A novel POPDC1 p.V183F variant leads to mild loss of POPDC1 and POPDC2 at the sarcolemma
We report here a novel variant in BVES (c.547G > T, p.V183F), which has been identified in two unrelated patients. The variant was present in homozygosity in both patients, while the unaffected family members were heterozygous or wild-type. The variant is not expected to cause a mis-splicing (SpliceAI max score 0.02) and is classified in VarSome (https://varsome.com/) as a variant of unknown significance (VUS). However, its identification in two unrelated patients with a similar phenotype suggests that there is sufficient evidence for the variant to be classified as VUS/likely pathogenic.
Patient 1 (PT1) was first investigated at age 17 for asymptomatic hyperCKemia, with values of 3000–3500 UI/L. Muscle MRI of the lower limbs showed bilateral hypotrophy and early fatty changes of the gastrocnemius medialis, which was also hyperintense on T2-STIR sequences (Fig. 1a). Muscle biopsy obtained from the same muscle displayed increased fiber size variability and dystrophic changes (Fig. 1c, d) with normal immunostaining for conventional sarcolemmal proteins (data not shown). Regarding the cardiological features, the patient neither complained of suspicious symptoms (fatigue, dyspnea, dizziness, palpitations) nor showed any structural or functional abnormality at baseline and follow-up visits. No arrhythmias were detected at baseline ECG and 24-h Holter monitoring. The echocardiogram and the cardiac MRI showed a structurally normal heart. After almost four years of follow-up, he only reported a single vasovagal syncope triggered by emotional stress, while his Holter ECG showed a para-physiological sinus bradycardia with normal chronotropic competence during the day; the other clinical and instrumental cardiac features remained unchanged. The patient did not agree to undergo invasive tests such as electrophysiological study and loop recorder implantation, which were proposed.
Patient 2 (PT2) had a clinical onset at age 47 with myalgias and burning pain in the lower limbs. After one year, he underwent renal transplantation for chronic kidney failure, likely due to hypertensive nephropathy. One month after transplantation, he developed proximal lower limb weakness, which rapidly progressed in the following years. His serum creatine kinase (CK) level was above 6000 UI/L, with subsequent fluctuations between 2500 and 9000 UI/L, and based upon a suspicion of an immune-mediated myopathy, he was treated with steroids, together with one infusion of intravenous immunoglobulins and chronic cyclosporine treatment, without benefit. Muscle imaging showed fatty replacement of gluteus minimus, adductor longus, and magnus, the left semimembranosus and gastrocnemius medialis bilaterally, together with relatively widespread abnormalities on T2-STIR images in the thigh muscles, especially in the anterior compartment, and in the gastrocnemii (Fig. 1b). An assay for myositis-specific antibodies turned out to be negative. At age 51, he could climb stairs using a handrail but could not raise from a chair without the use of arms. On physical examination, there was weakness of hip and knee flexion on the left side (Medical Research Council grade 4) and of knee extension bilaterally (grade 3). Electromyography was myopathic and nerve conduction studies were normal. Muscle biopsy from the right vastus lateralis showed, alongside myopathic changes (Fig. 1e, f), increased endomysial and perimysial fibrosis, and several necrotic fibers with myophagias; hypotrophic round fibers often concentrated in some fascicles; several nuclear clumps, and almost type II fiber uniformity on ATPase stainings were present. HLA class I staining was positive only in necrotic and regenerating fibers, and a mild reduction of sarcolemmal staining for caveolin-3 could be appreciated. Regarding the cardiological features, the patient did not complain of palpitations and did not have any syncope. His ECG was within normal range, only showing mild left axis deviation, and he did not show any arrhythmias on 24-h ECG monitoring; his heart rate was normal throughout the recording. The echocardiogram showed mild, non-pathological interventricular septum hypertrophy (13 mm), which could be explained by hypertension, normal biventricular function, and no functional or structural abnormalities. The patient could not complete cardiac MRI because of claustrophobia. He agreed to undergo electrophysiological study and loop recorder implantation, which however has not yet been performed.
The POPDC1 p.V183F mutation affects a residue that is strongly conserved (Phylo IP100 score = 7.844) and present in all three vertebrate POPDC isoforms and in invertebrate POPDC proteins (Fig. 1g). In the model of the Popeye domain of POPDC1, V183 is located in one of the β-strands (β4) and part of the jelly roll fold forming the roof of the cAMP binding Popeye domain (Fig. 1h). V183 faces into the core of the Popeye domain in close proximity to the predicted cAMP binding pocket. Direct contact between cAMP and V183 is not predicted, although, while the substitution preserves the hydrophobic character at this position, it is unclear if there is any impact on cAMP binding due to steric effects (Fig. 1i). The increased steric demand of phenylalanine may have a structural impact through clashes with other side chains in the β-folds (Fig. 1j). However, modelling the V183F mutation using Missense3D [20] did not predict any major structural aberrations to the Popeye domain.
Skeletal muscle biopsy material from both patients and from age- and sex-matched controls (CT1 and CT2) were sectioned and stained for either POPDC1 or POPDC2. Sections were also stained for SGCA to mark the sarcolemma of the fibers and served as a control for changes in the expression of POPDC isoforms (Fig. 2a, b), as previously reported [11, 40]. The expression level of POPDC protein and SGCA was measured, with the median level of each control sample set to one, and the differences between the patients and controls analyzed. An approximate 20% reduction (p < 0.0001) in the median SGCA-normalized POPDC1 staining intensity in the sarcolemma of fibers of PT1 was observed (0.790, 95% CI 0.744, 0.855; n = 167) compared to CT1 (1.000, 95% CI 0.982, 1.012; n = 161). A similar reduction of around 25% (p < 0.0001) was seen in PT2 (0.755, 95% CI 0.731, 0.771; n = 835) compared to CT2 (1.000, 95% CI 0.972, 1.024; n = 681) (Fig. 2c, d). Meanwhile, the SGCA-normalized level of POPDC2 in the sarcolemma was reduced by 35% (p < 0.0001) in PT1 (0.650, 95% CI 0.619, 0.681; n = 167) compared to CT1 (1.000, 95% CI 0.951, 1.033; n = 138), with a slightly milder reduction of 24% (p < 0.0001) between PT2 (0.763, 95% CI 0.715, 0.810; n = 453) and CT2 (1.000, 95% CI 0.938, 1.077; n = 339) (Fig. 2c, d). Mild reductions in the non-normalized expression levels of both isoforms at the sarcolemma were found (except for POPDC2 in PT2), while an increase in the cytoplasmic concentrations of POPDC2 was also observed in both PT1 and PT2 (Additional file 1: Fig. S1a, b, d, e). These changes led to mild reductions in the enrichment of POPDC1 and POPDC2 at the sarcolemma membrane compared to the cytoplasm, which is representative of how effectively the POPDC proteins are localized at the sarcolemma (Additional file 1: Fig. S1c, f). The changes in POPDC1 and POPDC2 expression were highly variable between individual fibers, with many fibers from the patients resembling control fibers with respect to POPDC protein expression while others showed major differences. A subpopulation of fibers from both PT2 and CT2 displayed a large increase in cytoplasmic levels of POPDC2, as can be seen in Fig. 2b. While the cause of this effect is unknown, it may reflect the increased age of PT2 and the matched control compared to PT1. All fibers were included in the analysis. Irregular and variable fiber sizes and morphologies were seen in both patients (Fig. 2a, b). The median fiber cross-sectional areas were lower in both patients, with an 80% drop (p < 0.0001) between CT1 (2968 μm2, 95% CI 2843, 3101; n = 296) and PT1 (628 μm2, 95% CI 586, 680; n = 327) and a 63% reduction (p < 0.0001) between CT2 (3269 μm2, 95% CI 3190, 3380; n = 1020) and PT2 (1215 μm2, 95% CI 1151, 1322; n = 1288) (Additional file 1: Fig. S1g).
A POPDC1 p.Q153X mutation leads to a severe loss of POPDC1 and POPDC2 at the sarcolemma
A recently reported nonsense mutation in BVES (c.457 > T, p.Q153X) is associated with early onset sinus bradycardia and AV-block and high serum CK levels without clinical signs for LGMD [15]. This mutation is predicted to lead to a truncation of POPDC1 within the Popeye domain, removing the cAMP binding domain and cytoplasmic C-terminal tail. A qualitative reduction in POPDC1 expression at the sarcolemma was reported in the affected patient, however no analysis of POPDC2 expression was performed [15]. We have now quantified the changes in expression levels of POPDC1 and POPDC2 in the muscle fibers contained in biopsies from the index patient, along with a matched control (Fig. 3a, b). The median SGCA-normalized POPDC1 intensity at the sarcolemma was around 75% lower (p < 0.0001) in the patient (0.266, 95% CI 0.250, 0.282; n = 65) compared to the matched control (1.000, 95% CI 0.962, 1.026; n = 238). The normalized POPDC2 sarcolemmal level was even further reduced, by 88% (p < 0.0001), between the patient (0.124, 95% CI 0.116, 0.130; n = 70) and the control (1.000, 95% CI 0.970, 1.038; n = 163) (Fig. 3c, d). The absolute changes in POPDC1 and POPDC2 staining intensity at the sarcolemma were similar (Additional file 1: Fig. S2a, d). A small decrease in POPDC1 and a moderate increase in POPDC2 were seen in the cytoplasmic levels (Additional file 1: Fig. S2b, e). This led to a highly consistent reduction in the enrichment of POPDC1 and POPDC2 at the sarcolemma of the patient’s muscle fibers, with minimal variability (Additional file 1: Fig. S2c, f). It was also noted that there was a greater than 2.5-fold increase (p < 0.0001) in the cross-sectional area of muscle fibers between the control (2031 μm2, 95% CI 1968, 2079, n = 401) and patient biopsies (5283 μm2, 95% CI 4707, 5952, n = 135) (Additional file 1: Fig. S2g).
Muscle fibers of Popdc2W188X/W188X mutants show a loss of POPDC1 and POPDC2 at the sarcolemma
A heterozygous POPDC2 (c.563G > A, p.W188X) mutation was previously reported in patients displaying AV-block [37]. No investigations into changes in the sarcolemmal expression of POPDC1 or the POPDC2 in skeletal muscle or heart tissue were performed due to a lack of biopsy material. We have created a homozygous Popdc2W188X/W188X knockin mouse [37]. The gastrocnemius was dissected from mutant and wild-type (WT) mice and sections were stained for POPDC1 (Fig. 4a) or POPDC2 (Fig. 4b) along with SGCA, and the differences in staining compared. A 56% reduction (p < 0.0001) was seen in the median SGCA-normalized POPDC1 intensity at the sarcolemma of each muscle fiber of the mutant (0.440, 95% CI 0.416, 0.467, n = 95) and WT (1.000, 95% CI 0.939, 1.043; n = 164) (Fig. 4c). POPDC2 at the sarcolemma was reduced by 43% (p < 0.0001) in the Popdc2W188X/W188X mutant (0.471, 95% CI 0.523, 0.661; n = 93) and WT (1.000, 0.976, 1.050; n = 143) (Fig. 4d). The absolute change in POPDC1 and POPDC2 at the sarcolemma was similar (Additional file 1: Fig. S3a, d) while only very mild changes in cytoplasmic levels of both POPDC isoforms were seen (Additional file 1: Fig. S3b, e). This resulted in a lowering of the excess of POPDC1 and POPDC2 at the sarcolemma compared to the cytoplasm, with minimal variability (Additional file 1: Fig. S3c, f). No major aberrations in fiber morphology were observed (Fig. 4a, b), although a 10% reduction in the average fiber cross-sectional area (p = 0.023) was seen in the mutant compared to wild type (Additional file 1: Fig. S3g).
Different POPDC mutations have a variable impact on sarcolemmal expression of POPDC1 and POPDC2
While all of the here studied mutations led to a reduction in the sarcolemmal expression level of POPDC1 and POPDC2, the effects were variable (Fig. 5). Comparing the changes between each biopsy and its respective matched controls showed that the reduction in POPDC1 and POPDC2 levels at the sarcolemma in the case of the two patients carrying the POPDC1 p.V183F variant was significantly less than that seen in the patient possessing the POPDC1 p.Q153X mutation (p < 0.0001; Fig. 5). While there was no difference in the effect on POPDC1 across the two V183F patients (p = 0.078; Fig. 5), the reduction in POPDC2 was around 10% greater in PT1 (p = 0.0030; Fig. 5). The effect in the Popdc2W188X/W188X mouse was less severe than in the patient expressing POPDC1 p.Q153X with respect to the loss of POPDC1 (p = 0.016; Fig. 5) and POPDC2 (p < 0.0001; Fig. 5). The effect on POPDC1 expression was however greater in the Popdc2W188X/W188X mouse mutant than in both patients expressing POPDC1 p.V183F as well as for POPDC2 in case of PT2 carrying the POPDC1 p.V183F variant (p < 0.0001; Fig. 5). No difference was seen between the impact on POPDC2 sarcolemmal expression in the case of PT1 carrying the POPDC1 p.V183F mutation and the Popdc2W188X/W188X mouse mutant (p = 0.32; Fig. 5).
The plasma membrane expression and trafficking of POPDC1 and POPDC2 is dependent on each other
The above findings, as well of those from other previously reported patients [11, 19, 40], suggest that a mutation in POPDC1 can alter the subcellular expression pattern of POPDC2, and vice versa, in skeletal muscle fibers. To investigate if the membrane expression of POPDC1 and POPDC2 is indeed dependent on each other, HEK293 cells were transiently transfected with either POPDC1 or POPDC2 possessing C-terminal ECFP and EYFP tags, respectively (Fig. 6a). Individual cells were segmented into cytoplasm and plasma membrane compartments using the lipophilic dye DiD to mark the plasma membrane and Hoechst-33342 to demarcate the nucleus. The extent of the plasma membrane localization of each protein was quantified by determining the relative level of each protein at the plasma membrane compared to the cytoplasm. The median level of localization of each protein across cells when singly expressed, or when co-expressed with the other POPDC isoform, was compared. When singly expressed, POPDC1 (0.565, 95% CI 0.502, 0.713; n = 15) and POPDC2 (0.506, 95% CI 0.403, 0.773, n = 15) were almost half as concentrated in the plasma membrane compared to the cytoplasm (Fig. 6b). However, when co-expressed, POPDC1 (4.461, 95% CI 3.692, 5.536; n = 46) and POPDC2 (5.666, 95% CI 4.254, 6.160, n = 46) were both effectively localized to the plasma membrane at levels significantly above the single expression conditions (p < 0.0001; Fig. 6b). No significant difference in cytoplasmic levels of POPDC1 and POPDC2 between the two groups was seen (Additional file 1: Fig. S4a). However, it was found that POPDC1 and POPDC2 plasma membrane expression when solely expressed was around 30% and 10% of the level observed in the co-expression system, respectively (p < 0.0001; Additional file 1: Fig. S4b).
The effect of a set of clinically identified POPDC mutations on the subcellular expression of POPDC1 and POPDC2 in HEK293 cells was then investigated. The POPDC1 p.V183F and p.Q153X mutations were introduced into the POPDC1-ECFP construct. Additionally, a POPDC1 p.S201F mutant was tested, having previously been reported to cause a significant loss in the plasma membrane expression of POPDC1 and POPDC2 in the muscle fibers of patients [40]. The POPDC2 p.W188X mutation was introduced into the POPDC2-EYFP construct. Each mutant construct was co-transfected with its corresponding wild-type partner and the change in the median plasma membrane localization of each protein, compared to the double wild-type expression, was determined. A significant reduction in the plasma membrane localization of POPDC1 was seen in the presence of POPDC1 p.Q153X (0.967, 95% CI 0.916, 1.243; n = 17), POPDC1 p.S201F (1.520, 95% CI 1.012, 1.735; n = 22), and POPDC2 p.W188X (0.983, 95% CI 0.830, 1.153; n = 24) compared to the wild-type pair (all p < 0.0001; Fig. 6c, d). Likewise, significant drops in the plasma membrane enrichment of POPDC2 compared to the wild-type pair was seen for POPDC1 p.Q153X (1.239, 95% CI 0.973, 1.369; n = 17), p.S201F (1.202, 95% CI 0.911, 2.078; n = 22), and POPDC2 p.W188X (0.682, 95% CI 0.573, 0.872; n = 24) (all p < 0.0001; Fig. 6c, d). The changes in the expression of the constructs in the cytoplasm and plasma membrane relative to the wild-type pair, which caused the observed changes in the level of plasma membrane localization, were as follows. The POPDC1 p.Q153X mutation led to an increased accumulation of POPDC1 (2.960, 95% CI 1.062, 5.464; p = 0.015) and POPDC2 (2.300, 95% CI 1.255, 4.554; p = 0.042) in the cytoplasm, while POPDC2 p.W188X led to an increase in the intracellular localization of the mutant POPDC2 protein (5.817, 95% CI 3.345, 7.374; p < 0.0001). The POPDC1 p.V183F mutant resulted in a mild drop in POPDC2 within the cytoplasm (0.5812, 95% CI 0.428, 0.806; p = 0.029). The cytoplasmic level of both isoforms was unchanged in the case of POPDC1 p.S201F (Additional file 1: Fig. S4c). The concentration of both proteins at the plasma membrane was comparable to wild type in the presence of the POPDC1 p.V183F mutant. POPDC1 p.Q153X led to mild reductions in the POPDC2 construct (0.618, 95% CI 0.270, 0.925; p = 0.35). The POPDC1 p.S201F led to major reduction of the mutant POPDC1 (0.233, 95% CI 0.158, 0.437) and its POPDC2 partner (0.276, 95% CI 0.184, 0.540) (both p < 0.0001). Additionally, POPDC1 when co-expressed with POPDC2 p.W188X was reduced at the plasma membrane (0.344, 95% CI 0.207, 0.656; p < 0.0001) (Additional file 1: Fig. S4d).
Direct interaction of POPDC1 and POPDC2
While POPDC1 has previously been reported to form homodimers [23, 26], the above results led us to search for evidence of a direct interaction between POPDC1 and POPDC2. Both POPDC1 and POPDC2 are prominently expressed in cardiac and skeletal muscle [2, 35] and immunostaining of isolated ventricular cardiac myocytes revealed overlapping expression domains for both isoforms (Fig. 7a). To identify any interactions between POPDC1 and POPDC2 in their native environment, a PLA was carried out using sections from mouse atrium and ventricle from wild-type mice, with ventricular tissue of Popdc1−/−/Popdc2−/− mutants serving as negative control (Fig. 7b). PLA signals were observed in atrial and ventricular sections of wild-type hearts, whereas no signal was present in sections of Popdc1−/−/Popdc2−/− mutants. While PLA signals were observed in cardiac myocytes of both chambers, the subcellular localizations differed between atrial and ventricular myocytes. In atrial myocytes, PLA signals were mostly localized at the sarcolemma, whereas in ventricular myocytes signals were found at the sarcolemma and within the cell boundaries. POPDC1 and POPDC2 have both been shown to reside in the sarcolemma and in the t-tubules of cardiomyocytes [1], with the higher level of t-tubules present in ventricular cardiomyocytes [28] likely contributing to the observed chamber-specific differences. This shows that POPDC1 and POPDC2 form complexes at the sarcolemma in cardiomyocytes. To confirm the interaction of POPDC1 and POPDC2, POPDC2 possessing a C-terminal FLAG-tag was co-expressed with a POPDC1-MYC construct in COS-7 cells. In addition, POPDC2-FLAG was also co-expressed with POPDC3-MYC. Cell lysates were precipitated with a MYC-tag antibody and subjected to Western blot analysis using FLAG-tag antibody. POPDC1 was found to specifically co-precipitate with POPDC2, but this was not the case for POPDC3 (Fig. 7c). Performing the same experiment with POPDC1 carrying a C-terminal HA-tag and POPDC3 with a MYC-tag demonstrated that POPDC3 can be co-precipitated with POPDC1 (Fig. 7d). This suggests that POPDC1 undergoes complex formation with POPDC2 and POPDC3, but no interaction is detectable between POPDC2 and POPDC3. The interaction of POPDC1 and POPDC2 was also further demonstrated to occur in native tissue by co-immunoprecipitation of POPDC1 from mouse heart lysates using a POPDC2 antibody (Fig. 7e). The co-precipitation of POPDC1 did not occur when lysates were used from the hearts of Popdc2 null mutant mice.
When lysates of COS-7 cells expressing POPDC1 and POPDC2 were examined using SDS-PAGE followed by a Western blot, evidence of POPDC1 forming hetero-oligomers with POPDC2 was seen. Both POPDC1 and POPDC2 isoforms have very similar molecular weights (POPDC1: 41.5 kDa, POPDC2: 40.5 kDa). Therefore, POPDC1 was tagged at the C-terminal with CFP (29 kDa), while POPDC2 was fused with FLAG (1 kDa), to enable differentiation of homo- (expected: 140.5 kDa) and heterodimers (expected: 111.5 kDa) based on their molecular weight on the blot. Use of an anti-CFP antibody showed a differing pattern of bands when POPDC1 was expressed alone compared to co-expression with POPDC2. The majority of POPDC1 was in a monomeric state (multiple bands of approx. 70 kDa) in both groups, as would be expected given the presence of SDS. However, weaker bands corresponding to homodimers of POPDC1 (approx. 140 kDa), as well as those matching the expected molecular weight of POPDC1–POPDC2 heterodimers (approx. 110 kDa) and heterotetramers (approx. 220 kDa) were also seen, despite the denaturing conditions during electrophoresis (Fig. 7f). Similar results were also seen when POPDC1 was co-expressed with POPDC3.
To investigate the stoichiometries of POPDC1–POPDC2 complexes within a cellular environment, a type-1 quantitative BRET (qBRET) assay was employed, following a previously reported protocol [13]. The NanoBRET platform [30] was utilized in the assay by tagging POPDC1 and POPDC2 at the C-terminus with NanoLuc luciferase (NL) or HaloTag (HT) fusion tags. These constructs were co-expressed in HEK293 cells at varying expression ratios, but at constant total expression levels (Additional file 1: Fig. S5) and the relationship to the BRET signal analyzed. Firstly, POPDC1-NL and POPDC1-HT were co-transfected to try and identify if homomeric interactions were present as previously reported [23, 26]. A clear hyperbolic BRET saturation curve indicative of a dimer was apparent (Fig. 7g). A rapidly saturating BRET curve was observed in the case of co-expression of POPDC2-NL and POPDC2-HT. Rapid saturation is a feature of higher order complexes, although such curve shapes make accurate determination of complex stoichiometry via type-1 qBRET studies difficult [13]. When POPDC1-HT and POPDC2-NL were co-expressed, the BRET saturation curve produced fitted to a dimer model. This suggests that the major POPDC1–POPDC2 complex is a dimer. Such heterodimers would have to compete against the tendency of POPDC1, and likely POPDC2, to form homodimers, which suggests that the heteromeric-interaction of POPDC1 and POPDC2 is favored.
We have previously shown that co-expression of POPDC1 or POPDC2 with the 2-pore domain potassium channel TREK-1 in Xenopus laevis oocytes leads to an increase in the outward K+ current compared to expression of TREK-1 alone [14]. This was attributed to a direct, cAMP-sensitive interaction between POPDC proteins and TREK-1. Incubation of the cells with 8-Br-cAMP, or the phosphodiesterase inhibitor theophylline, abolishes the increase in current in the presence of POPDC1 or POPDC2, respectively [14, 40]. To test if the formation of heteromeric POPDC1–POPDC2 complexes could modulate TREK1 current, one or both POPDC isoforms were expressed in Xenopus laevis oocytes alongside TREK-1 and two electrode voltage-clamp measurements used to determine the outward K+ current from cells. As expected, co-expression of POPDC1 or POPDC2 with TREK-1 led to a significant increase in TREK-1 current compared to TREK-1 alone (Fig. 7h, i). Furthermore, co-expression of POPDC1 and POPDC2 led to a significant additional increase in TREK-1 current above the levels observed with POPDC1 or POPDC2 alone. When the Xenopus oocytes were incubated with theophylline, the increase in TREK1 current after expression of POPDC1 and/or POPDC2 returned to baseline levels as previously reported [14, 40].
To help identify the domains responsible for the POPDC1–POPDC2 and POPDC1–POPDC3 interactions, two FLAG-tagged chimeras were constructed, consisting of the N-terminal and transmembrane domains of POPDC1 and the cytoplasmic region (including the Popeye domain) of POPDC2 and the inverse configuration (Fig. 7j, k). Each chimera was co-expressed in COS-7 cells with MYC-tagged POPDC1 or POPDC3. Precipitation of cell lysates using MYC-antibody led to co-precipitation of both chimeras in all cases.
To further map the sites in POPDC1, which mediate the interaction between POPDC1 and POPDC2, co-immunoprecipitation experiments in COS-7 cells were repeated in the presence of various C-terminal truncations of POPDC1 (Fig. 7l). The truncation mutants were C-terminally tagged with a MYC epitope and full-length POPDC2 with FLAG. Truncations were positioned to delete the C-terminal tail and end of the Popeye domain (Δ116), the C-terminal tail and half of the Popeye domain (Δ172), the C-terminal tail and the entire Popeye domain (Δ236) and the entire cytoplasmic region of POPDC1 (Δ245). After co-expression of these constructs with POPDC2, it was found that co-immunoprecipitation of POPDC2-FLAG was possible with all the truncation mutants, suggesting that the extracellular N-terminal region and transmembrane domains of POPDC1 were sufficient to form an interaction with POPDC2. A similar conclusion can probably be drawn for POPDC2, as the POPDC2 W188X mutant protein, which lacks the carboxy-terminal half of the Popeye domain and the carboxy terminus still retains the ability to interact with POPDC1 (Fig. 7m). These results, and behavior of the chimeric constructs, show that POPDC1–POPDC2 and POPDC1–POPDC3 interactions, occur at both the N-terminal/transmembrane domains and cytoplasmic portions of the proteins.
It was shown above that the POPDC1 p.Q153X, p.S201F, and POPDC2 p.W188X mutations led to mislocalization of POPDC1 and POPDC2 in HEK293 cells, as well as in skeletal muscle, while the POPDC1 p.V183F mutation had no effect in HEK293 cells and led to only mild changes in POPDC expression patterns in tissue. To investigate if a change in the interaction between POPDC1 and POPDC2 was responsible for this effect, a bimolecular fluorescence complementation (BiFC) assay was conducted (Fig. 7n). POPDC1 and POPDC2 wild-type and mutant constructs were tagged at the C-terminal with the split Venus domains VC155 and VN155, respectively, and expressed in HEK293 cells. Interactions between VC155 and VN155 lead to reconstitution of the Venus fluorophore. POPDC1-VN155 expressed with POPDC2-VN155 was utilized as a negative control, while mRFP was co-transfected into the cells to act as an internal transfection control to which BiFC signals were normalized. The mRFP signal was also used to define areas containing transfected cells within confocal microscopy images (approx. 50–100 cells per image) from which the BiFC signal was measured (Additional file 1: Fig. S6). As expected, wild-type POPDC1-VC155 and POPDC2-VC155 yielded a strong BiFC signal, which was set to equal 1, providing further evidence for the existence of POPDC1–POPDC2 complexes. No difference in the BiFC signal from POPDC1 p.V183F + POPDC2 compared to the wild-type pair was seen (n = 9, p = 0.54). However, a significant drop in the BiFC signal relative to the wild-type pair was observed in the presence of the POPDC1 p.Q153X (0.248, 95% CI 0.195, 0.301; n = 9), p.S201F (0.430, 95% CI 0.307, 0.553; n = 9), and POPDC2 p.W188X (0.590, 95% CI 0.449, 0.730; n = 5) variants (all p < 0.0001). However, a detectable BiFC signal greatly above background was observed in all groups suggesting the POPDC1–POPDC2 interaction was not fully abolished.
POPDC1 and POPDC2 may interact through a conserved interface in the αC-helix of the Popeye domain
Having demonstrated that POPDC1 and POPDC2 interact through both their N-terminal/ transmembrane and cytoplasmic regions, we focused on the role of the Popeye domain, which was previously reported to be involved in POPDC1 homomeric interactions [23, 26]. The cAMP binding domain of the prokaryotic cAMP-binding transcriptional regulator catabolite activator protein (CAP) shows the highest sequence similarity to the Popeye domain [43] and has therefore been used previously as a template for producing homology models of the Popeye domain [14]. CAP protein monomers dimerize through an α-helix at the C-terminal end of their cyclic nucleotide binding domain (CNBD), known as the C-helix, via a set of hydrophobic residues [31] (Additional file 1: Fig. S7a, b). As well as forming an interface between the CAP subunits, the C-helix also forms contacts with cAMP upon binding (Additional file 1: Fig. S7c) [34, 42]. The Popeye domains of POPDC1 and POPDC2 are predicted to be highly similar in structure to the CNBD of CAP [14], and the protein structure of the CAP dimer (PDB: 1G6N [31]) was utilized as a template to model the POPDC1–POPDC2 heteromeric complex (Fig. 8a). An α-helix is predicted to form at the C-terminal end of the Popeye domain and is referred to as the αC-helix in reference to the structures of PKA and other cAMP effector proteins [36, 46]. The model of the Popeye domain dimer possesses an interface between each αC-helix, analogous to the C-helix interface in CAP. Alignment of the amino acid sequences of the αC-helices of vertebrate POPDC1, POPDC2 and POPDC3, as well as the C-helix of CAP, reveals a high level of sequence conservation and in particular the invariant presence of a series of hydrophobic residues in each POPDC isoform, which, with the exception of one residue, were also present in the C-helix of CAP (Fig. 8b). Some of these residues are known to be involved in CAP dimerization (Additional file 1: Fig. S7b) [31, 34]. These hydrophobic residues show very strong structural alignment across the predicted structures of the αC-helix in POPDC1, POPDC2, and POPDC3 (Fig. 8c, d). The results shown above suggest that if normal POPDC1–POPDC2 interactions are disrupted, or absent, then the subcellular expression of both isoforms is altered. To determine if the CAP-aligned, highly conserved hydrophobic residues within the αC-helices of POPDC1 and POPDC2 are involved in complex formation, POPDC1-ECFP and POPDC2-EYFP were co-expressed in HEK293 cells, with each of the conserved hydrophobic residues in the αC-helix sequentially substituted to aspartic acid (Additional file 1: Figs. S8a, S9a). These substitutions were designed to disrupt the hydrophobicity of the putative helix-helix interface through the introduction of a negative charge. The median plasma membrane localization level of both POPDC isoforms across the cells was then determined as before and the difference to the wild-type pair analyzed. It was found that POPDC1 mutations F249D (0.812, 95% CI 0.584, 0.990; n = 56), I253D (0.841, 95% CI 0.571, 1.185; n = 27) and I257D (0.625, 95% CI 0.441, 0.741; n = 40) led to severe mislocalization of POPDC1 compared to the wild-type pair (3.676, 95% CI 3.176, 4.504; n = 34) (all p < 0.0001, Fig. 8e). These mutations had a similar effect on POPDC2 with F249D (0.845, 95% CI 0.650, 0.982; n = 56), I253D (1.202, 95% CI 0.790, 2.060; n = 27), and I257D (0.635, 95% CI 0.488, 0.779; n = 40) (all p < 0.0001) all leading to a reduction in the plasma membrane localization of POPDC2 compared to when co-expressed with wild-type POPDC1 (4.443, 95% CI 3.902, 6.662; n = 34). The other mutations within the αC-helix of POPDC1 did not lead to any significant changes in POPDC1 or POPDC2 plasma membrane localization (Fig. 8e). In POPDC2, the F233D (0.922, 95% CI 0.743, 1.277; n = 20), L237D (0.967, 95% CI 0.750, 1.267; n = 45), and I241D (0.735, 95% CI 0.642, 0.922; n = 42) mutations led to a significant disruption of POPDC1 plasma localization (all p < 0.0001) compared to when expressed with wild-type POPDC2 (Fig. 8f). These mutations also led to mislocalization of POPDC2 itself: F233D (0.778, 95% CI 0.651, 0.980; n = 20), L237D (1.121, 95% CI 0.860, 1.644; n = 45), and I241D (1.146, 95% CI 0.946, 1.479; n = 42) (all p < 0.0001). A mild reduction in POPDC2 plasma membrane localization was also seen with the I229D mutation (2.828, 95% CI 2.098, 3.337; n = 48, p = 0.049) (Fig. 8f). The L245D, L261D, and L264D mutations in POPDC1 and the I229D, L245D, and L248D in POPDC2, had no effect on the subcellular expression of either protein (except the minor change in POPDC2 expression in case of I229D). A loss in absolute plasma membrane expression was commonly seen in mutations that led to mislocalization of the proteins (Additional file 1: Figs. S8b, S9b), with lesser or no changes in cytoplasmic levels observed (Additional file 1: Figs. S8c, S9c). Residues that are aligned between POPDC1 and POPDC2 had very similar impacts on the plasma membrane localization of both isoforms (Fig. 8g, h). Mutation of the three residues at the core of the proposed αC-helices led to major losses in POPDC1 and POPDC2 plasma membrane localization, while substitution of the residues at the N- and C-termini of the helices had no, or only minimal effect (Fig. 8i).