Desmoplakin Cardiomyopathy, a Fibrotic and Inflammatory Form of Cardiomyopathy Distinct from Typical Dilated or Arrhythmogenic Right Ventricular Cardiomyopathy

Statistical Analysis

Continuous variables are reported as mean ± SD and were compared with the Student t test or ANOVA for multiple groups with Tukey posthoc analysis. Categorical variables are reported as percentages and were compared with contingency tables and the Fisher exact test. Subgroup comparison tests (Table 2) were calculated using adjusted residuals and Bonferroni correction for multiple comparisons.²⁴ Event free survival was estimated with the Kaplan-Meier method and compared with log-rank. Statistical analysis was performed using SPSS and Graphpad Prism.

Clinical Presentation of DSP Cardiomyopathy is Distinct from PKP2 Cardiomyopathy

The study included 107 patients with DSP mutations and 81 with PKP2 mutations by commercially available genetic testing or obligate carrier status (DSP obligate carrier n=4). Characteristics of the overall cohort are summarized in Table 1. Mutations in both groups were predominantly truncating and predicted to result in allelic loss of function through nonsense mediated RNA decay (DSP n=105, PKP2 n=81). Genetic variants are summarized in Supplemental Table 2&3 and Figure 1. The DSP cohort was 69% female with an average age at first evaluation of 36 ± 16 years (vs. 51% female for PKP2, p=0.01). LV-predominant cardiomyopathy was exclusively present in DSP cardiomyopathy (52 of 103, 51%) as compared to PKP2 cardiomyopathy (0 of 81, 0%). RV predominant cardiomyopathy was present in only 14 (14%) of DSP patients, while it was uniformly present among PKP2 patients with overt cardiomyopathy (32 of 79, 40%, p<0.001). Fewer DSP patients exhibited both normal LV and RV function as compared to PKP2 patients (36% vs. 60%, p=0.002), indicating a significantly greater probability of overt cardiomyopathy (clinical penetrance) for DSP mutation carriers. Curly hair and/or thick skin on the palms or soles (palmoplantar keratoderma) was commonly present in DSP patients (54/98, 55%) but not in PKP2 patients (1/46, 2%). T wave inversion in V1–3 were more common in PKP2 patients (50% vs 8%, p<0.001) and correlated with RV dysfunction (84% vs 27%, p<0.001), Supplemental Table 4. Although T wave inversions were uncommon among DSP patients, T wave inversions in V4–6 were more frequent with LV involvement (36% vs. 3% with normal ventricular function, p=0.006). DSP patients with MRI data were significantly more likely to exhibit LV LGE (23 of 57, 40%) as compared to PKP2 patients (5 of 51, 10%, p<0.001). Episodic chest pain was more commonly reported by DSP patients (21% vs. 4% for PKP2, p=0.001) at an average age 34 ± 19 years. In 15 of these DSP patients, acute myocardial injury (troponin elevations with no obstructive coronary disease by coronary angiography) was present, but no PKP2 patients had detected episodes of overt myocardial injury that presented with chest pain. The revised 2010 ARVC Task Force criteria were insensitive for clinical diagnosis in DSP patients as compared to PKP2 patients – for example, only 34% of DSP cases met task force criteria for definite diagnosis vs. 49% for PKP2 cases (p=0.02), despite the greater overall cardiomyopathy penetrance with DSP (as above).

Left Dominant Arrhythmogenic Cardiomyopathy with Fibrosis is the Primary Manifestation of DSP Cardiomyopathy

A total of 52 DSP patients demonstrated primary left ventricular involvement (79% of clinically affected DSP patients, 51% of the total genotype-positive cohort), as shown in Table 2. Of the LV predominant DSP cases, RV systolic function was normal in 69%, and only 7% with MRIs (2/29) had evidence of RV wall motion abnormalities. Only 3 out of 52 (6%) of the patients in this group met major imaging criteria for an ARVC diagnosis. Precordial T wave inversions in V1–3 were notably insensitive for detecting cardiomyopathy in this group (9 of 50, 18%) whereas T wave inversions in V4–6 were more common (17 of 47, 36%). The overall sensitivity of ARVC task force diagnostic criteria was low in this group (only 42% with a definite ARVC diagnosis). Most patients with available Holter monitors had frequent PVCs (21 of 29, 72%), consistent with an arrhythmogenic substrate.

Of patients with LV-predominant cardiomyopathy, 74% of patients with available MRIs (20/27) exhibited LV LGE (Table 2). LV LGE occurred in 26% (7/27) in the absence of overt LV systolic dysfunction. In a blinded review of 10 cases, the pattern of fibrosis was characteristically subepicardial in the inferior segment in all 10 cases with extension to the mid-myocardium in the septum in some cases (Figure 2, Supplemental Figure 1). Circumferential LGE with a primarily subepicardial distribution was present in 2 of 10 cases. The presence of intramyocardial fat adjacent to fibrosis was present in 4 of the 10 cases. The distribution of intramyocardial fat in or adjacent to regions of subepicardial fibrosis is highly similar to a recent report of histopathologic analyses on autopsy samples from sudden death cases associated with LDAC.¹⁵ None of the 10 cases from our review demonstrated evidence of RV fat or fibrosis, also consistent with the recent autopsy series from LDAC cases.²⁵

A Subset of DSP Patients Exhibits Predominant Right Ventricular Cardiomyopathy, Most Often with Concomitant LV Involvement

A comparison of the LV-predominant DSP patients with the minority of RV-predominant patients (14 of 103, 14%) is shown in Table 2. These patients were diagnosed later in life when compared to the LV predominant group (mean age of diagnosis: 47 ± 15 years vs. 35 ± 15 years, p=0.03). Though the ARVC task force criteria for a definite diagnosis were more often positive in this group, they were still insensitive given the presence of clinically confirmed cardiomyopathy (57% vs. 42% for LV-predominant, p=0.001). ECG criteria for ARVC also demonstrated modest sensitivity for the RV-predominant group (minor criterion positive in 6 of 14, 43%; major criterion positive in 2, 14%). In comparison, PKP2 patients with overt RV cardiomyopathy exhibited precordial T wave inversions in V1–3 in 84% of cases (26 of 31, p<0.001, Supplemental Table 4).

Concomitant LV dysfunction was present in 7 of the 14 RV-predominant DSP patients (50%), and three of these cases had LV LGE when MRIs were performed. The primary MRI imaging was available for review in one of these cases, and LV LGE was primarily subepicardial in the inferior and inferoseptal segments – a similar pattern of LV LGE as with patients with left dominant disease. Only 4 out of the entire DSP cohort (4%) had apparent isolated RV cardiomyopathy, a marked contrast from the PKP2 cohort. In the PKP2 cohort, LV involvement was only present in 5 of 81 cases (6%), and all of these cases occurred in the context of RV-predominant cardiomyopathy (Supplemental Table 4). Thus, apparently isolated RV cardiomyopathy was by far the most common presentation for PKP2 cardiomyopathy (84% of clinically affected PKP2 patients, p<0.001 vs. DSP).

Myocardial Inflammatory Episodes

Acute myocardial injury was observed in 15% of DSP patients (16/107) without a significant difference among the three clinically defined groups (Table 2). MRI data were available for 10 of these patients with 90% demonstrating LGE in the LV. Four of the patients with troponin elevation also underwent cardiac FDG PET which showed evidence of active inflammation in the myocardium (Figure 3, Supplemental Figure 2 and 3). Of these 16 patients, 7/16 (44%) exhibited LV systolic dysfunction and 1/16 (6%) exhibited both RV and LV systolic dysfunction. Of the 7 patients with normal ventricular systolic function and MRIs, 6 had evidence of LV LGE. In one of these cases with acute myocardial injury and normal ventricular systolic function, LV LGE was extensive (Figure 2).

Effect of Gender and Exercise in DSP Cardiomyopathy

A larger proportion of the DSP cohort was female as compared to the PKP2 cohort (69% vs. 51%, p=0.01). Therefore, we performed additional analyses to determine whether the greater proportion of females might be due to an effect of gender on disease burden. As shown in Supplemental Table 5 there was no significant difference between any of the clinical variables compared.

Additionally, since vigorous exercise has been implicated in the pathogenesis of ARVC²⁶, we compared exercise and sports history among 80 DSP patients for whom these data were available. Moderate to intense exercise was reported in 53% (42/80) of patients. There was no difference in exercise participation between normal ventricular function and left or right dominant cardiomyopathy groups (p=0.515). In contrast to expectation, a smaller proportion of probands reported a moderate or intense exercise history (35%, 11/31 vs. 63%, 31/49, p=0.02). Exercise history was not associated with a greater risk for experiencing severe ventricular arrhythmias among DSP patients (p=0.14).

Clinical Spectrum of DSP Cardiomyopathy Among Probands and Non-Probands

To establish the phenotypic spectrum and estimate the penetrance of DSP mutations, we compared DSP cardiomyopathy probands to their genotype-positive family members (Table 3). The non-proband DSP group demonstrated a smaller proportion of patients with LV dysfunction (33% vs 74%, p<0.001), higher left ventricular ejection fraction (53% vs 38%, p<0.001) and were less likely to have T-wave inversions in leads V4–6 (7% vs 39%, p<0.001). Despite the non-proband group demonstrating more individuals with normal LV function than the proband group, a high proportion of non-probands exhibited frequent PVCs (43% with PVC number >500/24h). The proportion of DSP non-probands who presented with chest pain, troponin elevation, and LV LGE was similar to probands. The overall penetrance of evident cardiomyopathy, frequent PVCs, sustained ventricular tachycardia and/or myocardial inflammatory episodes in the non-proband DSP group was 57% by a mean age of 40 ± 16 years. The penetrance of overt cardiomyopathy was significantly lower in non-proband PKP2 patients (28% vs 57%, p=0.003). While the DSP proband group was more likely to experience the combined VT outcome, the composite severe ventricular arrhythmia outcome was still relatively common within the non-proband DSP group (43% vs. 18%, p=0.005). This analysis highlights that 1) DSP patients’ family members carry a substantial risk of adverse events, and 2) the most sensitive tests for clinical diagnosis in DSP mutation carriers are cardiac MRI (with assessment of LV LGE, regardless of LV function by echocardiogram), rhythm monitoring for PVC burden assessment, and troponin level (if acute chest pain).

Truncating DSP Mutations are Homogeneously Distributed While Missense Mutations Cluster in Protein-Binding Domains

Specific mutation types and loci within the DSP gene coding sequence are shown in Figure 1 and listed in Supplemental Table 2. We also collated other reported DSP mutations from the literature to more comprehensively analyze whether mutations cluster in specific regions (Figure 1, Supplemental Table 1). Mutations in this study were primarily truncating mutations (i.e. frameshift, nonsense, or splice site mutations). Truncating mutations in this and other studies occur throughout the gene without evident clustering. In contrast, missense mutations were more likely to involve the plakophilin/plakoglobin and desmin binding domains as compared to relatively common gnomAD variants (70% vs. 39%, odds ratio 3.6 [95% CI, 1.7,7.6], p<0.001).^{23, 27}

Severe Ventricular Arrhythmia Outcome Analysis

The combined outcome of sustained ventricular tachycardia, sudden cardiac death or appropriate ICD therapy occurred in 28% (30/107) of DSP patients (N=19, 4, 7, respectively, for the individual outcomes) and 30% (24/81) of PKP2 patients (p=NS). In univariate analysis, LV EF <55% was strongly associated with events for DSP, whereas RV dysfunction was strongly associated with events for PKP2 (Table 4). An LV EF threshold of <35% failed to identify 14 of 27 (52%) of patients with severe arrhythmias, while an LV EF threshold of <55% failed to identify 4 of 27 (15%). Based on the genotype-specific univariate analysis results, Kaplan-Meier time to event analysis was performed stratified by LV systolic dysfunction for the DSP cohort and by RV systolic dysfunction for the PKP2 cohort, confirming the prognostic importance of these variables in the independent cohorts (Figure 4). Notably, a subset of DSP patients with normal biventricular function also demonstrated the primary outcome with some events occurring before 30 years of age. Only 2 out of 4 (50%) individuals who had events and were classified as normal ventricular function had cardiac MRIs performed, so the presence of fibrosis is unknown in these patients. Frequent PVCs were highly prevalent among both DSP (80%) and PKP2 (91%) patients who experienced severe ventricular arrhythmias (p=NS, Table 4).

Defining Features and Clinical Diagnosis of DSP Cardiomyopathy

Small case series have described an LV-predominant cardiomyopathy due to DSP mutations but the prevalence and variability of this finding has not been previously examined in a large cohort.⁶ Here, we found that DSP cardiomyopathy involves the LV in almost all cases and often without any apparent RV involvement. This finding was in clear contrast to PKP2 cardiomyopathy, which always involved the RV predominantly and most often in isolation. Not surprisingly, the ARVC task force criteria, validated primarily in PKP2-enriched cohorts, performed poorly for detection of clinically-affected DSP probands and non-probands. Even among DSP cardiomyopathy patients with apparent RV-predominant cardiomyopathy, T wave inversions exhibited poor sensitivity. In addition to LV systolic dysfunction, we identified cases in which significant LV fibrosis and/or frequent PVCs pre-existed overt LV systolic dysfunction or LV enlargement, a key distinction from typical DCM.^{34, 35} Taken together, these findings suggest that the primary, and most sensitive, clinical signs of DSP cardiomyopathy are LV systolic dysfunction, LV fibrosis on MRI, and frequent PVCs. The presence of curly hair and/or thick callused skin on the hands and/or soles of the feet (palmoplantar keratoderma) further heightens the likelihood of DSP cardiomyopathy, as recently reported for autosomal dominant DSP mutations.¹⁸ These findings were also common in our cohort (Table 1), though the sensitivity and extent of these findings were not fully characterized in this report.

Recurrent Left Ventricular Inflammation and Fibrosis as Primary Disease Mechanisms in DSP Cardiomyopathy

The presence of acute myocardial injury episodes among DSP mutation carriers has been suggested by prior case reports and case series^13,14,36,37, but the frequency of this finding and association with disease stage have not been clear in the absence of data from large cohorts that include non-probands. Here, we find that episodes of clinically apparent myocardial injury are common among probands (18%) and non-probands (13%). Moreover, we find that acute episodes of myocardial injury occur even in the presence of normal systolic function, suggesting that myocardial injury and fibrosis precedes overt systolic dysfunction in DSP cardiomyopathy (Figure 2). We used strict criteria to define acute myocardial injury (requiring troponin elevation with normal coronary angiography), likely leading to underestimation of the true prevalence of this phenomenon in the progression of the disease. In four of these cases, we found that episodes of injury were associated with cardiac FDG PET scans that showed typical findings for myocardial inflammation in the left ventricle. Consistent with this observation, evidence of inflammatory infiltrates and scarring has been demonstrated previously in limited histological analyses of DSP patient left ventricular myocardium.⁵ Interestingly, in contrast to the DSP group in our study, none of the PKP2 patients in this study had documented episodes of acute myocardial injury, despite reports of myocardial inflammation as a component of typical ARVC.^{32, 38} We speculate that episodic RV myocardial inflammation in ARVC may be less likely to cause episodic chest pain that leads to clinical recognition, as compared to predominant LV inflammation in DSP patients. Modulation of inflammatory signaling pathways may prove to be a novel therapeutic target for desmosomal-mediated cardiomyopathy, as recently demonstrated in a mouse model harboring homozygous mutations in DSG2.³⁹ Our findings here indicate that the window of opportunity for potential inflammatory modulation may be at an early disease stage, preceding development of LV systolic dysfunction.

In addition, 90% of cases of acute myocardial injury with available MRIs and 53% of all clinically affected DSP cases exhibited late gadolinium enhancement consistent with left ventricular fibrosis. A specific pattern of LGE was present, occurring mainly in the left ventricular subepicardium rather than mid-myocardial as is often seen with non-ischemic cardiomyopathy.^40,41 These imaging findings are consistent with a recent histopathologic analysis that have showed a distinct fibrosis pattern with DSP mutations compared to PKP2 mutations.¹⁵ Taken together, these observations implicate acute myocardial inflammatory episodes as a proximal cause of myocardial fibrosis and progressive dysfunction in the presence of DSP mutations. Our findings also strongly suggest that DSP cardiomyopathy should be considered on the differential diagnosis of other inflammatory myocardial processes, namely cardiac sarcoidosis and myocarditis.

DSP Cardiomyopathy Progression is Independent of Gender or Exercise History

We also examined whether penetrance and/or variable expressivity could be related to gender and exercise history. In ARVC cohorts with high proportions of PKP2 mutation carriers, both male gender⁴² and a history of strenuous exercise²⁶ have been linked to disease progression and ventricular arrhythmias. Interestingly, our DSP cohort had a substantially greater proportion of females (69%) than the PKP2 cohort (51%). Yet, comparison between males and females with DSP mutations did not reveal evidence of a greater disease burden among females. Although this finding requires replication in other substantially sized DSP cohorts, it indicates that male gender should not be considered a risk factor for ventricular arrhythmia risk stratification for DSP cardiomyopathy, as it has been for ARVC.^{31, 43, 44} In the present cohort, 53% (42/80) of patients participated in at least moderate exercise without a significant difference in ventricular dysfunction or arrhythmias between exercise groups. Paradoxically, a moderate or vigorous exercise history was less commonly reported by probands, who exhibited a greater disease burden. Although this finding is in contrast to reports for ARVC, it is consistent with a recent study of treadmill exercise in heterozygous DSP truncation mice, which showed evidence of improved inflammatory and other dysregulated cell-signaling markers following sustained exercise training (although a similar level of myocardial dysfunction persisted regardless of exercise group).⁴⁵ Taken together, our findings do not suggest an obvious interaction between exercise intensity and DSP cardiomyopathy severity, and, therefore, indicate further clinical studies should be performed before creating strict guidelines regarding recommended exercise levels for DSP patients

Risk Stratification for Severe Ventricular Arrhythmias Requires a Genotype-Specific Approach for DSP Cardiomyopathy

DSP cardiomyopathy has been associated with a high prevalence of ventricular arrhythmias since early descriptions.^{3,5, 11} However, risk factors specific to DSP cardiomyopathy have not previously been reported. We demonstrate here that risk stratification on the basis of variables studied in DCM or ARVC populations are inadequate for risk estimation in DSP cardiomyopathy. In particular, we found that the standard DCM LVEF threshold of <35% was an insensitive marker for future severe ventricular arrhythmias in DSP cardiomyopathy, with many events occurring in an EF range of 35–55%, and occasionally at an EF >55%. We also found that criteria predictive of events in ARVC, including RV systolic dysfunction, T wave inversions³¹, and male gender⁴⁴ had no basis at all for either clinical recognition or risk stratification in DSP cardiomyopathy.

Our study provides compelling evidence that a reduction in LV systolic function to <55% is associated with increased risk for DSP cardiomyopathy, and this risk is present regardless of RV involvement. This finding is in clear contrast to our PKP2 cohort, for which RV systolic dysfunction was the strongest marker of increased risk. Our data also suggest that frequent PVCs and LV fibrosis may contribute independently to risk prediction in DSP cardiomyopathy, though the lack of Holter and MRI data in a large proportion of the cohort limited the power of this analysis. Importantly, substantial risk may be present even with normal LV systolic function in the case of recurrent inflammatory episodes and/or extensive fibrosis on MRI. This work emphasizes the need to establish multicenter DSP cohorts with longitudinal follow-up to further advance risk prediction in this highly morbid condition.

Spectrum of Mutations and Variable Expressivity in DSP Cardiomyopathy

The location of individual truncating DSP mutations was relatively evenly distributed throughout the DSP coding sequence, and no clear correlation between truncating mutation location and clinical presentation was evident (Figure 1). These findings indicate a similar loss of function consequence of DSP truncating mutations through nonsense mediated RNA decay of mutant transcripts, rather than effects from truncated proteins, consistent with reduced DSP protein levels in skin and myocardium from patients with DSP truncating mutations.^{18, 30} The specific mutation location may contribute to the disease phenotype in the case of missense mutations (e.g. mutations in the desmin versus plakophilin/plakoglobin binding domains), but this study did not include enough missense mutation carriers to analyze this relationship. Supporting this concept, a prior case series of 10 patients with missense mutations vs. 17 patients with truncating mutations found that the former were less likely to have left ventricular dysfunction.²⁸ Also suggesting the possibility of locus-specific effects for missense variants is the evidence of mutation clustering in the plakophilin/plakoglobin-binding and desmin-binding protein domains, as shown in Figure 1.

Despite most DSP cardiomyopathy patients exhibiting a primary LDAC phenotype, an important subset (14%) exhibited RV-predominant cardiomyopathy. It is not clear from our study why this subset of individuals appeared to have a greater proclivity toward RV remodeling. Since most mutations in this study were truncating, with a presumed similar biologic effect, we speculate that other genetic variation or environmental factors influenced the clinical expression. This concept is supported by the fact that different individuals within the same families experienced either LV- or RV-predominant cardiomyopathy.