Female wildtype (WT) C57Bl/6 and IL10-KO (B6.129P2-IL10 ^tm1Cgn /J) mice were purchased from Jackson Laboratory (Bar Harbor, ME; National Institute on Aging, Bethesda, MD) and housed under specific pathogen-free barrier conditions until the age of 21–22 months in facilities accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International. Nineteen mice (eleven IL10-KO and eight WT) were euthanized by administered inhalation anesthesia in a plastic chamber under a vented hood using gauze soaked in pharmaceutical-grade isoflurane, followed by cervical dislocation once the mouse was unconscious. Death was verified by observed cessation of breathing and heartbeat. Tissues were harvested, weighed, flash-frozen in liquid nitrogen and stored at −80°C as described (Sikka et al., 2013; Malinina et al., 2020). All protocols were approved by the Johns Hopkins University Institutional Animal Care and Use Committee, and all experiments performed accordingly.

Frozen tissues were each pulverized under liquid nitrogen using a pre-chilled mortar and pestle, and powdered tissue was stored at −80C or in liquid nitrogen until use. To prepare lysates, we added powdered tissue (80–100 mg) to 500 uL ice-cold lysis buffer (50 mM Tris-HCl pH 7.4, 300 mM NaCl, 0.3% v/v Triton-X100, 5 mM EDTA, 1 mM DTT, 100 uM PMSF, 1 ug/mL Pepstatin A, 1X Thermo Scientific Halt Protease Inhibitor Cocktail #78430, 50 nM Thiamet G [OGA inhibitor; 1,2-dideoxy-2′-ethylamino-α-d-glucopyranoso-[2,1-d]-Δ2′-thiazoline; provided by G. W. Hart) and 50 mM UDP-GlcNAc [Sigma]) in a 2 mL Eppendorf tube on ice, and moved to liquid nitrogen as needed to keep frozen until further processing. Thiamet G and UDP-GlcNAc were included to maintain labile O-GlcNAc modifications. Samples were flick-vortexed to mix, incubated 10 min on ice, vortexed 10 s at high speed, then sonicated on ice 20 times (0.5 s bursts) and finally centrifuged 30 min (16,000 g, 4°C) to pellet insoluble material. Sonication greatly facilitates solubilization of lamins and associated proteins (Berk et al., 2013; Berk and Wilson, 2016). Supernatant protein concentrations were measured via Bradford assay and adjusted with lysis buffer to 1 ug/uL before use.

Protein extracts (40 ug total at 1 ug/ul in 1% SDS) were reduced with 15 uL of 15 mM DTT for 1 h at 56°C, alkylated by adding 15 uL 100 mM iodoacetamide and incubating in the dark for 30 min, and then TCA/Acetone precipitated. The protein pellet from each sample was digested overnight at 37°C by adding 100 uL Trypsin/LysC mixture [40 ug proteases in 1.2 mL of 100 mM triethylammonium bicarbonate (TEAB; Promega #V5071) or approximately 3.33 ug protease per sample]. Individual samples (40 ug) were labeled with a unique isobaric mass tag reagent (TMT 10-plex, Thermo Scientific) according to manufacturer instructions. Both pairing and labeling order of TMT reagent and peptide sample were randomized. Briefly, the TMT reagents (0.8 ug vials) were allowed to come to room temperature before adding 41 uL anhydrous acetonitrile, then vortexed and centrifuged. The entire TMT reagent vial was added to the 100 ug peptide sample and reacted at room temperature for 1 h. The reaction was quenched by adding hydroxylamine (8 uL) to a final concentration of 5%. All TMT-labeled samples were combined and vacuum centrifuged to dryness removing the entire liquid.

Data were analyzed using R version 4.0.4 and all new bioinformatics and statistical analyses described here are available (GitHub repository: https://github.com/aditharun/frailty-laminome). To assess potential differences between the IL10-KO and WT proteomes in heart or muscle, we used a rigorous statistical method developed by Rucinski and others to detect significant changes in protein abundance (Kammers et al., 2015). Starting from the raw peptide spectra data, we removed proteins with an Isolation Interference greater than 30% or which had missing values, yielding Supplementary Table S1. We then normalized the abundances to the lamin A abundance in each channel (i.e., the same sample), and performed a two-sample Student’s t-test using an empirical Bayes method (R script: https://github.com/aditharun/frailty-laminome/blob/main/code/analysis.R). This script produces output spreadsheets that can either include all proteins (Supplementary Table S2), or a subset that excludes as presumed contaminants mitochondrial, ribosomal and keratin proteins. For downstream analysis and succinct reference, we developed a script that matches protein accession numbers with gene names (https://github.com/aditharun/frailty-laminome/blob/main/code/get_gene_names.R) from NCBI (https://www.ncbi.nlm.nih.gov/gene) using the R API for the NCBI Database and searching for the accession number of a particular protein to return a unique identifying number. This number was then matched, also using the NCBI API, to a specific gene name (Supplementary Table S3). Some gene names were unavailable and curated manually in the final table (Supplementary Table S4), which was also curated manually (non-exhaustively) to indicate reported nuclear envelope transmembrane proteins [“NETs”; (Malik et al., 2010)] and nuclear-localized proteins.

The scripts used for volcano plots are available here: https://github.com/aditharun/frailty-laminome/blob/main/code/volcano-stratification.R.

Three identical custom peptide arrays were synthesized and printed on cellulose membranes by the Biopolymers and Proteomics Core Facility at the Koch Institute Swanson Biotechnology Center at Massachusetts Institute of Technology (Cambridge MA) as described (Frank and Dübel, 2005), using a Multipep automated peptide synthesizer (INTAVIS Bioanalytical Instruments AG, Koeln, Germany) as described (Frank and Overwin, 1996). The following proteins, all human, were each displayed as synthetic 20-mer peptides with seven amino acid offsets: Fabp4 (132 residues; NP_001433.1), Gins3 (216 residues; NP_073607.2), Perm1 (790 residues; NP_001356826.1), Tmem38a (299 residues; NP_076979.1), AldoA (364 residues; NP_908930.1), Csrp3 (194 residues; P50461-1), Lmcd1 (365 residues; Q9NZU5-1), emerin (254 residues; NP_000108.1), plus other proteins reported separately.

Arrays were incubated with C-terminally His-tagged residues 385–646 of human mature lamin A, affinity-purified from E. coli lysates. Protein expression was induced using 1 mM isopropyl-β-D-thio-galactoside (IPTG; 3 h, 37°C), and His-tagged polypeptides were affinity-purified as described (Simon et al., 2010), using Cobalt-charged TALON^® metal ion affinity resin (Takara Bio United States, #635502).

Peptide arrays were probed as follows, at room temperature (20°C–22°C) unless otherwise noted. After initial wetting (∼5 min in methanol), each array was washed three times in TBS (pH 7.0) in a polystyrene plate, and then incubated overnight in Membrane Blocking Solution (MBS), made by mixing 20 mL concentrated Casein blocking buffer (Sigma-Genosys #SU-07-250) with 80 mL TBS-T [Tris-Buffered Saline (150 mM NaCl, 20 mM Tris) plus 0.05% (vol/vol) Tween 20; pH 8.0] and 5 g sucrose, then adjusting dropwise with NaOH to final pH 7.0. The arrays were washed twice in TBS-T, then incubated 3 h in 30 mL MBS containing 150 ug recombinant lamin tail protein (final lamin concentration, 190 nM). After two washes in TBS-T, arrays were incubated 1 h in MBS containing 20 uL anti-lamin A/C mouse clone 14 (Millipore Sigma #05-714; 1:1000 dilution), washed twice in TBST and incubated 1 h in MBS containing 6 uL AP-conjugated donkey anti-mouse (Jackson ImmunoResearch #715-055-150; 1-to-5,000 dilution), and finally washed twice in TBS-T. To detect bound antibodies, arrays were washed twice in citrate-buffered saline (CBS: 137 mM NaCl, 2.7 mM KCl, 50 mM citric acid monohydrate, pH 7.0). The arrays were transferred to flat glass trays before adding 20 mL Color Developing Solution, made fresh by adding 80 uL BCIP solution [made by dissolving 60 mg BCIP (5-bromo-4-chloro-3-indolylphosphate p-toluidine salt; Sigma #51K1567) per mL absolute dimethylformamide (DMF; Cell Signaling Technologies #12767)] to 120 uL MTT solution [made by dissolving 50 mg of 3-(4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide (MTT) per mL 70% v/v DMF in water] and 100 uL MgCl2 (1 M stock) to 20 mL CBS. Note this specific form of BCIP is crucial to remove the purple color for re-use. The array was then incubated 30–35 min in Color Developing Solution, watching for color development, washed twice in PBS to stop the reaction, and imaged using an Azure Biosystems C600 imager, with the RGB capture setting (Cy2, Cy3, Cy5) and autoexposure. Arrays were stored at 4°C in PBS until the next stage, when they were stripped to remove lamins, antibodies and purple color in a multi-step process: 1) washed twice in 20 mL ddH2O, 2) incubated in 20 mL DMF until the purple color disappeared (typically ∼10 min), 3) washed three times in ddH₂O, 4) transferred into a plastic “pouch” (Kapak by Ampac, 404-24) and washed three times (10 min each) in ∼30 mL Stripping Mix A (PBS, pH 7.0, containing 8 M Urea, 1% SDS and 0.5% beta-mercaptoethanol) in a 40°C sonication bath, 5) washed three times in Stripping Mix B (10% acetic acid, 50% ethanol, in ddH₂O), and 6) washed three times in 100% ethanol. The stripped array was then imaged and stored at 4°C in PBS until the next cycle of probing and stripping.

Proteins immunoprecipitated from each sample were uniquely covalently marked using Tandem Mass Tags (TMT; Thompson et al., 2003), a stable isotope-based approach that allows all ten samples from the same tissue (e.g., hearts) to be pooled for multiplex analysis and quantification by mass spectrometry. Potential changes in lamin A/C association were measured after normalizing each protein to the lamin A/C signals in the same sample. Significant log2-fold changes in protein abundance were quantified based on normalized WT-to-IL10-KO signal ratios using the empirical Bayes method described by Rucinski and others (Kammers et al., 2015). Raw results for all identified proteins in skeletal muscle and heart are provided in Supplementary Table S1, and normalized results in Supplementary Table S2.

To focus further analysis, proteome datasets were filtered by removing ribosomal proteins, keratins and most mitochondrial proteins as presumed contaminants (Supplementary Table S3). Sarcomeric proteins, which are highly abundant and relatively insoluble in heart and muscle, were also presumed contaminants of immunoprecipitation but were not filtered out. Our final filtered Supplementary Table S4 was curated manually to note proteins that were: 1) known or predicted (Cytoscape version 3.8.2) to localize in the nucleus, 2) reported to bind lamin A/C directly, or 3) identified as nuclear envelope integral membrane proteins (“NETs”). NET classification was based on biochemical isolation of native NETs from liver, white blood cells or C2C12 myoblasts [Supplementary Table S4 in Korfali et al. (2010), Korfali et al. (2012), Worman and Schirmer (2015)], or from cultured mesenchymal stem cells, adipocytes or myocytes [Supplementary Tables S1, S2 in Cheng et al. (2019)].

We identified ∼340 proteins in heart and ∼500 proteins in skeletal muscle with high confidence (Supplementary Table S4). Sixty-two proteins were identified in both heart and muscle (examples shown in Figure 1), including chromatin regulators Smyd1 [adds the repressive H3K4me “mark” on histone H3 (Tracy et al., 2018)] and SetD2 [adds the H3K36me3 mark; (McDaniel and Strahl, 2017)], dystrobrevin-alpha [Dtna; (Aguilar et al., 2015)]; and Tmem38a (aka “Trimeric intracellular cation channel type A”), a nuclear membrane protein required for muscle-specific 3D genome organization (Robson et al., 2016) and genetically linked to Emery-Dreifuss muscular dystrophy (Meinke et al., 2020). Additional proteins identified in both tissues included NAMPT [controls the rate-limiting step in NAD biosynthesis, and the release of Clock-Arntl/BMAL1 heterodimers from Sirt1-mediated repression; (Ramsey et al., 2009); (Maynard et al., 2022)], Gapdh [glyceraldehyde-3-phosphate dehydrogenase; enters the nucleus during glucose starvation (Chang et al., 2015) and autophagy (Iqbal et al., 2021)], transcription factor Fhl1 (Roux et al., 2012) which interacts with emerin and lamin A and is genetically linked to EDMD (Ziat et al., 2016), and shuttling transcription factor Csrp3 [“muscle LIM protein”; a MyoD1 co-factor that enters the nucleus in response to mechanical force, and protects against muscular dystrophy (Mathiesen et al., 2019)].

A small number of nuclear envelope integral membrane proteins (“NETs”) and characterized lamin A/C-binding proteins were identified in the heart, including MLIP (Ahmady et al., 2011), Plcb4 (Dittmer et al., 2014), transcription regulator Kank2 (Dittmer et al., 2014), Sun2 [NET; (Haque et al., 2010)] and RyR2 [ryanodine receptor 2; (Kapiloff et al., 2001; Dridi et al., 2021)]. In skeletal muscle we identified Ryr1 [Ryanodine receptor 1; (Dridi et al., 2021), Matrin 3 (Depreux et al., 2015) and lamin A-proximal proteins Numa1, Pcbp1 (poly(rC)-binding protein, aka HnrnpE1)] and RanGap1 (Roux et al., 2012; Supplementary Table S4). Other expected partners such as emerin were not recovered from striated muscle, but were identified in brain, a soft tissue.

Most proteins identified in this study were not detectably affected in IL10-KO tissues, as seen in volcano plots for skeletal muscle (Figure 2A) and heart (Figure 2B). This was encouraging because it suggested these proteins reproducibly co-precipitated with native A-type lamins, whether due to genuine association (direct or indirect) or artifact. Curated examples of IL10-KO-unaffected proteins are shown for skeletal muscle in Table 1, and for heart in Table 2. Proteins unaffected by chronic inflammation (IL10-KO) were still of great interest as candidate (novel) partners for A-type lamins in these native tissues. Skeletal muscle candidates included Rragd [regulates mTORC1; (Tsun et al., 2013)], condensin subunits NCAPH (condensin complex subunit 2) and Smc3 (structural maintenance of chromosomes protein 3), chromatin repressor Sirt2 (NAD-dependent protein deacetylase sirtuin-2), scaffolding proteins 14-3-3 epsilon (Ywhae) and 14-3-3 gamma (Ywhag), and transcription factors Bin1 (Myc box-dependent-interacting protein 1), smoothelin, Thrap3 (thyroid hormone receptor-associated protein 3), Dmrt (doublesex- and mab-3-related transcription factor 2), Stat5B and Fhl3 (Table 1), and signaling kinase Akt2 (which can phosphorylate lamin A; Fan et al., 2023). Others included MAP “kinase kinases” (Map2k3, Map2k4, and Map2k6), multiple subunits of the 5′-AMP-activated protein kinase complex (Prkaa2, Prkab2, Prkag1, Prkaca, and Prkar2a), Ca++/calmodulin-dependent protein kinase type II (Camk2a and Camk2g), the protein phosphatase 1 complex (Ppp1cc, Ppp1r12b, Ppp1r3a), as well as protein phosphatase 1B (Ppm1b), protein phosphatase 2 (Ppp2cb), and “TRiC” chaperonin complexes [T-complex protein 1, Cct2/beta, Cct3/gamma, Cct4/delta, Cct7/eta, Cct8/theta; (Jin et al., 2019); Supplementary Table S4]. TRiC complexes are huge (∼1,000 kD) and might have co-precipitated as an artifact of density, but also reportedly associate with heterochromatin (Souès et al., 2003) and function in the nucleus (Pejanovic et al., 2012). Other novel candidates from heart included transcription factor Ndrg2, which protects against ischemia-reperfusion injury (Sun et al., 2013), several kinases (e.g., Speg, Skp1, Taok1, Sgk223), Pias2 (E3 SUMO-protein ligase), Mad1l1 (mitotic spindle checkpoint protein MAD1) and Ppp2ca (Ser/Thr-protein phosphatase 2A catalytic subunit alpha), which regulates an essential lamin-binding protein named Barrier to Autointegration Factor 1 (BANF1) and influences postmitotic nuclear assembly (Ahn et al., 2019).

Two proteins, Perm1 (PGC-1 and ERR-induced regulator in muscle protein 1) and Fam210A, were unaffected in IL10-KO hearts but intrinsically interesting as frailty-relevant candidates. Perm1 is an intrinsically disordered protein, highly expressed in heart and skeletal muscle, that regulates genes required for endurance exercise, mitochondrial biogenesis and oxidative capacity in muscle (Cho et al., 2016; Cho et al., 2019; Cho et al., 2021). Perm1 localizes primarily at sites of mitochondrial-ER contact but also enters the nucleus and promotes transcription of genes required for fatty acid oxidation (Huang et al., 2022). Perm1(aka C1orf170) showed selective binding to “progerin,” the toxic internally deleted form of prelamin A, in a yeast two-hybrid study (Dittmer et al., 2014). The other frailty-relevant candidate, Fam210A is genetically linked to human sarcopenia, bone fractures and reduced grip strength (Tanaka et al., 2018; Trajanoska et al., 2018; Tanaka et al., 2020). Fam210A regulates mitochondrial-encoded gene expression and localizes both in mitochondria and the cytoplasm but is relatively uncharacterized as a protein (Wu et al., 2021).

Certain candidates identified in both tissues were affected by IL10-KO in one tissue. For example, metabolic regulators Fabp4 and Tpi1, with opposite effects on the “browning” of white fat [Fabp4 inhibits; Tpi1 promotes; (Liu et al., 2022)], were both significantly increased in IL10-KO muscle (p < 0.0255 for Fabp4; p < 0.0374 for Tpi1; Table 1). Tpi1 controls nuclear acetate levels and influences global histone acetylation (Zhang et al., 2021). Cryab (aka Hspb5), which enters the cardiomyocyte nucleus in response to non-damaging endurance exercise (Antonioni et al., 2020), was identified in both tissues and showed higher association in IL10^tm hearts (p < 0.0146; Table 2). AldoA (Aldolase 1A retrogene 1), which has central roles in glycolysis and gluconeogenesis in the cytoplasm and regulates ribosome biogenesis in the nucleus (Schwarz et al., 2022), showed reduced association in IL10-KO hearts (p < 0.0031; Supplementary Table S4).

The subset of proteins from skeletal muscle that changed significantly (p < 0.05) in the frailty model are all listed in Figure 1 and detailed in Supplementary Table S4. Lamin A/C association was significantly reduced for Lmcd1 (p < 0.033), a transcription factor that increases skeletal muscle mass (Ferreira et al., 2019), and three other proteins of interest: Ppme1 [p < 0.034; methylates and inhibits protein phosphatase 2A (Xing et al., 2008; Kauko et al., 2020) and colocalizes with lamin A/C (Pokharel et al., 2015)]; Fxr1 (Fragile X syndrome-related protein 1; p < 0.049, Table 1), and Fbp2 (fructose-1,6-bisphosphatase isozyme 2; p < 0.021), a mitochondrial protein that also binds c-Myc and represses c-Myc-dependent transcription of TFAM, a master (positive) regulator of mitochondrial gene expression, in the nucleus (Huangyang et al., 2020).

The subset of proteins from heart that changed significantly (p < 0.05) in the frailty model are all listed in Figure 1 and detailed in Supplementary Table S4 . Lamin A/C association in heart was reduced for ER membrane protein HHATL (protein-cysteine N-palmitoyltransferase HHAT-like protein; p < 0.00034) and two mitochondrial proteins: Acsl1 (long-chain-fatty-acid CoA ligase 1; p < 0.00005) and Cpt1B (carnitine O-palmitoyltransferase 1; p < 0.0016; Table 2; see Discussion). Lamin A/C association increased for Myl6 (myosin light chain 6; p < 0.00076), Pygb (glycogen phosphorylase, brain isoform; p < 0.0088; Uno et al., 1998; enters nucleus: Sun et al., 2019), Cryab (p < 0.015), Aldoa (p < .0.003) and Gins3 (3.5-fold increase; p < 0.041; Table 2). Gins3 is a subunit of DNA helicase complexes (Kamada, 2012) that also regulates myocardial repolarization (Milan et al., 2009; Newton-Cheh et al., 2009) and is downregulated in metabolically unhealthy obese adults (Das et al., 2015).

There were many reasons why protein association with lamin A/C might have changed in IL10-KO tissues, from altered mRNA expression to changes in the posttranslational modifications, nuclear localization or stability of any given candidate protein. We therefore chose a proof-of-principal validation question: did these native proteomes include any novel lamin A-binding proteins? We knew it was feasible to use recombinant lamin A “tails” (residues 385–646) as probes to detect binding to SDS-PAGE-resolved partners such as emerin (Lee et al., 2001), and we previously used peptide arrays to map sites of emerin-emerin interaction (Berk et al., 2014) that were validated in mechanically stressed cells (Fernandez et al., 2022). We therefore chose a peptide array strategy to test candidate binding to recombinant lamin A.

We screened 11 candidates, favoring smaller proteins to fit more per array. Seven candidates are reported here; four will be reported with the brain proteome. Each candidate was displayed as a series of 20-mer synthetic peptides (staggered by seven residues) on cellulose (“peptide array”). The entire array was then probed with purified recombinant mature lamin A “tail” residues 385–646, which includes the flexible linker (residues 385–429) and Ig-fold domain (residues 430–544) shared with lamin C, plus disordered C-terminal residues 545–646 unique to mature lamin A. We used two identical arrays. One array was incubated first with recombinant lamin A, and then with primary (anti-lamin A/C) and secondary antibodies to detect the bound lamins, as shown in panels A of Figures 3–9 (first probe: LamA + Abs). To control for array ‘stickiness’ during their first use, our negative control for the first experiment was to incubate an identical array first (and solely) with antibodies, as shown in panels A of Figures 3–9 (first probe: control Abs-only). For the second full experiment, both arrays were urea-stripped, probed with lamin A protein and antibodies (Figures 3–9, second probe: LamA + Abs), then urea-stripped again and probed with antibodies alone (Figures 3–9, second probe: control Abs-only). The arrays also included peptides representing the lamin A tail itself (residues 385–664), and peptides unique to either lamin C (DEDEDGDDLLHHHHVSGSRR) or “progerin” (CGQPADKASASGSGAQSPQN and ADKASASGSGAQSPQNCSIM), or nuclear membrane protein emerin as positive control. Because many peptides gave high (Ab-only) backgrounds, spot signals were scored independently by two individuals as either weak (+), moderate (++) or strong (+++), compared to the corresponding Ab-only control (panels B of Figures 3–9). Potential sites of Lamin A-binding in each candidate were judged based on signal intensity and consistency (positive in both experiments, or consecutive overlapping peptides), and annotated in the full amino acid sequence (panels C of Figures 3–9). For candidates with known atomic structures, we used PyMol and blue shading to determine if putative lamin-binding peptides were plausibly solvent-exposed. For structurally uncharacterized candidates (Lmcd1, Tmem38a), we relied on Alphafold predictions.

We included lamin A peptides in the arrays both to query tail-tail interactions, and as a positive control to identify the epitope recognized by the detection system’s lamin A/C antibody. Our detection system (primary anti-lamin A/C and secondary antibodies) was strongly positive for lamin A peptides w16, w17 and x17 (black boxes, Figure 3A, first probe, control Abs-only). The overlapping peptides w16 (aa 464–483; MGNWQIKRQNGDDPLLTYRF) and w17 (aa 471–490; RQNGDDPLLTYRFPPKFTLK) are located in the Ig-fold shared by lamin A and lamin C; from this we deduced the epitope resided in residues 464–490 (gray; Figures 3C,D). Positive x17 (aa 611–630; ISSGSSASSVTVTRSYRSVG) was unique to lamin A, hence ruled out as the primary epitope. We speculate the epitope includes “⁴⁸⁰TYR,” since peptide x17 included homologous residues ‘⁶²⁵SYR’. However, this and other background positives remain unexplained.

We detected consistently strongest specific binding of lamin A tails to prelamin A peptide W14 (aa 450–469; KFVRLRNKSNEDQSMGNWQI) and slightly lower signals for overlapping peptides W19 (aa 485–504; PKFTLKAGQVVTIWAAGAGA) and X1 (aa 499–518; AAGAGATHSPPTDLVWKAQN; spots blue-boxed in Figure 3A; scored in Figure 3B). All three positives were located in the Ig-fold, as shown in the amino acid sequence (gray underbar, Figure 3C) and depicted in Figure 3D. The strongest positive, peptide w14, includes many solvent-exposed residues that snake around the bottom and back of the Ig-fold (dark blue in Figure 3D). The other positives, also largely solvent-exposed, occupy most of the “bottom” and “back” surfaces of the Ig-fold (light blue; Figure 3D). These proposed lamin tail-tail interaction regions do not overlap the binding site for BANF1 [pink; front view in Figure 3D; (Samson et al., 2018)]. Lamin A tails did not bind detectably to the lamin C-specific peptide “y3” (Figure 3A) or progerin-specific peptides (data not shown). In the context of lamin filaments, these results suggested that an unidentified region(s) of the lamin A tail (Ig-fold, A-specific disordered region, or both) can interact with the “bottom” and “back” of neighboring Ig-folds (Figure 3D), consistent with an elegant molecular crosslinking study of native lamin A filaments in living cells (Makarov et al., 2019).

The antibody-only control gave puzzlingly strong recognition of emerin, especially peptides y6 (aa 15–24; LLRRYNIPHGPVVGSTRRLY) and z11-z12 (aa 190–216; FMSSSSSSSSWLTRRAIRPENRAPGAG), and also broadly recognized the N-terminal region (y4–y10), middle (y16–y18; aa 85–118) and C-terminal region (z14–z18; aa 211–254) (Figures 4A, B; first probe, control Abs-only). We first checked, and ruled out, potential switching of the two images. Against this formidable and disappointing background, one peptide was weakly positive in both experiments: emerin z10 (aa 183–202; PTSSSTSFMSSSSSSSSWLT). This peptide resides in one of two large fragments of emerin, namely residues 1–132, and 159–220— each sufficient to bind lamin A (Berk et al., 2014). This result suggested lamin A contacts emerin residues 183–202 near the transmembrane domain, between two peptides (SAYQS region and R-peptide) involved in emerin-emerin association (Berk et al., 2014; Figure 4C).

We detected strong lamin A binding to three Perm1 peptides (Figures 5A, B), namely h5 (aa 441–460; SSGLVSTPVPRADAAGLAWP), i4 (aa 593–612; EENEEAEAAAAGQDPAGVQW) and j2 (aa 737–756; FVAFATWAVRTSDPHTPDAW), and consistent weaker binding to h16 (aa 529–548; QTATGAHGGPGAWEAVAVGP), as annotated in the full amino acid sequence of Perm1 (Figure 5C). We were unable to evaluate these peptides in a structural context, since Perm1 is almost entirely disordered, except to note that the proposed lamin A-binding peptides are all located in the C-terminal half of Perm1, and that two sites have predicted α-helicity (Figure 5D). Our conclusion that Perm1 can bind mature lamin A was unexpected, since a previous yeast two-hybrid study reported detectable binding of Perm1 (known as C1orf179) only to progerin (Dittmer et al., 2014).

Csrp3, also known as “muscle LIM protein” (MLP), is a nucleocytoplasmic shuttling protein that crosslinks and bundles F-actin and influences myocyte remodeling and responses to cardiac hypertrophy (Hoffmann et al., 2014). We observed weak lamin A tail binding to Csrp3 peptide s7 (aa 1–20; MPNWGGGAKCGACEKTVYHA) in both experiments, and weak binding to overlapping peptides s9–s11 (aa 15–48; KTVYHAEEIQCNGRSFHKTCFHCMACRKALDSTT) and t5–t6 (aa 127–146; VYAAEKVMGGGKPWHKTCFRCAICGKS) in the second experiment (Figures 6A, B). We were unconfident about the strong signals for peptides s14–s15 in the second experiment due to their high background signals in the first experiment (Figures 6A, B). The full amino acid sequence of Crsp3 is shown in Figure 6C. Csrp3 has two zinc-binding domains: LIM domain-1 (residues 7–66; Figure 6D) mediates Csrp3 dimerization, whereas LIM domain-2 (residues 119–176; Figure 6D, right) binds F-actin (Hoffmann et al., 2014). Our array data suggests lamin A binds disordered residues 1–6 and solvent-exposed residues 7–20 of LIM domain-1 (shaded blue in Figure 6D) and might therefore influence Csrp3 dimerization.

We detected strong consistent lamin A tail binding to Lmcd1 peptide v17 (aa 309–328; SGCDEIIFAEDYQRVEDLAW), and consistent but weaker binding to three peptides (u8–u10) that shared residues 113–132 (FDTITYEWAPPGVTQKLGLQ) (Figures 7A, B). Given the limitations of this assay we drew no conclusions from the faint signals seen with peptide t13 (aa 1–20), and other peptides that were positive only in the second experiment (Figures 7A, B). Convincing positives were annotated in the amino acid sequence of Lmcd1 (Figure 7C). Lmcd1 has a putative protein-protein interaction (“PET”) domain (residues 99–206), a disordered region (residues 200–235) and two predicted LIM zinc-binding domains (residues 241–306, and 307–365; Figure 7C). The strongest lamin A-binding site (peptide v17, in the second zinc-binding domain) is almost entirely solvent-exposed in the Alphafold-predicted structure of Lmcd1 (shaded blue in Figure 7D). The weaker lamin A-binding site (residues 113–132) in the putative protein-protein interaction region includes eight predictedly solvent-exposed residues (shaded blue in Figure 7D).

Lamin A tails showed consistently strong binding to overlapping AldoA peptides r16 (aa 281–300; SINLNAINKCPLLKPLLKPWALTF), r18 (aa 295–314; PWALTFSYGRALQASALKAW), r19 (aa 302–321; YGRALQASALKAWGGKKENL) and r20 (aa 309–328; SALKAWGGKKENLKAAQEEY; Figures 8A, B). We were intrigued to see that peptide r16 (aa 281–314) was mostly buried in the X-ray crystal structure (Figure 8D). The exceptions were residues P295, W296, A297 (“PWA”), on a concave surface (shaded blue in “front” view, Figure 8D), and residues K312, A313 and W314 (‘KAW’; shaded blue or white in Figures 8C, D). “PWA” was shared by overlapping AldoA peptides r16 and r18 (Figure 8B), and “KAW” by three strongly positive overlapping peptides (r18–r20; Figure 8B). We noticed similar “AW” residues in strongly positive peptides from Lmcd1 (³²⁶LAW; Figure 7B) and Perm1 (⁴⁵⁷LAW, ⁵³⁹GAW, ⁷⁵⁴DAW; Figure 5B); comparison suggested elements of a potential shared motif (see Discussion).

Lamin A showed consistently strong binding to Tmem38A peptide j13 (aa 36–55; YLKYEPGAVELSRRHPIASW; Figures 9A, B). Other consistent positives were k10 (aa 155–174; KGSGVALMSNFEQLLRGVWK) and k16 (aa 197–216; FTLQQTRWLPVSKASLIFIF). We disregarded weaker positives k12 (aa 169–188; LRGVWKPETNEILHMSFPTK) and L7–L8 (aa 274–299; QHSAMPAKSKEELSEGSRKKKAKKAD). We disregarded strong peptides k6–k7 because they were inexplicably positive in the antibody-only control (Figure 9A, first probe). Tmem38 is an integral membrane protein that localizes at the nuclear envelope inner membrane in muscle cells (Robson et al., 2016). Tmem38A is homologous to trimeric intracellular cation type-A (TRIC-A) channels, crystal structures for which were determined only in procaryotic and C. elegans orthologs (Kasuya et al., 2016). The structure of human Tmem38A has not been determined and was variously predicted to have three (Zhou et al., 2014) or four (Meinke et al., 2020) transmembrane helices. Our Alphafold prediction suggests five membrane-spanning helices (gray bars in Figure 9C; helices 1–3 colored green and helices 4–5 red in Figure 9E), three membrane-adjacent helices (residues 1–18, 236–249, and 285–295), and at least one “kinked” (partially bilayer-inserted) helix. Putative helix-2 and helix-5 are quite long and probably protrude beyond the lipid bilayer. Since Alphafold does not depict the lipid bilayer, we approximated crudely as depicted by gray lines in Figure 9D. This Alphafold-predicted topology suggests the C-terminus, lamin A-binding peptide j13 (residues 36–55) and most of lamin-binding peptide k10 (residues 158–174) are exposed and accessible in vivo (AF-Q9HF2-F1; residues 36–55 dark blue, residues 158–174 light blue, in Figures 9D–F). The third identified site (peptide k16) is predicted to localize in the lumen (Figure 9D), inaccessible to lamins in vivo.

Our evidence suggests lamin A “tails” (including the Ig-fold), which extend flexibly away from the filament backbone, have the capacity to interact with nearby tails either on the same filament, or possibly different filaments. These results independently support previous crosslinking evidence that lamin tails can associate with each other in the context of native filament networks (Makarov et al., 2019). Furthermore our lamin-binding peptides were located on the “bottom” and “back” of the Ig-fold, away from the binding site for BANF1 (Samson et al., 2018), suggesting these proposed tail-tail interactions would not interfere with BANF1, an essential partner. More work is needed to determine which region of lamin A (Ig-fold, or disordered A-specific tail) contacts neighboring Ig-folds, and whether such interactions affect filament dynamics in vivo (Makarov et al., 2019).

The genes encoding four novel lamin A-binding proteins identified in this study are worth testing for potential genetic linkage to “unmapped” EDMD patients. These proteins may also be relevant to understanding the pervasive muscle weakness phenotypes of frail patients.

Tmem38A has two unreconciled roles: on one hand, an intracellular cation channel at the nuclear inner membrane; on the other, a muscle-specific regulator of 3D chromosome organization (Czapiewski et al., 2016; Robson et al., 2016). Given the current uncertainty about the structure and topology of Tmem38a, our best prediction is that two of our three identified lamin-binding sites in Tmem38a are exposed and accessible to lamins in vivo (Figures 9D, E). Interestingly, none of our lamin A-binding peptides included residue N261 (boxed in Figure 9C), the site of a recently-reported EDMD-causing mutation, p.N260D [numbered without the initiating Met; (Meinke et al., 2020)]. Since residue N261 is predicted to be exposed (Figure 9D), we speculate that this mutation disrupts Tmem38A in some other way, for example by perturbing Tmem38A binding to a different EDMD-relevant partner such as emerin, Fhl1, Lemd3/Man1, Sun1, Sun2, Tmem201, Tmem43/Luma, Nesprin-1, Nesprin-2 or Nesprin-3 (Meinke et al., 2020). Further work on the structure and function of Tmem38a is needed to understand this fascinating and disease-relevant protein.

Lamin A gave no detectable binding to Fabp4 or Gins3 peptides (data not shown). Since negative results are inconclusive, further studies with full-length proteins will be needed to retest potential binding to lamin A, and also test lamin C. One enduring mystery is how lamins A and C, with residues 1–566 identical, nevertheless form separate filaments (Shimi et al., 2015) and have distinct roles in metabolism and lifespan (Fong et al., 2006; Lopez-Mejia et al., 2014) and tissue-specific 3D chromosome organization (Wong et al., 2021b). As lamin A/C-associated and IL10-KO-affected candidates important for metabolism (Garin-Shkolnik et al., 2014; Das et al., 2015; Hotamisligil and Bernlohr, 2015; Liu et al., 2022), Fabp4 and Gins3 both warrant further investigation.