unlikely to be pathogenic, based on available evidence. In all cases, establishing the pathogenicity of these variants requires more clinical evidence of co-segregation with disease.

The existing data supports the classification of connexin variants associated with EKVP into one of eight classes (Table 4), as has been done for connexin variants associated with deafness [106]. These classes encompass variants with impaired GJ formation (Class I), altered GJIC (Class II), HC gain-of-function (Class III), cytotoxic effects (Class IV), dominant effects on co-expressed connexins (Class V), altered turnover (Class VI), no known deleterious effects (Class VII), and variants that have not been studied in vitro (Class VIII). Importantly, variants may exhibit multiple defects and can thus be categorized into more than one category. This classification system will be useful in guiding future research into the treatments aimed at alleviating the specific molecular defects that underlie EKVP.

In five of the Cx30.3-associated EKVP case reports the authors discussed various attempts to treat patients including using topical (keratolytics, emollients, corticosteroids, retinoids) and systemic (retinoids) therapies [47,50,51,53,55]. It is important to note that these studies only report anecdotal responses to treatment in limited patient populations as, to the best of our knowledge, no clinical trials have been carried out. For example, topical keratolytics and emollients partially alleviated hyperkeratosis in a patient with the Cx30.3-V37M variant [50]. Furthermore, the retinoids etretinate and acitretin provided satisfactory symptomatic relief in patients harboring the Cx30.3-G12D and Cx30.3-C86S variants, although the extent of disease clearance was not described [47,51]. Encouragingly, acitretin treatment led to a complete clearance of skin disease in a patient with the Cx30.3-M190L variant [55]. It is unclear if ongoing treatment is required to maintain disease clearance in this patient, as significant disease relapse has been reported in at least one case following the discontinuation of retinoid treatment [69]. Despite these promising results in a limited patient cohort these treatments are not always effective as topical (corticosteroids, retinoids) and systemic (retinoids) treatments all failed to achieve long term clearance of disease in a patient homozygous for the Cx30.3-E99K variant [53].

Ten clinical case reports on Cx31-associated EKVP discussed various therapeutic strategies and their efficacy in patients [37,55,59,61,62,68–70,73,74]. Wilgoss et al. [37] reported that retinoid (etretinate or acitretin) treatment nearly cleared all clinical manifestations of EKVP linked to Cx31-R42P, while acitretin and/or etretinate showed partial to complete resolution of hyperkeratosis and mixed relief of erythema in six other studies involving ten patients harboring Cx31-G12D, Cx31-V30I, Cx31-R42P, Cx31-E100K or Cx31-F137L variants [55,59,61,69,73,74]. Responses to treatment in patients with Cx31-R42P [37,62] and Cx31-F137L [61,69] ranged from limited efficacy to near complete resolution. While the exact mechanism of action of retinoids is unknown, one putative mechanism is through the down-regulation of Cx31 by retinoids [107]. While this may indeed be useful in cases of cytotoxic Cx31 variants, caution is advised as at least one Cx30.3 variant is thought to cause EKVP through the dominant disruption of Cx31, and possibly other co-expressed connexins [32]. Thus, it is more likely that retinoids act independent of their effects on Cx31 expression. In another example, various therapies (topical corticosteroids, calcineurin inhibitors, and emollients) had limited efficacy in a 3 year old patient with EKVP [70]. Finally, therapeutic options for patients harboring Cx43 variants are virtually non-existent with evidence limited to reduced hyperkeratosis following acitretin treatment in a single patient harboring the Cx43-A44V variant [75].

Although retinoids represent the most promising treatment option for EKVP, these drugs are teratogens with long biological half-lives (∼120 days for etretinate) that may have serious consequences including negatively affecting bone growth in children [108]. For these reasons, there is a pressing need to identify or develop new therapeutic options for EKVP patients. Learning points may be obtained from advances in other inheritable genetic diseases. As an example, the FDA has recently approved gene therapy approaches to cure transfusion-dependent β-thalassemia [109] and severe sickle cell disease [110]. These exciting developments provide promise that gene therapy approaches may 1 day provide a cure for other rare inherited disorders such as EKVP. For instance, ablation of a cytotoxic connexin variant or correction of a missense connexin variant could theoretically cure EKVP. However, significant financial hurdles and therapeutic delivery approaches will continue to hinder the development of genetic approaches to treat or cure EKVP. Alternatively, repurposing existing FDA-approved drugs may have utility in the treatment of EKVP. For instance, based on the finding that some Cx31 variants exhibit impaired trafficking and defects in GJ formation that could be overcome via temperature changes [82] drugs that improve protein folding (i.e. pharmacological or chemical chaperones) may be efficacious in treating EKVP. This treatment approach may also alleviate ER stress and cell death associated with some Cx31 variants. Indeed, chemical chaperones have successfully restored the trafficking of select disease-linked Cx26 [111] and Cx50 [112] variants in vitro. One important caveat to consider is that the restored trafficking of variants to the plasma membrane may still exhibit impaired GJIC or reveal additional defects such as hyperactive HC function. To that end, EKVP caused by augmented HC function may be amenable to treatment with FDA-approved mefloquine and flufenamic acid that serve to block the HC channel pore [26,113,114]. In addition, a novel antibody capable of blocking hyperactive Cx30 HCs showed promising effects in a mouse model of Clouston syndrome, reducing both epidermal hyperproliferation and sebaceous gland hypertrophy [115]. Overall, the results from various pre-clinical studies are beginning to pave the way for novel therapeutic strategies that may effectively control or cure EKVP.

Tremendous strides have been made over the years in understanding the consequences of EKVP-associated connexin variants both in vitro and, less commonly, in vivo. There is a growing body of evidence that augmented HC function may underpin connexin-linked disorders [19,98,104,114–116]. Hyperactive or leaky HC may be closely associated with eventual keratinocyte death. Future studies utilizing well characterized systems should continue to discern connexin variant-dependent HC dysfunction and HC-independent cell death. In addition, deeper insight is needed to understand the consequences of connexin variant expression on keratinocyte proliferation, differentiation and viability in transgenic mice and/or organotypic epidermis [117,118]. Once established and validated, these models will complement two-dimensional cell cultures as platforms to test the efficacy of future therapeutics that might include chemical chaperones, antibodies, or gene editing strategies.

EKVP is a complex heterogenous disorder in terms of its underlying genetics and pathogenic mechanisms. Significant progress has been made in characterizing the underlying molecular and cellular mechanisms associated with the known pathogenic variants, paving the way for the future development of treatments targeting specific underlying defects.

The authors declare that there are no competing interests associated with the manuscript.

D.W.L. was supported by the Canadian Institutes of Health Research (148630) and Natural Sciences and Engineering Research Council (03869). S.A.L. was supported by an Ontario Graduate Scholarship.

Dale W. Laird: Supervision, Funding acquisition, Project administration, Writing — review and editing. Sergiu A. Lucaciu: Conceptualization, Resources, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing — original draft.

We would like to acknowledge Mengxi Sun who created an internal reference database of all known connexin gene variants linked to disease. The authors also thank Dr. Ruxandra-Maria Bogdan who helped estimate disease severity as presented in Tables 1–3.

CL

cytoplasmic loop

CTD

carboxy-terminal domain

EKVP

erythrokeratodermia variabilis et progressiva

EKV

erythrokeratodermia variabilis

GJ

gap junction

GJIC

gap junctional intercellular communication

HC

hemichannel

ODDD

oculodentodigital dysplasia

PPK

palmoplantar keratoderma

PSEK

progressive symmetric erythrokeratoderma

WT

wildtype