Poly(ADP-ribosylation) is present in murine sciatic nerve fibers and is altered in a Charcot-Marie-Tooth-1E neurodegenerative model

Introduction

Adherens junctions are protein complexes localized at intercellular junctions, characterized by the existence of a link to the actin cytoskeleton at their cytoplasmic face. The central molecules in these junctions are transmembrane proteins called cadherins. While extracellularly bound to a neighbor identical molecule, cadherins are intracellularly attached to several proteins that allow the anchorage of the actin microfilaments. In polarized epithelial cells, adherens junctions are grouped. They describe a band across the lateral cell faces which encircles the cell, called the zonula adherens or the adhesion belt, usually more basal than tight junctions (Alberts et al., 2002; Meng & Takeichi, 2009). Epithelial adherens junctions functions are not only structural but also regulatory, participating in cell signaling networks. For example, some adherens junctions proteins such as β-catenin are called NACos (Nuclear and Adherent junction Complex components). NACos can either localize at the adherens junctions or translocate to the nucleus where they act as transcription factors, allowing the coordination of the loss of cellular adhesion with cell rounding and mitosis entrance (Pérez-Moreno, Jamora & Fuchs, 2003; Cerejido, Contreras & Shoshani, 2004).

Clustered adherens junctions can also be found in other polarized cells. For example in the vertebrate peripheral nervous system, each axon is surrounded by Schwann cells and an extracellular matrix. These components altogether constitute the nervous fiber, characterized by longitudinal as well as radial polarity. Myelinating Schwann cells wrap around the axons of motor and sensory neurons to form the myelin sheath. E-cadherin autotypic junctions contribute to the overall stability of the Schwann cell, being located in all the autotypic junctions regions, namely the outer and inner mesaxons, the outer and inner loops, the paranodal regions and the Schmidt-Lanterman incisures (Fannon et al., 1995; Poliak et al., 2002).

Interestingly, Trembler-J (Tr-J) mice harbour a punctual missense mutation in peripheral myelin protein 22 (pmp-22) gene, constituting a model of hypomyelinating human peripheral neuropathy Charcot-Marie-Tooth-1E (CMT1-E, Li et al., 2013; formerly classified as CMT1-A; Valentijn et al., 1992). Tr-J mice display an increased number of Schwann cells (Robertson et al., 1997) as well as structural alterations in Schmidt Lanterman incisures and paranodes of their sciatic nerve fibers involving changes in autotypic adherens junctions components such as E-cadherin (Devaux & Scherer, 2005). Moreover, actin microfilaments abundance and distribution in sciatic nerve fibers are distorted (Kun et al., 2012a).

Poly-ADP-ribose (PAR) is a linear or branched polymer of up to 400 ADP-ribose units. It is synthesized by poly-ADP-ribose polymerases (PARPs) from NAD⁺ as a postranslational protein modification and catabolized mainly by the endo- or exo-glicosidic action of poly-ADP-ribose glycohydrolase (PARG) (Virag & Szabó, 2002). PAR can in turn interact non-covalently with different protein domains including PBZ (PAR binding zinc finger motif), WWE (with three conserved residues -tryptophans and glutamate-) and Macrodomain (Leung, 2014).

Excessive nuclear PARP and PAR occur in brains in the context of neurodegeneration for example in Alzheimer’s, Parkinson’s and Hungtington’s diseases (Love, Barber & Wilcock, 1999; Martire, Mosca & d’Erme, 2015; Vis et al., 2005; Cardinale et al., 2015). Besides, PAR quantity and/or PARP expression are altered in several other pathologies (Strosznajder, Jesko & Strosznajder, 2000; Virag & Szabó, 2002; Strosznajder, Jesko & Zambrzycka, 2005; Masutani, Nakagama & Sugimura, 2005; Lafon-Hughes et al., 2008; Cerboni et al., 2010; Strosznajder et al., 2012; Liu et al., 2014). Nevertheless, poly-ADP-ribosylation (PARylation) biology is still poorly understood.

Human PARP family has 17 members, 4 of which have PARylating activity. They are PARP-1, PARP-2, tankyrase-1 (TNKS-1) and tankyrase-2 (TNKS-2; Vyas et al., 2005). In spite of the fact that PARP-1 is the only member that is localized exclusively in the nucleus and PARG nuclear and cytoplasmic isoforms have been described, most PARylation studies are focalized in the cell nucleus. Nuclear PARPs regulate chromatin structure and somehow participate in nuclear networks regulating DNA replication, gene expression, DNA damage recognition and repair or telomere maintenance (Virag & Szabó, 2002). More recently, cytoplasmic roles of PARylation are being envisaged (Lehtio, Chi & Krauss, 2013; Vyas et al., 2013). For example, in epithelial cells TNKS-1 maps not only to the nucleus but also to endoplasmic reticulum, Golgi apparatus, secretion vesicles, lisosomes or epithelial lateral membrane. Moreover, TNKS-1 is recruited from the cytoplasm to the epithelial lateral membrane upon formation of E-cadherin-based adherens junctions (Yeh et al., 2006). Adherens junctions proteins vinculin and catenin have been recovered as PARylated proteins in immunoprecipitation experiments (Gagné et al., 2008; Gagné et al., 2012). Furthermore, VERO cells (green monkey renal epithelial cells) harbor a PAR belt associated to the epithelial adhesion belt which is synthesized during the cell–cell adhesion process. If actin polymerization is inhibited, the PAR belt is disaggregated. Conversely, if PAR belt synthesis is inhibited by the TNKS inhibitor XAV939, the actin cytoskeleton, cell shape and cell adhesion are altered, indicating that the PAR belt is structurally and functionally linked (directly or indirectly) to the actin cytoskeleton (Lafon-Hughes et al., 2014).

With the aim of contributing to understanding the biology of PARylation in the peripheral nervous system, we tested the following three hypothesis. First, as PAR exists associated to epithelial adherens junctions in VERO cells, it might be also found in peripheral nervous system Schwann cell non-compact myelin which is rich in autotypic adherens junctions. Second, as PAR is associated to the actin cytoskeleton in VERO cells and F-actin is highly increased in Tr-J sciatic nerve fibers, PAR might be more abundant in Tr-J than WT nerves. Third, as nuclear PAR increases have been described in central neuropathology, nuclear Schwann cell-PAR could be increased in neurodegenerative Tr-J peripheral nerve fibers.

In the present work, we have used immunohistofluorescence (IHF) and confocal microscopy to evidence for the first time the presence of PAR in Schmidt Lanterman incisures and paranode regions in WT sciatic nerves. The non compact myelin regions where PAR and actin colocalized mimicked E-cadherin well-described distribution. Besides, we demonstrated through filamentous actin (F-actin) and PAR signals quantification on cytoplasmic axonal and Schwann cell domains that Tr-J mice sciatic nerves have increased PAR.

Materials and Methods

Results

WT mice sciatic nerves contained epitopes recognized by BD anti-PAR antibodies, particularly evident in Schmidt Lanterman incisures, paranodes and outer loop regions

Nerve cryosections were subjected to indirect IHF with rabbit anti-PAR antibody as well as F-actin detection with phalloidin-543 probe and nuclear counterstaining with DAPI.

Throughout the work, figures illustrate fluorochrome signals corresponding to PAR (detected with BD anti-PAR antibody) in green, F-actin in red and DAPI in blue. Figures 1A and 1B show single confocal slices where PAR signals (absent in the control without primary antibody) can be seen.

For further analysis, colocalization was defined as the spatial overlap (in a single pixel) of two signals (red and green, correspondent to F-actin and PAR channels), being each of the signals above a certain intensity threshold (fixed arbitrarily in an experiment-specific way).

When pixels where F-actin and PAR signals colocalized were highlighted in yellowish/orange color, appealing images were obtained. The F-actin-PAR colocalization regions (overviewed in Fig. 1C) drew the known non-compact myelin regions in sciatic nerves where E-cadherin has been localized, namely Schmidt Lanterman incisures (Figs. 1D–1I), outer loops and paranodes (Figs. 1K–1P). The observed distribution of F actin-PAR is in close agreement to the E-cadherin distribution reported by Fannon and collaborators (see Figure 5 of Fannon et al., 1995).

ENZO anti-PAR antibody but not Tulip clone H10 anti-PAR antibody reproduced the signals obtained with BD anti-PAR antibody in WT and Tr-J sciatic nerves

ENZO mouse monoclonal antibody, raised against short to mid-PAR chains (2 to 50 units) showed a clean signal after direct IHF (see Fig. S2). In fact, ENZO anti-PAR depicted clearly the paranodes, Schmidt Lanterman incisures and outer loops (Figs. SA, Figs. S2B) in WT sciatic nerves. An overview shows again that Tr-J sciatic nerves, which harbor extra nuclei (Fig. S2C vs Fig. S2F) as well as F-actin increase (D vs G), do also depict PAR increase (E vs H). The same can be perceived comparing I (WT) and J (Tr-J) merged channels images. Thus, ENZO anti-PAR antibody qualitatively conduces to the same conclusions as BD anti-PAR antibody (and this has been confirmed even quantitatively in nuclear WT vs Tr-J signals; see the Figshare link on Table S1).

In contrast, Tulip H10-clone antibody, known to recognize preferentially long branched PAR chains above 20 residues (Kawamitsu et al., 1984), showed no signal at all although it was assayed both in WT and Tr-J sciatic nerves (data not shown).

Discussion

This study represents the first report of PAR presence and distribution in the peripheral nerve system. We demonstrated that WT mice sciatic nerves contained nuclear PAR as well as cytoplasmic PAR. The latter colocalized with F-actin at E-cadherin rich non-compact myelin regions including Schmidt Lanterman incisures and paranodes.

Like in VERO epithelial belt, PAR was detected with anti-PAR antibodies that are presumed to recognize short to middle chain length polymer (BD and ENZO anti-PAR antibodies). The finding of PAR associated to E-cadherin rich/actin anchorage regions in two different biological systems (a monkey epithelial cell line and mice nerves) suggests that PARylation may be involved in adherens junction biology. These findings have deep implications in terms of junctional structures and cell signaling.

It has been proposed that inhibition of axonal regeneration by myelin after injury is an aberrant effect of an otherwise physiological inhibitory function (Shen et al., 1998). Axon branching inhibition is essential for the structural economy of the peripheral nervous system, especially for long axons. One key mediator of these functions would be myelin-associated glycoprotein (MAG; Mukhopadhyay et al., 1994; Schafer et al., 1996; Shen et al., 1998). MAG is mainly found at the adaxonal side of the Schwann cell membrane and non-compact myelin regions rich in autotypic adherens junctions including Schmidt Lanterman inscisures and paranodes (Martini & Schachner, 1988; Ghabriel & Allt, 1980; Erb et al., 2006).

In WT mice we observed PAR at non-compact myelin regions. Besides, PAR has been involved in axonal regeneration inhibition (Brochier et al., 2015). Therefore, an hypothesis can be raised regarding the putative physiological role of the observed PAR in preventing axonal branching.

Through image quantification of immunohistofluorescent images, we have also shown that Tr-J sciatic nerve Schwann cells harbored excessive nuclear PAR. Interestingly, excessive nuclear PARP and PAR had already been reported in brains in the context of neurodegeneration, for example, in Alzheimer’s, Parkinson’s and Hungtington’s diseases (Love, Barber & Wilcock, 1999; Martire, Mosca & d’Erme, 2015; Vis et al., 2005; Cardinale et al., 2015). In fact, PARP-1-/- mice confirm the protective role of PARP-1 deficit towards injury induced by Aβ injections (that mimics Alzheimers disease), MPTP or 6-OHDA (used to induce Parkinson-like symptoms). Thus, the current prevailing interpretations are that different injuries induce PARP-1 overactivation leading to cell death by energy shortage due to NAD⁺ overconsumption and/or that DNA repair alterations (which may involve PARP-1 signaling) represent a common denominator in neurodegeneration (Martire, Mosca & d’Erme, 2015; Ross & Truant, 2016). However, in Tr-J, the absence of apoptotic morphology or long branched polymer (recognized by H10 clone anti-PAR antibody) suggest that the excessive nuclear PAR was not associated with a DNA repair response.

Our current work underscores the existence of a comparable increase of cytoplasmic PAR in a Charcot-Marie-Tooth model. Interestingly, the first scientific work that communicates an increase in human brain poly(ADP-ribosyl)ation with neurodegeneration (Love, Barber & Wilcock, 1999) is focused on the strong nuclear PARP and PAR signal in Alzheimer’s disease patient brains. Such work does also show the existence of cytoplasmic PARP and PAR immunoreactivity, which is not further remarked. Other authors measure a significant PARP activity increase in hippocampus homogenates in contexts of excitotoxicity or amyloid beta peptide presence or, conversely, study the effect of PARP inhibition with 3-aminobenzamide (3-AB) (Strosznajder, Jesko & Strosznajder, 2000; Strosznajder, Jesko & Zambrzycka, 2005). They also arrive at the conclusion that PARP has a central role in neurodegeneration. We agree with that conclusion but in our opinion it has to be emphasized that neither PARP activity measurements nor PARP inhibition with 3-AB are evidencing exclusively nuclear PARP or PARP-1 activity. These results could be reflecting the activity and the role of a pool of PARPs (including PARP-1, PARP-2, TNKS-1 and TNKS-2), some of which are cytoplasmic.

In contrast to WT, Tr-J sciatic nerve axons harbor a strong PAR signal. PAR presence in peripheral axons resembles the work of Brochier et al. (2015) in the central nervous system, since they detect PAR in crushed optic nerve axons. Of note, we did not see any obvious difference in PAR signal intensity at the nerve cutting ends (we skipped them for quantification purposes just in case). Brochier et al. (2015) do also demonstrate that axonal growth inhibiting signals such as Nogo, MAG or astrocyte-produced chondroitin sulfate proteoglycans (CSPGs), induce neuronal PAR accumulation in primary cortical neurons. Likewise, the increase of MAG expression in the peripheral nervous system under neurodegenerative conditions (Kinter et al., 2013) might be related to the high levels of PAR in the Tr-J nerve fibers.

About 80% of the basal PAR pool is conserved in PARP-1 ^−∕− primary cortical neurons (Brochier et al., 2015). In contrast, the induced PAR increase depends on PARP-1 activity. PARP-1 is a nuclear enzyme; just a small PARP-1 fraction has rarely been localized in the cytoplasm of cancer cells (Donizy et al., 2014). Therefore, in Brochier et al. experiments, either neuronal PARP-1 can somehow reach the cytoplasm under certain conditions or PARP-1 is part of a signaling cascade that activates a cytoplasmic PARP that in turn synthesizes the cytoplasmic PAR. Interestingly, the PARP inhibitors which restore the axonal growth that is diminished by Nogo, MAG or CSPGs are not PARP-1 specific. Moreover, according to the same authors (Brochier et al., 2015), in a microfluidic-based culture platform, most growth rate restoration by the PARP inhibitor is achieved through PJ34 administration in the axonal compartment, sugesting that most PJ34 effects can be reached through the inhibition of a cytoplasmic PARP. No increase in axonal regeneration nor improvement in motor function recovery were observed after optic nerve crush or dorsal hemisection of the spinal cord, in PARP-1 ^−∕− mice or after systemic administration of the specific PARP-1 inhibitor velaparib (Wang et al., 2016).

Disturbances in PMP22 are associated with abnormal myelination in a range of inherited peripheral neuropathies both in mice and humans (Robertson et al., 1997). PMP22 is critical for actin-mediated cellular functions (Lee et al., 2014). Interestingly, actin microfilaments assembly and disassembly is essential during myelin sheath formation in the peripheral and central nervous system (Park & Feltri, 2011; Feltri, Suter & Relvas, 2008; Nawaz et al., 2015; Zuchero et al., 2015). A lamellipodia-like structure driving myelin wrapping has been described in peripheral and central myelination process (Salzer, 2012; Feltri, Poitelon & Previtali, 2015; Nawaz et al., 2015; Zuchero et al., 2015). Besides, PMP22-deficient nerves depict early abnormal junctions and permeability of myelin (Guo et al., 2014). Tr-J mice sciatic nerve has marked alterations in junctional proteins including delocalized E-cadherin (Devaux & Scherer, 2005). In turn, cumulative evidence indicates that adherens junctions proteins (E-cadherin, catenins, vinculin), some of which can act as NACos, may play a significant role in the myelination process (Tricaud et al., 2005; Perrin-Tricaud, Rutishauser & Tricaud, 2007; Ye et al., 2009; Peng et al., 2010; Beppu et al., 2015; Basak et al., 2015). For example, E-cadherin enhances neuroregulin-1 (NRG1) signaling and promotes Schwann cell myelination (Basak et al., 2015). It has to be emphasized that NRG1 is one of the major and best characterized extrinsic signals that control myelination (Salzer, 2015). Conversely, an aberrant localization of E-cadherin can be a potent inhibitor of Wnt/ β-catenin (Su et al., 2015) and in mammalian CNS, dysregulation of the Wnt pathway inhibits timely myelination (Fancy et al., 2009; Dai et al., 2014).

Considering our previous results regarding the actin cytoskeleton/PAR belt dialogue in VERO cells (Lafon-Hughes et al., 2014) and the highly increased actin observed in the Tr-J sciatic nerve (Kun et al., 2012a), the observed PAR increase was an expected result. Moreover, PAR-digested regions seemed to harbor lower F-actin signals, suggesting that PAR might be somehow participating in the holding or anchorage of at least part of the F-actin cytoskeleton network. In the nerves, like in VERO cells, the actin-PAR connection seems to be active.

We don’t know if PARylation alterations are upstream actin and E-cadherin modifications. It is even unknown which proteins are PARylated in the sciatic nerve. Looking for putative cytoplasmic PARylation candidates –adherens junctions/actin anchorage proteins or microfilaments components-, we have inspected the macrodomain-recognized ADP-ribosylome of liver epithelium; catenin (CTNNA 3), vinculin, and β-actin are in the list of affinity-enriched proteins. As the macrodomain recognizes MAR and in some cases PAR (Martello et al., 2016), the method does not differentiate mono-ADP-rybosylated (MARylated) from PARylated proteins. Interestingly, this data fit with another proteomics work showing that in human HEK 293 embryonic kidney cells (Gagné et al., 2012), vinculin and catenin (β-catenin) were recovered in the pool of proteins bound to a catalytically inactive GFP-PARG or “PARG-DEAD domain” (indicating their probable PARylation or association to PARG) but not immunoprecipitated with clone H10 antibody (indicating that they do not bind long branched chains). Actin was not recovered in non-stimulated cells. Instead, β-actin was enriched in PARylated complexes in MNNG-treated cells, indicating its participation in the responses induced by this alkylating agent (Gagné et al., 2012). In our knowledge, a single group claimed that actin is “the unique endogenous acceptor of PAR” in an Octopus brain cytoplasmic subcellular fraction (De Maio et al., 2013).

Monomeric G-actin has been long recognized as a MARylating target of bacterial toxins, shifting the equilibrium towards microfilament disassembly (De Maio et al., 2013). Coherently, the knockdown of PARP-14, which harbors MARylating activity (Vyas et al., 2005) and is a focal adhesion protein, results in cells that are unable to retract protrussions efficiently, generating highly elongated extensions (Vyas et al., 2013). The knockdown of the macrodomain-containing enzymatically-inactive PARP-9 results in abnormal membrane blebbing in the absence of typical apoptotic nuclear DNA hypercondensation (Vyas et al., 2013), suggesting that PARP-9 is also involved in actin cytoskeleton dynamics. As PARP-9 is enzymatically inactive and PARP-14 has just MARylating activity (Vyas et al., 2005), neither of them could be responsible for the synthesis of the PAR that we observed in mice sciatic nerves. Nevertheless, the existence of some sort of regulatory interaction among PARP-9 or PARP-14 and the PARP responsible for PAR synthesis cannot be discarded.

The analogy with epithelial cells (Yeh et al., 2006; Lafon-Hughes et al., 2014), points to TNKS involvement in cytoplasmic PARylation. Mice deficient in either one TNKS are viable but deficiency of both TNKS results in embryonic lethality, demonstrating that TNKS are essential but at least partially redundant (Chiang et al., 2008). Accordingly, TNKS-2 knockdown cells display no detectable phenotype whereas TNKS-1 knockdown cells display mitotic defects and diminished viability (Vyas et al., 2013), probably preventing further dissection of the molecular mechanisms involved.

Interestingly, a study on structural basis and sequence rules for substrate recognition by TNKS identifies a TNKS-binding motif (Guettler et al., 2011) which is present in vinculin and catenin but not in β-actin.

TNKS is necessary for canonical Wnt signaling (Kartner et al., 2010). TNKS inhibition induces axin stabilization and blocks Wnt signaling (Bao et al., 2012). Besides, axin has been identified as a regulatory and therapeutic target in newborn brain injury and remyelination (Fancy et al., 2011). Moreover, anti-TNKS weapons promote myelination (Casaccia, 2012). To be more precise, the TNKS inhibitor XAV939 accelerates oligodendrocyte progenitor differentiation in cell cultures, improves myelination and remyelination (following hypoxia or lysolecithin) in ex vivo mouse cerebellar slice cultures and diminishes the demyelinating effects of lysolecithin in mice spinal cord in vivo (Fancy et al., 2011). It has been demonstrated that the TNKS PARylation target molecule axin is involved. It is likely that some of the adherens junctions/actin anchorage proteins are PARylation targets too, acting in concert with axin to coordinate cell adhesion and migration with differentiation.

Together our findings highlight the presence of PAR in specific regions of peripheral nerves and its increase in Tr-J, opening a window to further explore the possible roles of cytoplasmic PAR associated to adherens junctions/actin cytoskeleton in the whole nervous system. The advances in this field are expected to contribute in the future to a more precise design of therapies, not only for CMT patients, which represents 1 in every 2500 persons in the general population (Li et al., 2013), but also for other neurodegenerative disease patients.

Conclusions

• PAR was detected in mice sciatic nerves, colocalizing with F-actin at non-compact myelin regions of peripheral nerve fibers which are rich in adherens junctions. The existence of PAR in adherens junctions regions in systems as divergent as Schwann cells and VERO epithelial cells suggests that PAR may be a previously overlooked inherent component of the E-cadherin rich adherens junctions.

• PAR was in excess in a Charcot-Marie-Tooth demyelinating disease model in which sciatic nerves present an excess of F-actin. Like in VERO cells, our result argue in favor of a structural (direct or indirect) connection of PAR with F-actin that deserves further investigation.

• Nuclear PAR was present in WT and moderately increased in Tr-J Schwann cells, putatively affecting chromatin structure and functions.

Supplemental Information