Multiple TSC22D4 iso-/phospho-glycoforms display idiosyncratic subcellular localizations and interacting protein partners


  • Sonia Canterini,

    1. Department of Psychology, Section of Neuroscience, Pasteur Institute – Cenci Bolognetti Foundation and ‘Daniel Bovet’ Neurobiology Research Center, Sapienza University of Rome, Italy
    Search for more papers by this author
  • Valentina Carletti,

    1. Department of Psychology, Section of Neuroscience, Pasteur Institute – Cenci Bolognetti Foundation and ‘Daniel Bovet’ Neurobiology Research Center, Sapienza University of Rome, Italy
    Search for more papers by this author
  • Stefania Nusca,

    1. Department of Psychology, Section of Neuroscience, Pasteur Institute – Cenci Bolognetti Foundation and ‘Daniel Bovet’ Neurobiology Research Center, Sapienza University of Rome, Italy
    Search for more papers by this author
  • Franco Mangia,

    1. Department of Psychology, Section of Neuroscience, Pasteur Institute – Cenci Bolognetti Foundation and ‘Daniel Bovet’ Neurobiology Research Center, Sapienza University of Rome, Italy
    Search for more papers by this author
  • Maria Teresa Fiorenza

    Corresponding author
    • Department of Psychology, Section of Neuroscience, Pasteur Institute – Cenci Bolognetti Foundation and ‘Daniel Bovet’ Neurobiology Research Center, Sapienza University of Rome, Italy
    Search for more papers by this author


M. T. Fiorenza, Department of Psychology, Section of Neuroscience, Via dei Sardi 70, Sapienza University of Rome, Rome, Italy

Fax: +39 6 49917871

Tel: +39 6 49917869



Proteins of the TSC22 domain (TSC22D) family, including TSC22D1 and TSC22D4, play pivotal roles in cell proliferation, differentiation and apoptosis, interacting with other factors in a still largely unknown manner. This study explores this issue by biochemically characterizing various TSC22D4 forms (both iso- and glyco-phospho-, namely the splice variants 42 and 55 kDa and the post-translationally modified 67 and 72 kDa forms) and their subcellular localization and protein partners during cerebellar granule neuron (CGN) differentiation. The TSC22D4-42 form is mostly cytosolic, and is the only TSC22D4 form that associates with TSC22D1.2 in undifferentiated but not differentiated CGNs. In contrast, TSC22D4-55 is prominently associated with the nuclear matrix in differentiated but not undifferentiated CGNs. As for TSC22D4-67, it is localized in the cytosol and nuclei of undifferentiated CGNs and enters mitochondria of differentiated CGNs, associating with apoptosis-inducing factor. TSC22D4-72 is modified by O-linked beta-N-acetylglucosamine (O-GlcNAcylated) and phosphorylated and is always associated with chromatin irrespective of CGN differentiation. The various subcellular localization patterns and interacting protein partners of TSC22D4 forms during CGN differentiation suggest the existence of form-specific function(s) and provide a novel framework to further investigate the biological functions of TSC22D proteins.

Structured digital abstract


apoptosis-inducing factor


cerebellar granule neuron


TGFβ1-stimulated clone 22 domain


TSC22D4 is a protein belonging to the TGFβ1-stimulated clone 22 domain (TSC22D) family of tumor suppressors [1] that control multiple biological processes such as cell proliferation, differentiation, senescence, apoptosis and embryo development [2-10]. TSC22D proteins include widely expressed and evolutionarily conserved members, encoded by a single locus, bunched (bun), as short and long isoforms with different N-termini [11] in Drosophila melanogaster, and by four distinct loci (TSC22D1, TSC22D2, TSC22D3 and TSC22D4), in mammals [12]. To date, studies of the functional activities of mammalian TSC22D family members have mostly focused on TSC22D1, formerly known as TSC-22 [13], expressed with the long TSC22D1.1 (110 kDa) and short TSC22D1.2 (11 and 15 kDa) [12] transcript variants apparently playing opposite roles. Indeed, human and murine TSC22D1.1 and TSC22D1.2 proteins promote cell proliferation/survival and senescence/death, respectively [6, 14], probably by competing with each other for heterodimerization with TSC22D4 [6]. A similar situation is also observed in Drosophila, where the long BunA isoform promotes cell growth, whereas the short BunB and BunC isoforms antagonize such effects [11]. In addition to transcript variants, additional heterogeneity and functional complexity are also given to the TSC22D1 protein by post-translational modifications. Indeed, mouse mammary gland expresses two additional variants (18 and 20 kDa) that probably arise from post-translational modifications of TSC22D1.2 [14]. A high level of complexity of forms is also displayed by TSC22D4. In fact, we previously reported that mouse cerebellar granule neurons (CGNs) express multiple TSC22D4 forms (42, 55, 67 and 72 kDa), the subcellular localization of which varies during postnatal cerebellum development and CGN differentiation/physiological conditions [15, 16]. Among these forms, TSC22D4-42 and TSC22D4-55 arise from alternative splicing [17], whereas TSC22D4-67 and TSC22D4-72 aore probably derived from post-translational modification of TSC22D4-42 [15, 16]. Other than heterodimerization with TSC22D1 [6, 18], actual TSC22D4 functions are still undetermined. It was recently shown by two-cell hybrid assay that TSC22D4 is capable of heterodimerizing with apoptosis-inducing factor (AIF) [19], a mitochondrial protein that is essential for redox metabolism and represents the first identified component of the caspase-independent cell-death cascade [20, 21].

Here we have studied TSC22D4 forms by determining which post-translational modifications they are subjected to, the nuclear/cytoplasmic compartments to which they localize, and the protein partners that they associate with, depending on CGN differentiation. We show that different forms display specific and differentiation-dependent subcellular localizations, and may physically associate with TSC22D1.2 or AIF, suggesting a functional involvement in the regulation of CGN differentiation.


The TSC22D4 protein is post-translationally modified by O-glycosylation and phosphorylation

Transfection experiments using siRNA directed against Tsc22d4 mRNAs previously performed in our laboratory [15] demonstrated that all TSC22D4 forms derive from TSC22D4 mRNAs encoding TSC22D4-42/55, suggesting that they are the result of post-translational modification(s). To investigate this possibility, we performed a preliminary in silico analysis of the TSC22D4-42 amino acid sequence (see 'Experimental procedures' for details) (Fig. 1A), showing that several serine and threonine residues may be phosphorylated while others may be O-glycosylated. Moreover, 11 of these residues are modifiable by either phosphorylation or O-glycosylation, raising the possibility of so-called ‘yin yang’ control [22]. Only a single asparagine residue is susceptible to N-glycosylation.

Figure 1.

Post-translational TSC22D4 modifications. (A) In silico identification of TSC22D4 amino acid residues susceptible to O-glycosylation (open rectangles) and/or phosphorylation (closed rectangles). Asterisks indicate amino acid residues that are susceptible to ‘yin yang’ control, and the open circle indicates the single asparagine residue susceptible to N-glycosylation. (B) Total protein extracts from DIV7 CGNs were incubated with calf intestinal phosphatase (CIP), N-glycanase, or a mixture of O-glycosidases and processed for western blotting using antibody against TSC22D4. (C) Relative abundance (mean ± SEM of total pixels per band) of individual TSC22D4 forms obtained in three independent experiments performed as in (B). Solid bars, TSC22D4-72; dark gray bars, TSC22D4-67; light gray bars, TSC22D4-55; open bars, TSC22D4-42. Asterisks indicate significant difference (< 0.01) between TSC22D4 bands for digested versus undigested extracts. (D) Protein extracts from DIV7 CGNs were immunoprecipitated with anti-TSC22D4 or control IgG (IgG lanes), and immunocomplexes were revealed by western blotting with anti-TSC22D4 (left) or the anti-O-GlcNAc CT110.6 antibody, either directly or after a β-elimination reaction (middle). Protein degradation caused by β-elimination was ruled out by western blotting with antibody against TSC22D4 (right). The arrow indicates the IgG heavy chain.

The occurrence of Ser/Thr phosphorylation in TSC22D4 forms was assessed by treating protein extracts of CGNs cultured in vitro for 7 days (DIV7) with calf intestinal phosphatase, followed by SDS/PAGE fractionation and western blotting. Treatment with calf intestinal phosphatase caused the 72 kDa band to strongly diminish/disappear and the 67 kDa band to correspondingly increase (Fig. 1B,C), indicating that the latter band is the unphosphorylated precursor of the former. The possibility of TSC22D4 post-translational modification by N-glycosylation or O-glycosylation was then directly addressed by treating CGN protein extracts with glycosidases specific to N-linked or O-linked carbohydrate groups. Extract incubation with N-glycanase had no apparent effect on any band, indicating the absence of N-linked glycosylation, as predicted by in silico analysis. By contrast, extract treatment with a mixture of O-glycosidases, including β-N-acetylglucosaminidase, resulted in significant reduction of the 67 kDa band and a corresponding increase of the 42 kDa band (Fig. 1B,C). The 55 and 72 kDa bands were apparently unaffected, suggesting that, at least for the 72 kDa band, phosphorylation protects the protein sugar moieties from enzymatic removal. Protein extract digestion using an O-glycosidase mixture lacking β-N-acetylglucosaminidase had no effect on the 67 kDa band (data not shown), indicating that most sugar residues of this band were β-linked N-acetylglucosamine (O-GlcNAc). Additional evidence for O-GlcNAcylation was also provided by using an antibody against O-GlcNAc (clone CTD110.6) or by using a chemical β-elimination reaction, a treatment that specifically removes O-GlcNAc residues [23, 24]. To this end, TSC22D4 was immunoprecipitated from DIV7 CGNs using anti-TSC22D4 antibodies, then fractionated by SDS/PAGE and blotted to membranes. The membranes were then probed with either the antibody against TSC22D4 (Fig. 1D, left) or the CTD110.6 antibody, either directly or following a β-elimination reaction (Fig. 1D, middle). Both the 72 and 67 kDa bands, but not the 55 and 42 kDa bands, appear to be O-GlcNAcylated based on positive immunostaining with anti-O-GlcNAc antibody and negative immunostaining with this antibody after the β-elimination reaction. The possibility of protein degradation in the 72 and 67 kDa bands caused by the β-elimination reaction was ruled out by immunostaining with anti-TSC22D4 antibodies (Fig. 1D, right). The results for positive and negative controls for the specificity of anti-TSC22D4 antibodies used in these experiments are shown in Fig. S1.

TSC22D4-72 and TSC22D4-55 forms display a chromatin and nuclear matrix localization, respectively

Because in vivo and in vitro CGN differentiation is accompanied by progressive TSC22D4 reduction in the nucleus and a relative increase in the cytoplasm [15], we studied the subcellular distribution of TSC22D4 forms during CGN differentiation. To this end, cytoplasmic and nuclear fractions from in vitro cultured undifferentiated (DIV1) and differentiated (DIV7) CGNs were probed with anti-TSC22D4 antibodies in western blotting experiments (Fig. 2A,B). TSC22D4-42 was always localized in the cytoplasm irrespective of the CGN differentiation stage. In contrast, TSC22D4-67 was quite abundant in DIV1 CGN nuclei, but disappeared from this compartment at DIV7, with a corresponding increase in TSC22D4-72, indicating that TSC22D4 phosphorylation is a prominent feature of differentiated CGN nuclei. In contrast, the localization/abundance of TSC22D4-55 did not vary significantly during CGN differentiation.

Figure 2.

Expression and sub-nuclear localization of TSC22D4 forms during in vitro CGN differentiation. (A) Western blotting of TSC22D4 forms in whole lysates, cytoplasmic (Cytp) and nuclear fractions (Nuclei) of DIV1 and DIV7 CGNs. Fractions are identified using histone H1 (nuclear fraction) and βIII tubulin (cytoplasmic fraction). Nuclear fraction purity is also indicated using Sp1. (B) Relative abundance of cytoplasmic versus nuclear TSC22D4 forms (mean ± SEM), obtained in 3–5 independent experiments performed as in (A). Pixels of individual bands were normalized to Sp1 and βIII tubulin and expressed as fractions of total pixels/band in the lane. Solid bars, TSC22D4-72; dark gray bars, TSC22D4-67; light gray bars, TSC22D4-55; open bars, TSC22D4-42. Asterisks indicate a significant difference between TSC22D4 forms of DIV7 versus DIV1 (*< 0.05; **< 0.001). (C) DIV1 and DIV7 CGNs were processed for preparation of chromatin- and nuclear matrix-associated TSC22D4 forms, and equivalent amounts of each fraction were analyzed by western blotting with anti-TSC22D4 antibodies. Fractions were identified using histone H1 (chromatin) and Lamin B1 (matrix). The ratios between pixels for the 72 kDa TSC22D4 and histone H1 bands were 0.98 ± 0.15 at DIV1 and 0.57 ± 0.11 at DIV7.

The significant variation of nuclear TSC22D4 abundance during CGN differentiation prompted us to investigate whether this was related to a specific sub-nuclear localization of TSC22D4 forms (Fig. 2C). To this end, lysates of DIV1/DIV7 CGNs were biochemically fractionated into chromatin (fraction marker histone H1) and nuclear matrix (fraction marker Lamin B1). The TSC22D4 forms present in these sub-nuclear compartments were then determined by SDS/PAGE and western blotting. The chromatin fraction was marked by the constant presence of TSC22D4-72, irrespectively of CGN differentiation stage, while other TSC22D4 forms were apparently absent. For the nuclear matrix, no TSC22D4 forms were detectable in DIV1 CGNs, but TSC22D4-55 and TSC22D4-42 were consistently observed in DIV7 CGNs.

TSC22D4-67 is localized in mitochondria and physically interacts with AIF in differentiated CGNs

In light of the punctate immunofluorescence staining of TSC22D4 in DIV7 CGNs [15] suggesting mitochondrial localization, we performed western blotting of purified cytosolic and mitochondrial fractions from adult mouse cerebellum, showing that the mitochondria contain all TSC22D4 forms but that TSC22D4-72 and TSC22D4-67 were the most abundant forms in both fractions (Fig. 3A). By contrast, a similar analysis of cytosol and mitochondria from PN3 mouse cerebellum showed that TSC22D4 forms were barely detected in the mitochondrial preparations at this developmental stage (Fig. S2). Although highly enriched, the mitochondrial fraction still showed limited cytosolic contamination (6–8%, evaluated by the presence of glyceraldehyde-3-phosphate dehydrogenase) and/or other cytoplasmic components. Sub-mitochondrial TSC22D4 localization in adult cerebellum was then investigated by preparing pure mitochondrial components, namely the outer mitochondrial membrane, the inter-membrane space, the inner mitochondrial membrane and the mitochondrial matrix [25]. Western blotting assays of equivalent amounts of mitochondrial sub-fractions derived from the same preparation (Fig. 3B) showed that TSC22D4-67 was mostly localized at the level of the inner mitochondrial membrane, and, to a lesser extent (~ 30% of the total mitochondrial TSC22D4-67), in the outer mitochondrial membrane, closely paralleling the sub-mitochondrial localization of AIF [26].

Figure 3.

TSC22D4-67 sub-cytoplasmic localization and association with AIF. (A) Western blotting of TSC22D4 forms in whole lysate, cytosolic and mitochondrial fractions from adult cerebellum. Fraction purity is indicated using glyceraldehyde-3-phosphate dehydrogenase (GAPDH, cytosol) and AIF (mitochondria). (B) Adult cerebellar mitochondria were fractionated into outer mitochondrial membrane (OM, marker Bcl-2), inter-membrane space (IMS, marker cytochrome c), inner mitochondrial membrane (IM, marker AIF) and mitochondrial matrix (MX, marker Hsp60), and processed for western blotting with anti-TSC22D4 antibody. The relative abundances of TSC22D4-67 in the outer mitochondrial membrane and inner mitochondrial membrane were 27.2 ± 1.3% and 72.8 ± 1.1%, respectively, and those of AIF were 28.9 ± 0.9% and 71.1 ± 1.2%, respectively. (C) Adult cerebellar mitochondria were immunoprecipitated (IP) with anti-AIF or anti-TSC22D4 antibodies, and immunocomplexes were revealed by western blotting (WB) with anti-TSC22D4 or anti-AIF antibodies, respectively. (D) Total protein extracts from DIV1 and DIV7 CGNs were processed as described in (C). Arrows indicate the IgG heavy chain.

The similar TSC22D4 and AIF distributions in mitochondrial membranes prompted us to determine whether TSC22D4 interacts with AIF in vivo, as previously suggested by pull-down and two-hybrid approaches [19]. To this end, immunoprecipitation experiments were performed on adult cerebellum mitochondria and DIV7 CGNs using either anti-AIF or anti-TSC22D4 antibodies, and the immunocomplexes were analyzed by western blotting using anti-TSC22D4 or anti-AIF antibodies (Fig. 3C,D). It appeared that the mature 62 kDa AIF form [21] specifically interacts with TSC22D4-67 in adult cerebellum mitochondria and DIV7 CGNs (Fig. 3C,D), but this interaction was not observed in DIV1 CGNs (Fig. 3D). Antibodies were validated for their efficacy in immunoprecipitation assays as shown in Fig. S3.

The 42 kDa TSC22D4 form associates with TSC22D1.2

The existence of TSC22D4–TSC22D1 heterodimers in cell lines [6, 18] prompted us to investigate by immunoprecipitation assays whether this interaction is mediated by a specific TSC22D4 form(s) in isolated CGNs. We preliminarily investigated TSC22D1.1/2 expression during CGN differentiation, observing a strong decrease in TSC22D1.2 isoforms in DIV7 CGNs (Fig. S4). Immunoprecipitation assays were then performed by incubating protein extracts from DIV1/DIV7 CGNs with either anti-TSC22D4 or anti-TSC22D1 antibodies, and then revealing immunocomplexes by western blotting with anti-TSC22D1 or anti-TSC22D4 antibodies (Fig. 4). TSC22D4-42 (but not the other TSC22D4 forms) appeared to be specifically associated with the 15 kDa TSC22D1.2 isoform in DIV1 CGNs. Heterodimers of TSC22D4 and other TSC22D1 proteins, including the 11 kDa TSC22D1.2 form and TSC22D1.1, were not observed in DIV1 and DIV7 CGNs.

Figure 4.

TSC22D4-42 and TSC22D1-2 heterodimerization in DIV1 CGNs. Total protein extracts of DIV1 CGNs were immunoprecipitated with anti-TSC22D1 (A) or anti-TSC22D4 (B) antibodies, and immunocomplexes were revealed by western blotting with anti-TSC22D1 and TSC22D4 antibodies, respectively. Arrows indicate the IgG heavy chain.


Interest in TSC22D4 was recently increased by the finding that, in a melanocyte cell line, this protein plays a balancing role in cell proliferation versus senescence [6] by physically interacting with TSC22D1, a well-established tumor suppressor [2, 7, 8, 27]. Here we demonstrate that TSC22D4 forms display a variety of specific protein partners and subcellular localizations in undifferentiated and differentiated CGNs (summarized in Table 1), suggesting that the described interaction of TSC22D4 with TSC22D1 [18] is only one of a number of form-specific functions still to be characterized. We demonstrate that TSC22D4–TSC22D1 heterodimers result from the specific association of the TSC22D4-42 and 15 kDa TSC22D1.2 isoforms, and that this association is restricted to the cytosolic compartment of undifferentiated (DIV1) CGNs. The lack of TSC22D4-42–TSC22D1.2 heterodimers in DIV7 CGNs is apparently related to a strong decrease in TSC22D1.2 expression accompanying CGN differentiation, favoring the idea that these heterodimers are involved in the onset of differentiation rather than full expression of differentiative traits. In addition to its interaction with TSC22D1.2 in the cytosol, TSC22D4-42 is also consistently localized in the nuclear matrix fraction of differentiated CGNs, suggesting other functional role(s) independent of TSC22D1.2. A prominent nuclear matrix localization was also displayed by the splice variant TSC22D4-55 in differentiated but not undifferentiated CGNs. However, because this form lacks both the TSC box (a putative DNA-binding domain) and the leucine zipper that are typical of TSC22D proteins [13], its functional significance remains obscure.

Table 1. Subcellular localizations and protein partners of TSC22D4 forms. Undifferentiated CGNs were obtained from DIV1 in vitro cultures or PN3 cerebella, and differentiated CGNs were obtained from DIV7 in vitro cultures or adult cerebella
FormPost-translational modificationLocalizationHeterodimerization
42 kDaNoCytosol (undifferentiated)NoYesNo
Cytosol (differentiated)NoNoNo
Nuclear matrix (differentiated)NoNoNo
Mitochondria (differentiated)NoNoNo
55 kDaNoCytosol (undifferentiated/differentiated)NoNoNo
Nuclear matrix (undifferentiated/differentiated)NoNoNo
67 kDaO-glycosylation (others?)Cytosol (undifferentiated/differentiated)NoNoNo
Mitochondria (differentiated)NoNoYes
72 kDaPhosphorylationChromatin (undifferentiated/differentiated)NoNoNo

In contrast to splice variants TSC22D4-42 and TSC22D4-55, TSC22D4-67 and TSC22D4-72 arise from O-GlcNAcylation and phosphorylation. Even though the existence of other TSC22D4 post-translational modifications in these forms cannot be ruled out at present, O-GlcNAcylation is of particular interest because it represents a type of dynamic glycosylation that is unique to nucleocytoplasmic proteins [28]. In fact, TSC22D4-67 and TSC22D4-72 display cytoplasmic and nuclear localization, respectively, ruling out the possibility of N-linked/O-linked glycosylation patterns typical of secretory or external membrane surface-bound proteins. The TSC22D4-67 form is similarly abundant in the cytosol of both undifferentiated and differentiated CGNs, but also displays a prominent localization in mitochondria of differentiated but not undifferentiated CGNs, where it is associated with AIF. Initially considered as soluble inter-membrane space protein [20], AIF was subsequently shown to be N-terminally anchored to the inner mitochondrial membrane [29], and more recently was also identified at the level of the outer mitochondrial membrane at a smaller level (~ 30% of total mitochondrial AIF) in cortical neurons [26, 30], a localization that possibly represents a strategy for rapid release of uncleaved 62 kDa AIF upon exposure of cortical neurons to a pro-apoptotic stimulus [26]. Similar 70% and 30% distributions between the inner and outer mitochondrial membranes, respectively, were also observed in the present experiments for both AIF and TSC22D4-67, reinforcing the idea that mitochondrial TSC22D4-67 function(s) is mediated by heterodimerization with AIF. To our knowledge, mitochondrial TSC22D4-67 represents the first AIF interactor identified by a direct biochemical approach. In addition to mitochondrial localization and interaction with AIF, TSC22D4-67 is also a constant component of CGN nuclear sap in both undifferentiated and differentiated CGNs. This localization, and the finding that the quantity of TSC22D4-67 significantly decreases during CGN differentiation, while chromatin-associated TSC22D4-72 correspondingly increases, support the possibility that it represents the phosphorylation substrate required for TSC22D4-72 production. Finally, the TSC22D4-72 form appears to be a stable chromatin component at all CGN differentiation stages, probably representing the TSC22D4 form that is most directly/indirectly involved in the regulation of gene expression.

Experimental procedures


Cerebella were obtained from our colony of Swiss Webster CD1 mice (Charles River Laboratories, Calco, Italy). Mice were killed by cervical dislocation (adults) or decapitation (infant/juvenile). The experimental protocol and related procedures were approved by the Italian Ministry of Public Health. Mice were raised according to Sapienza University guidelines for the care and use of laboratory animals and under Sapienza Veterinary Office supervision and approval. All efforts were made to limit the number of animals used.


The antibodies used are listed in Table 2.

Table 2. Antibodies used
Primary antibodies
Anti-TSC22D4Aviva Systems Biology, San Diego, CA; #ARP373391 : 1500
AbCam, Cambridge, UK; #ab562361 : 500
Anti-TSC22D1Protein Tech Group, Chicago, IL; #10214-11 : 1000
Santa Cruz Biotechnology, Santa Cruz, CA; #sc-278441 : 200
Anti-AIFEpitomics, Burlingame, CA; #1020-11 : 1000
Santa Cruz Biotechnology; #sc-555191 : 200
Anti-histone H1Santa Cruz Biotechnology; #sc-86161 : 200
Anti-Lamin B1AbCam; #ab160481 : 4000
Anti-Sp1Santa Cruz Biotechnology; #sc-140271 : 150
Anti-βIII tubulinAbCam; #ab479671 : 400
Anti-Bcl-2Santa Cruz Biotechnology; #sc-73821 : 200
Anti-cytochrome cSanta Cruz Biotechnology; #sc-71591 : 100
Anti-Hsp60Enzo Life Sciences, Farmindgale, NY; #ADI-SPA-8281 : 1000
Anti-O-GlcNAc, clone CTD110.6Covance, Richmond, CA; #MMS-248R1 : 1000
Anti-glyceraldehyde-3-phosphate dehydrogenaseMillipore, Milan, Italy; #MAB3471 : 500
TSC22D4 synthetic peptideAviva Systems Biology; #P23304 1 : 300 
Secondary antibodies
Horseradish peroxidase-conjugated goat anti-rabbit IgGThermo Scientific, Rockford, IL; #324601 : 650
Horseradish peroxidase-conjugated goat anti-mouse IgGThermo Scientific; #324301 : 650
Horseradish peroxidase-conjugated rabbit anti-goat IgGVector Laboratories, Burlingame, CA; #BA-50001 : 40 000
Horseradish peroxidase-conjugated goat anti-mouse IgMSanta Cruz Biotechnology; #sc-20641 : 200

CGN isolation and in vitro culture

CGNs were isolated from cerebella of postnatal 4–6-day-old mice as previously described [31], and cultured in vitro for 1 day (DIV1, undifferentiated) or 7 days (DIV7, differentiated) in Dulbecco's modified Eagle's medium containing 25 mm KCl, 2 mm glutamine, 2% B27, 100 mg·mL−1 penicillin/streptomycin, 5% fetal bovine serum and 10 mm cytosine-β-d-arabinofuranoside (Ara-C).

Protein dephosphorylation and deglycosylation

Total DIV7 CGN lysates prepared as described previously [16] were incubated for 3 h at 37 °C with 0.3 IU·mL−1 calf intestinal alkaline phosphatase (New England Biolabs, Ipswich, MA, USA). Enzymatic N- or O-deglycosylation was performed using the ProZyme®, Glyko®, PRO-Link Extender™ kit (GLYCO; Prozyme, Cambridge, UK) according to the manufacturer's instructions. To determine the presence of O-GlcNAc moieties, DIV7 CGN lysates were immunoprecipitated with anti-TSC22D4 antibodies and then analyzed by western blotting. Blots were probed with anti-TSC22D4 antibodies or subjected to chemical β-elimination [24] by treatment with 55 mm NaOH at 40 °C overnight (removing O-GlcNAc residues), and then incubated with an antibody against O-GlcNAc.

Preparation of nuclear and cytoplasmic fractions

Total nuclear and cytoplasmic fractions were prepared from DIV1 and DIV7 CGNs as described by Bohinski et al. [32]. Cell monolayers were lysed with 10 mm HEPES, pH 7.9, 10 mm KCl, 1.5 mm MgCl2, 0.5 mm dithiothreitol, 0.05% Nonidet P-40 and protease and phosphatase inhibitor cocktails (Roche Diagnostics, Indianapolis, IN, USA), and centrifuged at 1000 g for 10 min. The supernatant (total cytoplasmic fraction) was diluted with Laemmli buffer, and the pellet (total nuclear fraction) was resuspended in 5 mm HEPES, pH 7.9, 300 mm NaCl, 1.5 mm MgCl2, 0.2 mm EDTA, 0.5 mm dithiothreitol, 26% glycerol and protease/phosphatase inhibitors, then extracted for 30 min in ice, cleared by centrifugation at 24 000 g, and finally diluted with Laemmli buffer.

Western blot assays

Protein concentration was routinely determined using the DC Protein Assay (Bio-Rad Laboratories, Hercules, CA, USA), loading equal amounts of total protein per lane on a 10% or 4-20% gradient Mini-Protean TGX pre-cast gel for electrophoresis (Bio-Rad Laboratories). Fractionated proteins were transferred to poly(vinylidene difluoride) membranes (Roche Diagnostics) and analyzed by western blotting using SuperSignal West Dura reagents (Thermo Scientific/Pierce, Rockford, IL, USA).

Preparations of chromatin and nuclear matrix fractions from in vitro cultured CGNs

Purified chromatin and nuclear matrix fractions were prepared from 5 × 106 in vitro cultured CGNs as described by Reyes et al. [33] with minor modifications. Briefly, cells were lysed using cytoskeleton buffer containing 10 mm PIPES, pH 6.8, 100 mm NaCl, 300 mm sucrose, 3 mm MgCl2, 1 mm ethylene glycol tetraacetic acid (EGTA), 1 mm dithiothreitol, 0.5% v/v Triton X-100 and protease/phosphatase inhibitors. The lysate was then centrifuged at 5000 g to separate soluble proteins (supernatant) from the cytoskeleton framework (pellet). The pellet was then dissolved in cytoskeleton buffer and digested with 1 mg·mL−1 RNase-free DNase I for 15 min at 37 °C. Chromatin-associated proteins (including 95% of total histones) were made soluble by 0.25 M ammonium sulfate and recovered in the supernatant after centrifugation at 5000 g for 3 min (chromatin fraction). The pellet was further extracted with 2 m NaCl in cytoskeleton buffer buffer for 5 min at 0–4 °C, and centrifuged at 5000 g for 3 min at 4 °C, removing the supernatant (wash). The remaining pellet (matrix fraction, containing nuclear matrix-associated proteins) was dissolved in urea buffer (8 m urea, 0.1 m NaH2PO4, 0.01 m Tris/HCl, pH 8.0).

Preparation of cytosol, mitochondria and sub-mitochondrial fractions

Mitochondria were prepared as described by Frezza et al. [34]. Briefly, 5-7 adult mouse cerebella (0.6 g) were homogenized in 6 mL of IBm1 buffer (1 m Tris/HCl, pH 7.4, 1 m KCl, 1 M EDTA, 1 m sucrose, 10% BSA). Following low-speed centrifugation at 1000 g for 5 min to remove unbroken cells and nuclei, the supernatant was centrifuged at 8000 g for 10 min. The resulting supernatant (S1) and pellet (P1) were collected separately. S1 was further cleared by centrifugation at 100 000 g for 1 h (cytosol). P1 was re-suspended in 2 mL of IBm2 (1 m Tris/HCl, pH 7.4, 0.1 m EGTA/Tris, 1 m sucrose) and centrifuged at 8000 g for 10 min, giving a pellet (mitochondria) that was dissolved in lysis buffer (10 mm Tris, pH 8.0, 100 mm NaCl, 2 mm EDTA, 1% Triton X-100 and protease/phosphatase inhibitors).

Sub-mitochondrial fractions (inner mitochondrial membranes, mitochondrial inter-membrane space, outer mitochondrial membranes and mitochondrial matrix) were prepared from 8 to 10 adult mouse cerebella exactly as described by Zhang et al. [25]. Fraction purity was assessed by western blotting using anti-Bcl-2 (for outer mitochondrial membranes), anti-AIF (for inner mitochondrial membranes), anti-Hsp60 (for the mitochondrial matrix) and anti-cytochrome c (for the mitochondrial inter-membrane space).

Immunoprecipitation assays

Immunoprecipitation assays were performed on both purified mitochondria from adult cerebella and total lysates of 8 x 106 DIV1/DIV7 CGNs. The following antibodies (5 μg·mL−1) were added to lysates: mouse/rabbit monoclonal anti-AIF (Santa Cruz Biotechnology, Santa Cruz, CA/Epitomics, Burlingame, CA, USA), rabbit polyclonal anti-TSC22D4 (AbCam, Cambridge, UK/Aviva Systems Biology, San Diego, CA), rabbit/goat polyclonal anti-TSC22D1 (Protein Tech Group, Chicago, IL, USA/Santa Cruz Biotechnology) or an equal amount of IgG corresponding to the host species of the primary antibody and then incubated overnight at 0–4 °C. Following addition of protein A/G-conjugated agarose beads and incubation for 4 h at 0–4 °C, immunoprecipitates were rinsed with immunoprecipitation buffer and eluted by boiling in Laemmli buffer for 5 min. Immunocomplexes were fractionated by SDS/PAGE and processed by western blotting.

In silico determination of putative TSC22D4 post-translational modification sites

The in silico analysis of putative TSC22D4 post-translational modification sites, including O-glycosylation, N-glycosylation and phosphorylation, was performed using the website

Statistical analysis

Statistical significance was determined by repeated-measures ANOVA and Tukey's post hoc test, using SPSS software (SPSS Inc, Chicago, IL, USA).


We thank Rossella Puglisi DAHFMO, Section of Histology & Medical Embryology, University La Sapienza of Rome, for advice on sub-nuclear protein fractionation. This work was supported by grants from the Pasteur Institute/Cenci Bolognetti Foundation, Sapienza University, Italy (2008–2010) and Sapienza University ‘Ateneo’ (2011–2012).