Nuclear membrane diversity: underlying tissue-specific pathologies in disease?

Abstract

Human ‘laminopathy’ diseases result from mutations in genes encoding nuclear lamins or nuclear envelope (NE) transmembrane proteins (NETs). These diseases present a seeming paradox: the mutated proteins are widely expressed yet pathology is limited to specific tissues. New findings suggest tissue-specific pathologies arise because these widely expressed proteins act in various complexes that include tissue-specific components. Diverse mechanisms to achieve NE tissue-specificity include tissue-specific regulation of the expression, mRNA splicing, signaling, NE-localization and interactions of potentially hundreds of tissue-specific NETs. New findings suggest these NETs underlie tissue-specific NE roles in cytoskeletal mechanics, cell-cycle regulation, signaling, gene expression and genome organization. This view of the NE as ‘specialized’ in each cell type is important to understand the tissue-specific pathology of NE-linked diseases.

Introduction

As the complexity of multicellular organisms increased during evolution, so did the number of proteins that localize and function at the nuclear envelope (NE) [1,2]. In animals these proteins include lamins, which form nuclear intermediate filaments, and growing numbers of NE transmembrane proteins (NETs). Many NETs, such as LEM-domain proteins (see Barton et al., this issue), localize specifically at the inner nuclear membrane (INM) and influence chromatin [3,4]. Other INM NETs, such as SUN-domain proteins, bind within the NE lumen to NETs known as Nesprins in the outer nuclear membrane (ONM), together forming complexes that Link the Nucleoskeleton and Cytoskeleton (LINC) [5,6]. LINC complexes have major roles in mechanical force transduction to the nucleus [7–9]. Nuclear pore complexes (NPCs), in addition to their canonical roles in nucleocytoplasmic exchange, also have tissue-specific roles [10–13]. For example, specific nucleoporins anchor dynein-dependent movement of the nucleus in developing neurons (see Razafsky and Hodzic, this issue).

Interest in tissue-specific NE proteins arose from the discovery of tissue-specific human diseases known as ‘laminopathies’, which posed a paradox: A-type lamins (encoded by LMNA) are expressed in most differentiated tissues, yet mutations in LMNA can selectively affect striated muscle or cause Hutchinson-Gilford progeria syndrome, restrictive dermopathy, Dunnigan-type familial partial lipodystrophy or peripheral neuropathy [14,15]. These distinct pathologies might be explained by selective disruption of tissue-specific partners for A-type lamins. Indeed several lipodystrophy-causing mutations lead to amino acid substitutions mapping to the surface of the Ig-like fold in the tail domain of A-type lamins [16–18]. Other tissue-specific partners might be affected by overexpression of lamin B1 [19] or by mutations in genes coding for widely expressed NETs linked to muscular dystrophy (emerin, nesprin, SUN, LUMA) [20–23], osteopoikilosis (MAN1) [24] or Pelger-Huet anomaly/HEM/Greenberg skeletal dysplasia (LBR) [25,26]. Supporting this concept, loss of the widely expressed LEM-domain protein MAN1 in Drosophila causes tissue-specific defects [27]. Indeed, certain widely expressed NETs have tissue-specific splice variants. For example the LEM-domain protein LAP2/TMPO has at least six experimentally confirmed splice variants [28,29] with developmentally regulated expression in Xenopus [30]. Similarly many Nesprin splice variants are expressed preferentially in specific tissues or during development [31,32]; myogenesis favors shorter splice variants [33], and ovaries express a tissue-specific nesprin-2 epsilon isoform [34]. Significantly expanding the concept of tissue-specificity, a host of potentially disease-relevant new NE membrane proteins is emerging from the field of NE proteomics as discussed below.

Tissue-specific NETs

A- and B-type lamins (LMNA, LMNB1, LMNB2) are differentially expressed in specific cell types with a wide range of functional consequences [35,36]. However this diversity pales in comparison to the number of distinct NET proteins identified so far, starting 15 years ago with a liver-specific NET named UNCL [37]. Since then, several hundred NETs have been identified in multiple proteomic studies [2,38–41], with at least 93 confirmed experimentally (Table 1). Three of these proteomic studies used identical methods to isolate NEs from peripheral blood leukocytes, muscle and liver; this strategy enabled comparisons that revealed only ~15% of NET proteins were shared [2,39,41]. Thus, the expression of most NETs is limited to specific tissues. Other tissue-specific NE proteins that do not have transmembrane spans have also been identified individually as binding partners for lamins in specific tissues such as heart [42].

Table 1. Summary of validated NETs.

Summary of NETs with validated localization at the NE, including tissue-preferential expression and known functions. Tissue specificity is based on comparison of peptide recovery and spectral counts from liver, muscle and leukocyte NEs in [2]. Any NET for which one or more peptide spectra were identified per mass spectroscopy run in a tissue, was considered normally expressed in that tissue, even if it was enriched in another tissue. NETs with 0–1 peptide spectrum per run in specific tissue(s) might be contaminants; in this case AND if another tissue(s) had >10 times more spectra per run, this NET was considered enriched in the other tissue(s). NE targeting of most NETs was confirmed by exogenous expression of tagged constructs; other NETs were localized by antibodies and electron microscopy. NETs annotated ONM or INM were localized by super resolution microscopy. ‘T’ indicates localization at the NE based on resistance to pre-fixation extraction with triton X-100, a classical assay for protein interaction with the nuclear lamina; cytoskeletal-associated ONM proteins may also resist detergent extraction. Not all NETs were tested by triton-resistance. ER, endoplasmic reticulum; PM, plasma membrane; Mito, mitochondria; CID, cardiac intercalated discs. Certain NETs are also listed as ER-localized, in most cases, if the number of spectral counts in NE fractions was <4 times that in microsome fractions. Asterisks indicate NETs that were confirmed (NET46; NET19), but their original hypothetical ORF designation was removed from the NCBI database. References pertain to NE targeting: for functional references see [1].

Tissue-Specificity	Targeting		Functions
tissue name	INM ONM	other subcellular localizations	Chromatin	Cytoskeleton	Cell Cycle	Signaling	Disease	References
liver, blood, and muscle
P0M121	PoM							141
gp210	PoM							142
NET3, NDC1	PoM							40
LBR	INM							143
LAP1	INM							144
LAP2	INM							145
emerin	INM	CID, PM, ER						50
MAN1	INM							146
SUN1	INM							38
SUN2	INM							147
nesprinl	ONM							148
nesprin2	ONM							149
LUMA								38
NET13								45
C20orf3	T	ER						39
TMEM66								ur
Tmeml94	INM							41
RHBDD1								41
AADACL1	INM	ER						39
METTL7A	INM	ER						39
DSCD75		ER						ur
NET8	T	ER						40
NET32	T							150
NET24	ONM	ER						45
STT3A	INM	ER						39
NET5, Sampl	INM							94
FLJ20254	INM	ER						41
NET25	T							151
NET26								40
NET53, nesprin3	T							152
NET56, Dullard	T	ER						40
NET31	ONM							40
MBOAT5	INM							41
liver enriched
UNC-50								37
NET4	ONM	ER						40
NET33	INM	ER						45
NET39	INM							45
EGFR		ER						45
NET34	INM	ER						45
NET11		ER						45
NET47	INM	ER						45
NET55	INM							45
NET62								45
NET46*	INM							45
muscle enriched
TRIC-A	INM	ER						41
WFS1		ER						41
POPDC2		ER						41
VMA21		ER						41
KLHL31								41
ATP1B4								153
ATP1B1		ER						154
RYR1		ER						155
blood enriched
HVCN1	T							ur
LOC55848								ur
SLC38A10		ER						ur
LRRC8A		ER						ur
FAM3C	T	ER						ur
ABCB1								156
NHE-1		ER						157
InsP 3R		ER						158
NET23, STING	T	ER, PM, Mito						45
LOC84233		ER						39
LOC79415	INM							39
LOC29058	T	ER						ur
SECll-like 3	INM							39
MARCHV	T	ER						39
NET20	INM							45
NET50	INM							45
AYTL1								ur
NET45		ER						45
glandular tissues enriched
nesprin4								87
testis enriched
SUN3								159
Spag4L, SUN4								160
early development enriched
NET19*								161
fat enriched
NET29A	INM							45
NET29B	T							84
migrating P cells, hyp7 precursors and intestinal primordium
Unc-83								44
liver and blood enriched
NET59, nicalin	INM	ER						45
SQSTM1	T							ur
IAG2	INM	ER						39
TAPBPL	T	ER						39
TMUB1		ER						ur
Tmeml99	T							39
NET51	INM	ER						40
NET9	T	ER						150
NET38	INM	ER						ur
liver and muscle enriched
NET37	INM	ER						150
TMTC3	T	ER						ur
NET30	INM	ER						45
blood and muscle enriched
LOC54968		Mito						41
CKAP4								41
TMEM41A	INM							39
TMEM109	T							39
nurim	INM							162
Non-transmembrane tissue-specific NE proteins
NPC core components (nucleoporins)
BS-63								60
Nupl33								58
Nupl55								59
Nup358, RANBP2								57
Nup50		nucleoplasmic						163
Lamins
Lamin C2								164
Lamin B3								165
heart enriched
MLIP								42

The impact of these newly identified NETs on current thinking about functional diversity at the NE is illustrated in Figure 1. Each plot shows the gene expression distribution across a range of tissues for the NETs isolated from either blood leukocytes [39] (Figure 1A), liver [2] (Figure 1B) or skeletal muscle [41] (Figure 1C). For each tissue only NETs enriched at the NE compared to the ER (an expected contaminant) are shown. Human transcriptome data from BioGPS [43] was transformed to z-scores (a measurement of distance from the mean in standard deviation units) across tissues, and then plotted as a heatmap. Many NETs identified in each tissue were expressed preferentially in that tissue (Figure 1). Some NETs were also expressed in other tissues: for example a splattering of NETs identified in blood leukocytes were also expressed variously in muscle, heart, liver, thyroid, and/or prostate (Figure 1A). For liver the NETs overlapping with other tissues clearly fell into distinct sets for different tissues. For example, a cluster of overlap (red) is observed with skeletal muscle that is clearly distinct from a cluster with thyroid and prostate (Figure 1B). Not surprisingly, many NETs identified in skeletal muscle are also expressed in heart and cardiac myocytes (Figure 1C). Collectively, very few of these NETs are expressed in the pancreas, kidney, brain, ovary or skin, identifying these organs as potentially rich sources of new undiscovered NETs. Among the identified NETs that were also expressed in liver, lung or brain, many were not expressed in fetal versions of those tissues. These findings collectively highlight a major new concept — the high level of tissue-specific or tissue-restricted expression among NE membrane proteins — and reveal an open frontier in biomedical research to define the NET proteomes of important tissues including the brain, pancreas, ovary, kidney and skin.

Figure 1. Tissue specificity of the NE proteome.

Gene expression distribution across a range of tissues, ‘heat-mapped’ for genes encoding NE transmembrane proteins (NETs) identified in proteomic studies of either (A) white blood cells (‘blood leukocytes’) [39], (B) liver [2] or (C) skeletal muscle [41]. Only proteins that were enriched in the NE preparations over an equivalently prepared microsome fraction (5-fold based on peptide spectral counts) are included. This criterion increases the certainty of NET designations, sometimes problematic because the ER is continuous with the ONM and ER proteins are hence legitimate components of NE preparations. Each proteomic dataset was correlated with human transcriptome data from BioGPS [43], transformed to z-scores (a measurement of distance from the mean in standard deviation units) across tissues, and then plotted as a heat-map. Colors indicate probes whose signal was above (red) or below (blue) the average, respectively, for particular tissues. The datasets in A–C were plotted separately but scaled to the same height for presentation.

Tissue-specific import or localization at the NE

Tissue-specificity can also be achieved by selective targeting to the NE. This could be as simple as expressing a tissue-specific partner involved in retention at the INM. For example, C. elegans Unc83 only localizes at the NE when Unc84 is present, though here Unc84 is widely expressed [44]. Several NETs fail to localize at the NE when expressed exogenously in fibroblasts, but target the NE in more differentiated or appropriately specialized cells [45]. WFS1 targets almost exclusively to the NE in muscle [41] but localizes primarily in the ER in other tissues [46]. Indeed, some NETs have variable localizations: LUMA also co-localizes with adherens junctions and myocardial intercalated discs [38,47], and emerin also localizes in the ONM and ER and intercalated discs of cardiomyocytes [48–53]. Differential control of NET protein localization could be achieved by mechanisms ranging from tissue-specific partners to tissue-specific posttranslational modifications. The latter mechanism (differential regulation), interestingly, might be influenced by the tissue-specific control of nucleocytoplasmic transport.

The structure and composition of NPCs, long thought to be uniform [11–13], can vary: about one-third of NPCs have variable subunit stoichiometry [54], and specific nucleoporins show tissue-specific differences in expression. For example the integral membrane nucleoporin gp210, though typically present at NPCs, is absent in certain tissues but is expressed preferentially in muscle [55]. Experimental reduction of gp210 expression during early development specifically affects muscle and neuronal differentiation [56]. Levels of Nup358/RanBP2 increase during myogenesis and alter the architecture of these cytoplasmic NPC filaments as visualized by atomic force microscopy (AFM) [57]. It is tempting to speculate that these changes influence nuclear import or reinforce NPC connections to the cytoskeleton in muscle. Defects in Nup133 cause specific deficits in neural development [58], while mutation of Nup155 in humans or its loss in mice causes atrial fibrillation and sudden death, suggesting a cardiac-specific function of this NPC component [59]. There are also tissue-specific variants of Nup358/RanBP2, POM121 and Nup98/96 [60–62]. Interestingly certain novel tissue-specific NE components identified by proteomics have features characteristic of nucleoporins [63–65]; it will be interesting to determine if they are tissue-specific contributors to NPC function.

The two faces of the NE — tissue-specific impacts on chromatin and the cytoskeleton

Lamins and widely expressed NETs such as LAP2β, LBR and emerin interact with transcriptionally silent chromatin [3,4,66]. For example, LBR binds heterochromatin protein 1 [67]. This interaction helps stabilize peripheral heterochromatin, since LBR knockout in combination with lamin A knockout disrupts and in some cell types completely inverts the spatial distribution of heterochromatin in tissues from nocturnal mammals [68]. NET23/STING, which contributes to the state of chromatin compaction, is also widely expressed, but its levels vary considerably between different cell types. Accordingly, the level of expression of NET23/STING roughly correlates with the extent of chromatin compaction in the various cell types examined [69]. SUN2 binds the chromatin remodeler BRD4 [70]. LAP2β binds histone deacetylase 3 (HDAC3) [71] and, in co-complex with lamin B1, influences the positioning of several lymphocyte-specific genes [72]. Another such complex may involve LAP1, sumoylation of which influences HDAC4-dependent transcriptional repression of the Cox-2 locus [73]. Emerin, like LAP2, is a ubiquitous NET that can associate with ubiquitous repressors such as HDAC3 or germ cell-less (GCL) [71,74–76]. However, despite GCL being ubiquitous, its reduction in Drosophila and in humans has a tissue-specific effect on the germ line [77,78]. Emerin also interacts with Lmo7, which is relatively tissue-specific [79]. Thus, even widely expressed NETs have the potential to co-function with tissue-specific proteins or respond to tissue-specific control.

There are now spectacular examples of tissue-specific 3D spatial organization of specific chromosomes and individual genes in the nucleus (reviewed in [3,4,80]). For example, in myoblasts the MyoD locus is maintained near the NE in a zone inaccessible to the TAF3 subunit of the TFIID activating complex; during myotube differentiation the MyoD locus moves to the nuclear interior where it can be activated [81]. Intriguingly this level of regulation can operate in both directions: light stimulation of plant (Arabidopsis) cells repositions the CAB locus from the nuclear interior to the periphery where it becomes active [82]. While some widely expressed NETs contribute to gene repositioning [72], the tissue-specific positioning of genes and chromosomes relative to the NE can also require individual tissue-specific NETs [39,83]. For example, the tendency of chromosome 5 to localize near the NE in liver cells depends on NET45 and NET47 (Figure 2A). NET47 shows extremely preferential expression in liver; NET45 is also expressed preferentially in liver, at levels 50-fold (mRNA) and 20-fold (protein) higher than leukocytes; this enrichment is not reflected in Table 1 due to higher NET45 peptide levels in ER/microsomes [83]. Another NET that affected chromosome positioning, NET29/TMEM120A, is specifically important during adipocyte differentiation [84]. Dynamic ‘tethering’ of silent chromatin, and the roles of individual NETs, are major frontiers in understanding gene regulation and we postulate that tissue-specific NETs will influence the spatial positioning of many developmentally important genes [3,4].

Figure 2. Tissue-specific functions of NETs.

(A) NET45 and NET47, which are highly expressed in liver, function to maintain chromosome 5 near the NE specifically in liver cells. HepG2 liver cells transfected with a control siRNA or siRNA-downregulated for both NET45 and NET47 were visualized by staining DNA with DAPI (blue) and by hybridization with probes that ‘painted’ chromosome 5 (green) as in [83]. Chromosome 5 loses its normal peripheral localization when both liver NETs are absent. (B) Although most NETs are excluded from the base of the microtubule spindles in mitosis when the NE has broken down, NET5/Samp1 and Tmem214 are not [41,94]. Arrowheads indicate NET accumulation around the spindle poles. NET5/Samp1 is also important for the tight centrosome association with the NE; other tissue-specific NETs might also influence cytoskeletal organization or cell polarity.

Tissue-specific diversity is also a feature of LINC complexes [85] and their multifunctional association with different cytoskeletal components [86]. The four Nesprin genes and two Sun-domain genes give rise to multiple isoforms controlled, in part, by tissue-specific alternative mRNA splicing [31], or in the case of Nesprin4 by restricting transcription to limited cell types [87]. LINC complex functions are further modulated by tissue-specific partners. For example NET5/Samp1, which has several isoforms and was expressed in only 3 of 11 tissues tested [83], and the widely expressed diaphanous formin, FHOD1, both contribute to actin-dependent nuclear migration [88,89] mediated by transmembrane actin-associated nuclear (TAN) lines, which include the ‘giant’ isoform of nesprin-2 (nesprin-2G) and SUN2 [90–92]. NET5/Samp1 also interacts with lamin A and SUN1 [93]. NET5/Samp1 contributes to centrosome positioning during interphase, and accumulates at the spindle base during mitosis [94] (Figure 2B). Centrosome positioning is critical to orient the nucleus and cytoskeleton in highly polarized cell types such as neurons, polarized epithelia or during antigen presentation in immune cells. Certain novel tissue-specific NETs also accumulate at the spindle base (Figure 2B), or co-localize with microtubules at the nuclear surface [41]. The unexpectedly large number of novel partners for SUN-domain proteins discovered in plants suggests an even wider range of functions [95].

Distinct partners for LINC complexes may also explain their ability to interact with high-tension versus low-tension actin filaments at the NE [8] and influence mechanotransduction signalling in specific cell types [7–9]. The transcription factor megakaryoblastic leukaemia 1, which responds to mechanical stress [96], is functionally impaired in cells from Lmna^−/− mice suggesting it functions downstream of LINC- and lamin A-dependent force transduction [97]. Other NETs involved in signalling cascades include emerin (e.g., β-catenin signalling) [98], MAN1 (e.g., rSmad-mediated and TGF-β signalling) [24,99–101] and NET25/LEM2 (e.g., ERK signalling) [102]. Interestingly, NET25/LEM2 is widely expressed but is strongly induced in muscle; consistent with this pattern, heterozygous knockout mice have a muscle-specific defect whereas homozygous knockouts have gross developmental defects across tissues [103]. MAP kinase signalling is affected by loss of emerin [104] or lamin A/C [105]— another layer of tissue-specific control involving NE proteins. Finally, signalling is influenced by tissue-specific NETs. For example NET39 is liver-enriched (based on 5-fold enrichment over the ER/microsome fraction in Table 1), but is also expressed in muscle where it affects mTOR signaling [106]. Similarly the blood and liver-enriched NET45/DAK is present and affects dsRNA innate antiviral signaling in tissue culture cells [107].

The NE as a tissue-specific interface

NE proteins– ubiquitous and tissue-specific— collectively define the functionality of the nucleus in each cell type and may be particularly critical in certain tissues or at specific times. For example, myogenic differentiation requires LINC complexes [7], and lamin A variants linked to striated muscle disease have defects in LINC complex-dependent nuclear positioning [108]. LINC complexes are also particularly important for nuclear migration during retinal development [109,110], centrosome function during neurogenesis [111] and glutamate receptor density at neuromuscular junctions [112], and may co-function with lamin B2 during neuronal migration [113].

The ratios of specific A- and B-type lamins are major determinants of cell type-specific mechanics that dictate whether differentiated blood cells can exit the bone marrow [114]. Not surprisingly, the nucleoskeleton is dramatically reorganized during adipogenesis [115]. Specific lamins also support the function of specific transcription regulators, including the muscle transcription factor myoD [116,117] and the cell cycle and differentiation regulator pRb [118–120]. Several tissue-specific NETs also affect cell cycle regulation [121], a potentially novel contribution to tissue-specific pathology.

In one case, the mechanism of tissue-specificity is clear: nesprin4 is expressed in very few tissues including the cochlea, and nesprin4 gene mutations cause high frequency hearing loss [87,122]. However, most NE-linked diseases are caused by mutations in genes encoding widely expressed proteins. These proteins might disrupt binding to a tissue-specific partner(s), leading to pathology. The concept that NE proteins form functional complexes that may include tissue-specific partners is underscored by human genetics, since EDMD-like phenotypes can be caused by mutations in the genes encoding either lamin A/C, emerin, nesprin1, nesprin2, SUN1, SUN2 or FLH1, all of which interact directly or indirectly [20,22,23,123,124]. The FLH1 and nesprin disease variants are both linked to expression of a specific splice variant [23,124]. LINC complexes also appear to function tissue-specifically. For example, nesprin interactions with desmin, an important muscle-specific intermediate filament protein, may underlie its clinical manifestation as muscular dystrophy [125]. Alternatively, the nesprin ANC-1 interacts with the signaling protein RPM-1 in C. elegans brains and then influences β-catenin signaling during neural development [126]. In contrast, the mammalian nesprin-2 directly interacts with α-catenin and is a positive regulator of Wnt signaling in keratinocytes, COS-7 and HaCaT cells [127]. LINC complex interactions with dystroglycan are important for maintaining the positioning and connectivity of lumbar neurons to avoid breaking their connections to the muscle during muscle contraction [128]. The NETs LAP1 and emerin also interact; depleting either one causes muscular dystrophy while depleting both generates a more severe phenotype [129].

Novel NETs implicated in tissue-specific disease

Several tissue-specific NETs are linked to disease. Mutations in the genes encoding VMA21 and Tmem70, both identified in isolated muscle NEs, respectively cause human myopathy [130] and neonatal encephalocardiomyopathy [131]. WFS1 is preferentially expressed in muscle and retina [43] and mutations in its gene cause Wolfram syndrome, characterized by optic atrophy, deafness and/or diabetes [132–134]. Mutations in LRRC8A, encoding a leukocyte-specific NET, specifically inhibit B-cell development [135]. Other tissue-specific NETs show promise as disease candidates. For example the NET Tmem38A is expressed in both smooth and skeletal muscle [2,41] and its depletion in mice affects smooth muscle causing hypertension [136] and skeletal muscle where the explanted muscle exhibits stronger initial contractile force but rapidly fatigues [137]. Depletion of another muscle-specific NET, Popdc2, leads to mice with stress-induced bradycardia [138], suggesting roles in cardiac muscle pathology.

Future outlook

Tissue-specific NETs are exciting new pieces of the laminopathy puzzle. We estimate that over half of all NETs are still unidentified in mammals, and, as many are not conserved in evolution [1], most NETs are unidentified in lower organisms. The 598 identified mammalian NETs [2] are enriched in their tissues of origin and certain other tissues (Figure 1), whereas distinctive unexplored tissues such as skin and brain likely harbor many novel tissue-specific NETs. Furthermore, very little is known about most of the tissue-specific NETs identified thus far. Only ~10% of NETs identified by proteomics have been tested for NE targeting [2]. Unfortunately, domain or family resemblances have little predictive value, since several validated NETs appear to be splice variants of proteins that function in the cytoplasm, mitochondria or plasma membrane. For example the Na,K-ATPase beta m subunit, which normally functions in a larger complex as a Na⁺/K⁺ translocating ATPase at the surface of most cells, can also localize independently of the rest of the complex at the NE of neonatal skeletal muscle and C2C12 cells where it functions in transcriptional regulation [139]. Tissue-specific NETs may have arisen during evolution by amplification and adaptation of cytoplasmic proteins for secondary nuclear functions, as proposed [139]; alternatively, some ancestral proteins may have functioned at the NE as proposed for intermediate filaments [140]. What is clear is that tissue-specificity of the NE arises in many ways and makes this organelle critical for integrating genome function in specific cell types.