Nuclear matrix-associated protein SMAR1 regulatesalternative splicing via HDAC6-mediated deacetylationof Sam68Kiran Kumar Nakkaa,b, Nidhi Chaudharya, Shruti Joshia, Jyotsna Bhatc, Kulwant Singhb, Subhrangsu Chatterjeec,Renu Malhotrad, Abhijit Ded, Manas Kumar Santrae, F. Jeffrey Dilworthb,f, and Samit Chattopadhyaya,1

aChromatin and Disease Biology Laboratory, National Centre for Cell Science, Pune 411007, India; bSprott Centre for Stem Cell Research, Ottawa HospitalResearch Institute, Ottawa, ON, K1H 8L6, Canada; cDepartment of Biophysics, Bose Institute, Kolkata 700054, India; dAdvanced Center for Treatment,Education and Research, Tata Memorial Centre, Navi Mumbai 410210, India; eCancer and Epigenetics Lab, National Centre for Cell Science, Pune 411007,India; and fDepartment of Cellular and Molecular Medicine, University of Ottawa, Ottawa, ON, K1H 8L6, Canada

Edited by Mark Groudine, Fred Hutchinson Cancer Research Center, Seattle, WA, and approved May 15, 2015 (received for review September 29, 2014)

Pre-mRNA splicing is a complex regulatory nexus modulated byvarious trans-factors and their posttranslational modifications to cre-ate a dynamic transcriptome through alternative splicing. Signal-induced phosphorylation and dephosphorylation of trans-factorsare known to regulate alternative splicing. However, the role ofother posttranslational modifications, such as deacetylation/acetyla-tion, methylation, and ubiquitination, that could modulate alternativesplicing in either a signal-dependent or -independent manner remainenigmatic. Here, we demonstrate that Scaffold/matrix-associated re-gion-binding protein 1 (SMAR1) negatively regulates alternative splic-ing through histone deacetylase 6 (HDAC6)-mediated deacetylationof RNA-binding protein Sam68 (Src-associated substrate during mitosisof 68 kDa). SMAR1 is enriched in nuclear splicing speckles and asso-ciates with the snRNAs that are involved in splice site recognition.ERK–MAPK pathway that regulates alternative splicing facilitates ERK-1/2–mediated phosphorylation of SMAR1 at threonines 345 and 360and localizes SMAR1 to the cytoplasm, preventing its interaction withSam68. We showed that endogenously, SMAR1 through HDAC6maintains Sam68 in a deacetylated state. However, knockdown orERK-mediated phosphorylation of SMAR1 releases the inhibitorySMAR1–HDAC6–Sam68 complex, facilitating Sam68 acetylation andalternative splicing. Furthermore, loss of heterozygosity at theChr.16q24.3 locus in breast cancer cells, wherein the human homologof SMAR1 (BANP) has been mapped, enhances Sam68 acetylation andCD44 variant exon inclusion. In addition, tail-vein injections in micewith human breast cancer MCF-7 cells depleted for SMAR1 showedincreased CD44 variant exon inclusion and concomitant metastaticpropensity, confirming the functional role of SMAR1 in regulationof alternative splicing. Thus, our results reveal the complex molecularmechanism underlying SMAR1-mediated signal-dependent and-independent regulation of alternative splicing via Sam68 deacetylation.

SMAR1 | Sam68 | HDAC6 | alternative splicing | deacetylation

Newly synthesized pre-mRNAs undergo multiple posttranscrip-tional gene-regulatory events, such as capping, splicing, cleavage,

and polyadenylation. Of these, splicing is most stringently regulated,because it is a prerequisite for spatiotemporal generation of splicevariants observed in 95% of human genes (1). Accuracy of alternativesplicing (AS) is modulated by cis-regulatory elements that includesplice enhancers and silencers and/or trans-acting factors, viz. theserine–arginine (SR)-rich family proteins, heteronuclear ribonucleo-proteins (hnRNPs) and various chromatin modifiers (2, 3). Themajority of these trans-factors are distributed along the nuclear matrix(NM), which is a fibro-granular structural framework, organized aschromatin and ribonucleoprotein (RNP) domains. Chromatin do-mains are involved in replication and transcription, whereas RNPdomains are actively associated with cotranscriptional and post-transcriptional gene-regulatory events (4, 5). Thus, NM is consideredto be the site for pre-mRNA processing; however, the precise role ofNM-associated proteins in the regulation of AS remain elusive.

Scaffold/matrix-associated region-binding protein 1 (SMAR1) isone such NM-associated protein identified in mouse double-pos-itive thymocytes (6). The human homolog of SMAR1, BANP, ismapped to chromosome 16q24.3 locus. Loss of heterozygosity(LOH) in this locus is implicated in multiple cancers (7, 8).SMAR1 delays tumor progression (9, 10) and inhibits the ex-pression of multiple genes (11, 12). Clinical specimens of differenthistological grade breast cancer tissues indicated that SMAR1expression is drastically down-regulated in the higher grades ofinfiltrating ductal carcinomas (13).Cancer cells often take advantage of AS because the spliceosome

machinery goes berserk in cancers, resulting in the expression ofprotein isoforms that promote growth and metabolism of tumorcells (14–16). Signal-transduction pathways—such as Ras, PI3-kinase,AKT, ERK, PRP4, and SRPK1—have been reported to bederegulated in cancerous cells (17). Posttranslational modifications(PTMs) induced by kinases and phosphatases on regulatory factorsmodulate tumor progression and metastasis, as exemplified byCD44 gene AS (18–21). CD44, a cell-surface transmembraneglycoprotein, expresses variant isoforms (CD44v) that increase

Significance

Multiple studies highlight the role of various proteins in regu-lation of alternative splicing; however, the regulatory role ofdistinct posttranslational modifications during alternative splic-ing that contribute to tumorigenesis is enigmatic. Here we reporta previously unidentified noncanonical mechanism of regulationof alternative splicing modulated by deacetylation of RNA-bindingprotein Sam68 (Src-associated substrate during mitosis of68 kDa) via Scaffold/matrix-associated region-binding protein 1(SMAR1)–histone deacetylase 6 (HDAC6) complex. SMAR1 incomplex with HDAC6 maintains Sam68 in a deacetylated state.We observed that ERK-1/2–dependent phosphorylation ofSMAR1, knockdown of SMAR1, or loss of heterozygosity facili-tates CD44 variant exon inclusion via Sam68 acetylation andthus confers invasive and metastatic potential in breast tumorcells. Our findings provide key insights into regulation of alter-native splicing and the potential for therapeutic interventionduring tumor metastasis.

Author contributions: K.K.N., F.J.D., and S. Chattopadhyay designed research; K.K.N., N.C.,S.J., J.B., K.S., S. Chatterjee, R.M., and A.D. performed research; A.D., M.K.S., F.J.D., andS. Chattopadhyay contributed new reagents/analytic tools; K.K.N., N.C., S.J., J.B., S. Chatterjee,and S. Chattopadhyay analyzed data; and K.K.N., N.C., and S. Chattopadhyay wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. Email: samit@nccs.res.in.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1418603112/-/DCSupplemental.

E3374–E3383 | PNAS | Published online June 15, 2015 www.pnas.org/cgi/doi/10.1073/pnas.1418603112

Dow

nloa

ded

by g

uest

on

June

20,

202

0

metastatic propensity in multiple tumor types (22). Generation ofCD44v is modulated by AS events, which require activation of theERK–MAPK pathway (23). However, a comprehensive role fortrans-factors that negatively influences CD44 AS in a signal-dependent and -independent manner remains to be established.Our present study revealed SMAR1 as a negative regulator of

CD44 gene AS in a signal-dependent and -independent mannerby modulating Sam68 (Src-associated substrate during mitosis of68 kDa) deacetylation. Here, we document that SMAR1, incooperation with histone deacetylase 6 (HDAC6), interacts withSam68 and maintains it in a deacetylated state, concomitantlyinhibiting the inclusion of CD44 alternate exons. Furthermore,we show that ERK activation, knockdown of SMAR1, LOH atchromosome 16q24.3 locus, and/or inactivation of HDAC6 en-hances Sam68 acetylation and facilitates CD44 variant exoninclusion. Overall, our results demonstrate the role of NM-associated protein SMAR1 in the regulation of AS.

ResultsSMAR1 Is Enriched in Splicing Speckles and Regulates AS. SC35-enriched nuclear speckles are considered as a hallmark of proteinsinvolved in pre-mRNA splicing (24). Presence of a stretch of serineand proline residues, in addition to interspersed arginine residuesin the protein-interaction domain of SMAR1 (SI Appendix, Fig.S1A), inspired us to examine the distribution of SMAR1 in theRNP matrix. Initially, we found that the highly abundant SMAR1protein had a homogenous distribution within the nucleus, makingit difficult to resolve its distribution in the RNP matrix and/or itsassociation with SC35-splicing speckles using immunofluorescence(IF) of intact cells. However, in situ immunostaining of NM (25)allowed us to reveal that SMAR1 is enriched in the RNP-richsplicing speckles and colocalizes with SC35 (Fig. 1A, −RNase A).Treatment of NM with RNase A abolished the occurrence of thesenuclear speckles (Fig. 1A, +RNase A), suggesting that SMAR1-enriched speckles are RNPs in nature. Further, RNA immuno-precipitation (RIP) assays showed an association of SMAR1 withspliceosome component snRNAs, U1, U2, U4, U5, and U6snRNAs that are involved in splice site recognition (Fig. 1B). In-terestingly, affinity of SMAR1 for U5 snRNA is higher than it isfor any other snRNA. The specific association of SMAR1 withsnRNAs, but not with GAPDH mRNA, prompted us to investi-gate its effect on the regulation of AS. In vivo splicing assays withminigenes harboring the cis-regulatory elements controlling theAS of either FAS (26) or CD44 (23) revealed that SMAR1knockdown augments the inclusion of alternative exons, whereasoverexpression reduced the exon inclusion (Fig. 1 C and D and SIAppendix, Fig. S1 B and C). However, to eliminate the possibilitythat overexpression or knockdown of SMAR1 alters the stabilityof mRNA splice variants per se, we performed mRNA stabilityassays using Actinomycin D. Inhibition of transcription by usingActinomycin D in HEK293T cells that were previously cotrans-fected with either SMAR1 overexpression or knockdown plasmidsalong with FAS and CD44 minigenes resulted in no change in theobserved ratio of spliced mRNAs at any given time point, in-dicating that SMAR1 expression does not contribute toward thedifferential stability of mRNA splice variants (Fig. 1 E and F).Together, these data indicate that SMAR1 specifically regulatesAS without affecting the stability of splice variants.

SMAR1 Knockdown or LOH at Chr.16q24.3 Enhances CD44 VariantExon Inclusion and Promotes Lung Metastasis. Considering that thelevels of SMAR1 are drastically down-regulated in the highergrades of breast tumors (13), and the inclusion of CD44 variantexons confers invasive and metastatic propensity to tumor cells(27, 28), we set out to decipher the molecular mechanismunderlying SMAR1-mediated regulation of CD44 AS. Semi-quantitative exon-specific RT-PCR analysis of CD44 splicevariants using different primer sets (20) (Fig. 2A and SI Appendix,

Table S1) in HeLa cells that are stably expressing a SMAR1-tar-geting shRNA (Fig. 2B) revealed an enhanced incorporation ofvariant exons v2–v6 (proximal variant exons; Fig. 2C). Interestingly,we observed no significant change in the incorporation of variableexons v7–v10 (distal variant exons; SI Appendix, Fig. S1D, Left) andstandard CD44 (CD44s). In corollary, overexpression of SMAR1(Fig. 1D) inhibited the inclusion of proximal variant exons (Fig.1E), whereas distal variant exons and CD44s remain unaltered

Fig. 1. SMAR1 is enriched in splicing speckles and regulates AS. (A) IFanalysis of in situ nuclear matrices extracted from HeLa cells. Cells were ei-ther untreated (−) or treated with RNase A (+) before fixation and stainedwith SC35 (green) and SMAR1 (red). (B) Real-time quantitative PCR (RT-qPCR)analysis of immunoprecipitated RNA eluates from α-SMAR1 antibody andcontrol antibody (IgG) in formaldehyde-fixed HeLa cells. Error bars representSD from three independent experiments. (C) Schematic representation ofFAS minigene with alternative exon E6. FAS minigene was cotransfectedalong with Flag-SMAR1 (FSM) and SMAR1–sh745. Semiquantitative RT-PCRanalysis (C, Left) with mapped primers and relative quantification (C, Right;n = 3) showed that knockdown of SMAR1 favors inclusion of alternate exonE6. (D) A schematic representation of luciferase-based CD44 pET-v5-Lucminigene construct. CD44 minigene construct was cotransfected along withFSM or SMAR1–sh745. Semiquantitative RT-PCR analysis (D, Left) and relativequantification (D, Right; n = 3) indicated an increase in the inclusion of v5alternate exon. (E and F) A measure of mRNA stability of alternativelyspliced transcripts upon SMAR1 overexpression and knockdown. Cells wereharvested at different time points as indicated after treatment with Acti-nomycin D and analyzed for stability through measurement of temporalchanges in abundance of each splice variant, CD44 (E) and FAS (F). A ratio ofrelative exon inclusion during overexpression or knockdown of SMAR1during different time points showed no effect on the stability of mRNAsplice variants of target genes. Error bars represent SD from three inde-pendent experiments. FV, Flag vector; shCntrl, Control shRNA; shSM, SMAR1shRNA (sh745). Student’s t tests were used to calculate the P values. Differ-ences with P < 0.05 were considered statistically significant.

Nakka et al. PNAS | Published online June 15, 2015 | E3375

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

20,

202

0

(SI Appendix, Fig. S1D, Right), indicating that SMAR1 negativelyregulates the inclusion of CD44 variable exons v2–v6.Human breast cancer cells such as MDA-MB-468 and T47D

harbor LOH, whereas MCF-7 cells lack LOH (7, 29). Westernblot analysis showed that MDA-MB-468 and T47D cells expressminimal amounts of SMAR1, whereas MCF-7 cells express sig-nificant amounts of SMAR1, corroborating with the LOH status(Fig. 2F). Because SMAR1 expression inversely correlates withCD44 variant exon inclusion, we investigated CD44 AS in con-text of LOH. In MDA-MB-468 and T47D cells, we observed acategorical increase in the inclusion of variant exons v2–v10,whereas CD44 variant exon inclusion is negligible in SMAR1

expressing MCF-7 cells (Fig. 2G and SI Appendix, Fig. S1E, Left).Nevertheless, ectopic expression of SMAR1 in LOH-harboringMDA-MB-468 and T47D cell lines by using adeno-SMAR1 (Ad-SM) (Fig. 2H) consistently decreased the inclusion of proximalexons (Fig. 2I), whereas no significant change was observed in thelevels of distal variant exons and CD44s (SI Appendix, Fig. S1E,Right). Together, these data indicate that LOH at 16q24.3 locus,wherein SMAR1 has been mapped, contributes to CD44 AS.Considering that the inclusion of CD44 variable exons confers the

invasive and metastatic proclivity to tumor cells, we investigated themetastatic potential of MCF-7SMAR1−/− cells. Tail-vein injections insyngeneic nonobese diabetic/severe combined immunodeficiency

Fig. 2. SMAR1 knockdown or LOH in breast cancer cells enhances CD44 AS and promotes lung metastasis. (A) Schematic representation of organization ofconstant and variable exons in CD44 gene. The relative location of PCR primers used to perform exon-specific PCR is indicated. (B and C) Western blot analysis ofSMAR1 (B) and semiquantitative RT-PCR analysis of CD44 variant exon inclusion (C) in HeLa cells stably knock down with nonsilencing (NS) and different SMAR1–shRNA lentivirus (sh1, sh3, and sh5). (D and E) Western blot analysis of SMAR1 (D) and semiquantitative RT-PCR analysis of CD44 variant exon inclusion (E) upontransduction with recombinant SMAR1 adenovirus (Ad-SM). (F and G) Western blot analysis of SMAR1 (F) and semiquantitative RT-PCR analysis of CD44 variantexon inclusion (G) in the indicated breast cancer cell lines. (H and I) Western blot analysis of SMAR1 (H) and semiquantitative RT-PCR analysis of CD44 variant exoninclusion (I) upon transduction of the indicated LOH-containing breast cancer cells using Ad-SM. (J) Tail-vein injections in NOD/SCID mice with MCF-7 cells thatwere previously stably transduced with nonsilencing (NS; n = 6) and SMAR1–shRNA (sh3; n = 6) lentivirus. Lung metastasis assays were performed, and lungnodules are indicated by arrows (J, Upper Left) and were quantified (n = 3) as shown in the graph (J, Right). Histopathological analysis [hematoxylin/eosin (H/E)staining] of the lung sections (J, Lower Left). Student’s t tests were used to calculate the P values. (K ) Immunohistochemistry (IHC) analysis of the lungsections (n = 3) obtained from metastatic breast tumor nodule sites with the indicated proteins (α-SMAR1 and various α-CD44v antibodies).

E3376 | www.pnas.org/cgi/doi/10.1073/pnas.1418603112 Nakka et al.

Dow

nloa

ded

by g

uest

on

June

20,

202

0

(NOD/SCID) mice indicated that MCF-7SMAR1−/− cells showed aprofound increase in invasive and metastatic potential, especially tothe lungs, as evidenced by histopathological analysis (Fig. 2J).Immunostaining of the histological sections of metastatic lung nod-ules with α-SMAR1 and -CD44 variable exon-specific antibodiesindicated enhanced expression of CD44v in the metastatic nod-ules, wherein the amount of SMAR1 is minimal (Fig. 2K). Alltogether, these results indicate that knockdown or absence ofSMAR1 enhances CD44 alternate exon inclusion, and the meta-static propensity served as a hallmark of CD44v inclusion.

Activation of ERK–MAPK Pathway Translocates SMAR1 to the Cytoplasm.ERK–MAPK signaling is considered indispensable and a prerequisitefor CD44 gene AS (18, 23). Considering the similar effects exertedon CD44 gene AS upon low levels of SMAR1 and ERK–MAPK

signal activation, we predicted an intricate interplay betweenSMAR1 expression and ERK–MAPK pathway. Towards this pre-diction, activation of ERK–MAPK pathway using phorbol12-myristate 13-acetate (PMA) in HeLa cells showed no changein SMAR1 levels (SI Appendix, Fig. S2). However, IF analysisrevealed that SMAR1, a primarily nuclear protein, translocated tothe cytoplasm upon ERK activation (Fig. 3A). This observation wasreaffirmed by Western blot analysis of SMAR1 in PMA-stimulatednuclear-cytoplasmic fractions (Fig. 3B). To further determine thatSMAR1 localization is ERK-dependent, we used ERK-1/2 in-hibitor U0126 and observed an abrogation in the translocation ofSMAR1 to the cytoplasm (Fig. 3 A and C). Furthermore, treatmentwith a nuclear export inhibitor, Leptomycin B, also inhibited thecytoplasmic translocation of SMAR1 (Fig. 3 A and D), reinforcingthat activated ERK-1/2 kinase induces nuclear export of SMAR1.

Fig. 3. Occupancy of SMAR1 along the CD44 gene locus and ERK–MAPK signal-induced translocation of SMAR1 to cytoplasm. (A) IF analysis of HeLa cells treatedwith PMA (50 ng/mL, 2 h), U0126 (25 μM, 30 min), and Leptomycin B (10 nM, 10 min), and stained for SMAR1 (red; indicated by arrows) and DNA (DAPI; blue).(B–D) Western blot analysis of whole-cell lysates and nuclear (Nuc) and cytoplasmic (Cyto) fractions of HeLa cells upon treatment with PMA (B), U0126 (C), andLeptomycin B (D) for the indicated time points with the mentioned antibodies. Lamin and Paxillin served as nuclear and cytoplasmic controls, respectively.(E) Schematic representation of CD44 gene locus mapped with exons and introns. Regions indicated are: DP, distal promoter; PP, proximal promoter; e, constant exon;i, intron; v, variable exon. (F) RT-qPCR analysis of α-SMAR1 immunoprecipitated chromatin from HeLa cells treated with or without PMA. The occupancy of SMAR1 isrepresented as relative enrichment normalized with input DNA and control IgG. (G) RT-qPCR analysis of immunoprecipitated CD44 pre-mRNA in the eluates ofα-SMAR1. Graph represents the relative enrichment of SMAR1, normalized with input and control IgG. Error bars represent SD from three independent experiments.

Nakka et al. PNAS | Published online June 15, 2015 | E3377

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

20,

202

0

Collectively, these results indicate that ERK–MAPK signal mimicsSMAR1 knockdown by facilitating its cytoplasmic localization, thuspromoting CD44 gene AS.

SMAR1 Is Accumulated Along the CD44 Gene Body and Binds to CD44Pre-mRNA. An intricate association of chromatin remodelers withsplicing machinery and the chaperoning of pre-mRNA by chro-matin during AS incited us to analyze the distribution of SMAR1along the CD44 gene locus (Fig. 3E). Chromatin immunoprecipi-tation (ChIP) assays indicated an endogenous distribution ofSMAR1 along the intronic and exonic regions of CD44; however,activation of the ERK–MAPK pathway reduced the same (Fig. 3F).Furthermore, occupancy of RNA polymerase (Pol) II (19) inverselycorrelates with that of SMAR1 and also served as positive control(SI Appendix, Fig. S3 A and B). Considering that some of thechromatin proteins, such as heterochromatin protein 1γ (HP1γ)and RNA Pol II, bridge the nascent pre-mRNA to the chromatin(21), we further investigated the association of SMAR1 with CD44pre-mRNA. RIP assays revealed an intricate association of SMAR1with CD44 pre-mRNA in untreated cells (Fig. 3G); however, theassociation was reduced upon ERK-1/2 activation (Fig. 3G, redline). The association of SMAR1 with pre-mRNA containing var-iable exons v3 and v5 was significantly high, indicating that SMAR1associates with CD44 transcript and likely sequesters the variableexons from being incorporated into the mRNA.

Activated ERK-1/2 Kinase Phosphorylates SMAR1 at Threonines 345and 360. ERK–MAPKs are classical protein kinases that phos-phorylate numerous substrates with proline-directed serine/threo-nine residues (18, 30). SMAR1 in its protein-association domainharbors two putative threonine followed by proline (TP) residuespositioned at 345 and 360 (Fig. 4A). Phospho-TP (pTP) assays using

α-pTP antibody revealed time-dependent phosphorylation ofSMAR1 upon ERK-1/2 activation (Fig. 4B). SMAR1 phosphory-lation at TP residues was observed within 15 min of ERK activation,and the levels in the nucleus decreased by 60 min (Fig. 4B, lanes2 and 5), indicating nuclear export of phosphorylated protein. Aconcomitant increase of pTP–SMAR1 in the cytoplasmic fractionswas observed within 15 min until 120 min of ERK activation.However, treatment with U0126 inhibited the phosphorylation ofSMAR1 as well as its cytoplasmic localization (SI Appendix, Fig.S4A). Furthermore, in vitro kinase assay, by incubating recombinantGST–SMAR1 with activated recombinant ERK2 in a concentra-tion-dependent manner, indicated phosphorylation of SMAR1 (Fig.4C). However, treatment with calf intestinal phosphatase (CIP)abated SMAR1 phosphorylation, affirming SMAR1 as a substrateof activated ERK kinase (Fig. 4D). To further ascertain theseobservations, we generated threonine-to-alanine mutants (T345A/T360A) in p3X-Flag-SMAR1. IF (Fig. 4E) and pTP assays (Fig. 4F),along with subcellular fractionation studies (SI Appendix, Fig. S4B),illustrated that T345A/T360A mutants are phosphorylation-deficientand failed to exhibit cytoplasmic localization, rendering insensitivityto ERK activation. To further strengthen these observations, wegenerated in-house antibodies against T345/T360 phospho-SMAR1(Pep3) and T345A/T360A mutant-SMAR1 (Pep6) (SI Appendix,Fig. S4C). Analysis of PMA-stimulated nuclear and cytoplasmicfractions using these antibodies reaffirmed phosphorylation-mediated differential localization of SMAR1 (SI Appendix, Fig.S4D). Despite Pep6-antibody–detected endogenous unphosphory-lated SMAR1, which is primarily localized in the nucleus, it failed torecognize the cytoplasmic as well as nuclear threonine-phosphory-lated SMAR1. These results confirm the specificity of the antibodyas well as phosphorylation of SMAR1 at T345 and T360.

Fig. 4. ERK-1/2 kinase phosphorylates SMAR1 at T345 and T360. (A) Schematic representation of putative ERK-mediated SMAR1 phosphorylation sites atT345 and T360 (highlighted in red). (B) In vivo pTP analysis of α-SMAR1 eluates by immunoblotting with α-pTP antibody in both the nuclear and cytoplasmicfractions of HeLa cells activated with PMA for the indicated time points. IP, immunoprecipitation. (C) In vitro phosphorylation of recombinant GST–SMAR1protein upon incubation with 20 or 80 units (U) of recombinant ERK2 in the presence of ATP. GST served as a control substrate. (D) In vitro dephosphorylationof GST–SMAR1 by calf intestinal phosphatase (CIP; 100 U). (E) IF analysis of HeLa cells transfected with FSM wild-type (FSM) or T345A/T360A mutant [FSM(Mut)] plasmids and stained with Flag (red, indicated by arrows) and DNA (DAPI). (F) In vivo pTP assays of α-Flag–immunoprecipitated eluates from the nuclearand cytoplasmic fractions of HeLa cells transfected with the indicated plasmids and treated with PMA (30 min). (G) RT-PCR monitoring of CD44 alternate exoninclusion in HeLa cells transfected with the indicated plasmids and left untreated or treated with PMA. (H) Western blot analysis of whole-cell lysate andnuclear and cytoplasmic fractions from HeLa cells transfected with the indicated plasmids.

E3378 | www.pnas.org/cgi/doi/10.1073/pnas.1418603112 Nakka et al.

Dow

nloa

ded

by g

uest

on

June

20,

202

0

To further demonstrate the biological significance of T345/T360 phosphorylation of SMAR1 and CD44 AS, we examinedthe CD44 variant exon incorporation upon expression of wild-type Flag-SMAR1 (FSM) and T345A/T360A mutant Flag-SMAR1 (FSM-Mut). As depicted in Fig. 4G, unactivated cellsexpressing FSM, as well as FSM-Mut, showed decreasedCD44v inclusion. Interestingly, activation of the ERK pathwayenhanced variant exon inclusion in cells expressing either vectoralone or FSM, but not in cells expressing FSM-Mut. This resultcan be attributed to the presence of FSM-Mut in the nucleus(Fig. 4H), which impedes the ability of ERK-1/2 to induce aswitch in CD44 variable exon inclusion. Together, these resultsindicate that phosphorylation of SMAR1 at T345/T360 is crit-ical and necessary for ERK–MAPK signal-induced inclusion ofCD44 variant exons.

SMAR1 Knockdown Promotes AS Independent of the ERK–MAPKPathway. Given that SMAR1 knockdown facilitates CD44gene AS in a manner similar to that of activation of the ERK–

MAPK pathway (23), we investigated a possible leaky activa-tion of ERK-1/2 upon SMAR1 knockdown. Toward this hy-pothesis, we checked for the expression of phosphorylated ERK-1/2(activation of ERK) in the nuclear extracts from SMAR1knockdown HeLa cells. Interestingly, unlike the PMA-stimu-lated nuclear extracts that showed the presence of phospho–ERK-1/2 (Fig. 5A, lane 2), SMAR1-knocked-down nuclearextracts did not show any of its expression (Fig. 5A, lane 4), in-dicating ineffective activation of ERK-1/2. However, we observeda consistent increase in CD44 variant exon inclusion uponSMAR1 depletion (Fig. 5B, lane 4), indicating that SMAR1knockdown-mediated regulation of the CD44 gene AS is ERK–MAPK signal-independent. This finding incited us to investigate

Fig. 5. SMAR1 associates with Sam68 and modulates its acetylation status. (A and B) Western blot analysis of whole-cell, nuclear, and cytoplasmic fractions (A)and RT-PCR analysis of CD44 alternate exon inclusion (B) of HeLa cells transduced with the indicated lentivirus and subsequently treated with either PMA orU0126, highlighting signal-independent regulation of AS. (C) Immunoprecipitation assay of PMA-activated nuclear extracts using α-SMAR1 and α-Sam68 anti-bodies. Eluates were immunoblotted with the indicated antibodies. In, Input; IgG, Control IgG; IP, immunoprecipitate. (D) A schematic map of Sam68 protein withthe KH domain highlighted in red (Upper). An in silico docking analysis of the association of SMAR1 (green) and Sam68 (white) (Lower). (E) Sam68 acetylationassays performed by immunoprecipitation with α-acetyl lysine (Ac.K) antibody in the nuclear extracts of HeLa cells that were stably transduced with the indicatedlentivirus, and the eluates were immunoblotted (IB) with the mentioned antibodies. (F) Nuclear extracts of HeLa cells activated with PMA for the indicated timepoints were assayed for Sam68 acetylation. (G) A schematic representation of myc-tagged full-length (FL) and truncated versions of Sam68 protein. (H) HeLa cellsthat were transfected with the indicated constructs of myc-tagged Sam68 and treated with PMA were assayed for acetylation by using α-Ac.K antibody. *IgG.

Nakka et al. PNAS | Published online June 15, 2015 | E3379

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

20,

202

0

the molecular mechanism underlying SMAR1-mediated signal-independent regulation of AS.

SMAR1 Interacts with Sam68 in a Signal-Dependent Manner. Sam68,a STAR-family protein, associates with splice-regulatory ele-ments on pre-mRNAs and plays a strategic role in CD44 gene AS(18). Of significance, SMAR1 association with CD44 pre-mRNAprompted us to investigate its association with Sam68. Immu-noprecipitation (IP) assays in HeLa cell nuclear extracts revealedan endogenous association of SMAR1 with Sam68 (SI Appendix,Fig. S5A). However, the association of SMAR1 with Sam68 isprevented in ERK-activated nuclear extracts due to the differ-ential localization of SMAR1 to the cytoplasm (Fig. 5C, lanes 2and 6). Furthermore, IP analysis of nuclear extracts treated withRNase A abolished the SMAR1–Sam68 interaction. (SI Appen-dix, Fig. S5B). Together, these data suggest that endogenously,SMAR1 associates with Sam68 in an RNA-dependent manner,and this association is perturbed due to differential localizationof SMAR1 upon activation of ERK-1/2. This finding led us tohypothesize that SMAR1 in complex with Sam68 might form aninhibitory complex that perturbs AS, which inspired us to de-lineate the SMAR1–Sam68 nexus that regulates AS.

Depletion of SMAR1 and/or Activation of ERK-1/2 Enhances Sam68Acetylation. In silico docking studies of SMAR1 and Sam68revealed that C-terminal domain (CTD) of SMAR1 (residues 248–371) associates with Sam68 and shares a complimentary shape withthe lysine-rich (21 lysine residues) KH domain of Sam68 (Fig. 5Dand SI Appendix, Fig. S5 C–E and Table S2). PTMs on trans-factorsplay a decisive role during AS (18, 31, 32), and SMAR1 modulatesPTMs such as deacetylation/acetylation through its association withHDAC1. The association of SMAR1 with the lysine-rich KH do-main of Sam68 led us to hypothesize that SMAR1 might modulatethe acetylation/deacetylation dynamics of Sam68. IP assays usingα-acetyl-lysine (α-Ac.K) antibody revealed enhanced acetylation ofSam68 upon SMAR1 knockdown (Fig. 5E, lanes 2 and 3), despitetotal Sam68 levels remaining unaltered, emphasizing that endoge-nously Sam68 is maintained in a deacetylated state and knockdownof SMAR1 facilitates Sam68 acetylation. Because ERK activationprevents SMAR1–Sam68 association, we investigated Sam68 acet-ylation upon ERK activation. Consistently, we observed an accu-mulation of acetylated Sam68 upon ERK activation (Fig. 5F, lanes3–6). Of significance, acetylation kinetics of Sam68 indicated thatphosphorylation of SMAR1 (Fig. 4B, 15 min) precedes Sam68acetylation upon ERK activation (Fig. 5F, 30 min). Strikingly, in-hibition of ERK-1/2 with U0126 prevented Sam68 acetylation (SIAppendix, Fig. S5F). Furthermore, Sam68 acetylation in SMAR1knockdown cells treated with U0126 convincingly demonstratedthat Sam68 acetylation depends on the presence or absence ofSMAR1 in the nucleus (SI Appendix, Fig. S5G). Together, thesedata reveal that knockdown of SMAR1 and/or activation of ERK-1/2 are equally effective in facilitating Sam68 acetylation.To further establish the significance of Sam68 acetylation during

AS, we generated various deletion constructs of Sam68 (Fig. 5G)and verified their acetylation status as well as the efficiency tomodulate AS. Results showed that myc-Sam68 full-length (1–443),myc-Sam68 (90–443), and myc-Sam68 (90–257) are acetylatedupon ERK activation (Fig. 5H, lanes 2, 4, and 8), whereas myc-Sam68 (257–443), lacking lysine-rich KH domain, failed to showany acetylation (Fig. 5H, lane 6). Acetylation of the minimal do-main, myc-Sam68 (90–257), harboring 21 lysine residues, but notmyc-Sam68 (257–443), emphasized that lysine residues are thetarget sites of Sam68 acetylation. CD44 splice reporter minigenesplicing assays in MCF-7Sam68−/− cells (SI Appendix, Fig. S5H) in-dicated that the lysine-rich KH domain is indispensable for CD44AS (SI Appendix, Fig. S5I). In addition, we investigated whetherthe mechanism of Sam68 acetylation-mediated regulation is ap-plicable during AS of other target genes. Toward this investigation,

we assayed FASminigene AS (SI Appendix, Fig. S5J) and one of thetarget genes of Sam68, Bcl-X gene AS (33) (SI Appendix, Fig. S5K).Our results from these studies indicated that acetylation of Sam68through the mechanism of ERK activation facilitates AS. Together,these results distinctly highlight that deacetylation/acetylation switchon Sam68 is a critical regulatory event during AS.

HDAC6 in Complex with SMAR1 Deacetylates Sam68 and RegulatesAS. Next, we investigated the specific deacetylase of Sam68 thatcoordinates with SMAR1 and regulates AS. Our results revealedthat, despite SMAR1 association with both HDAC1 (21) andSam68, Sam68 failed to interact with HDAC1 (SI Appendix, Fig.S6 A–C). However, a study by Wang et al. on genome-wide oc-cupancy of HDACs revealed the enrichment of HDAC2 andHDAC6 on to the coding regions of multiple genes (34). Anassociation of SMAR1 with HDAC6 (35), but not with HDAC2(SI Appendix, Fig. S6 D and E), and an association of HDAC6 withRNA Pol II (33) and with U5 snRNA (SI Appendix, Fig. S6F) in-dicated that SMAR1 might cooperate with HDAC6 to deacetylateSam68. Consistent with our hypothesis, IP assays clearly showedthat SMAR1, as well as Sam68, associate with HDAC6 (Fig. 6A).Concurrently, sequential IP experiments revealed the coexistence ofSMAR1–HDAC6–Sam68 as a trimeric complex (Fig. 6B). Fur-thermore, the existence of this trimeric complex was confirmed bysize-exclusion chromatography using HeLa cell nuclear extracts (SIAppendix, Fig. S6G, Upper, fractions 16–26). However, the trimericcomplex was observed to be RNA-sensitive because RNaseA-treated nuclear extracts showed the dissociation of high-molecular-weight HDAC6–SMAR1 and Sam68 complexes (SI Appendix, Fig.S6G, Lower, fractions 16–26). Thus, an in silico analysis of HDAC6–SMAR1–Sam68 complex in the presence of RNA component iscompatible with the positioning of RNA at the interacting site ofSMAR1–Sam68, and both the electrostatic and solvation energiescontribute largely for the stabilization of protein–trimeric andRNA–protein–trimeric complexes (Fig. 6C and SI Appendix, TableS3). Furthermore, positioning of the RNA enhances the total energystabilization of the protein–trimeric complex (SI Appendix, TableS4). Based on the above observations, it can be inferred thatSMAR1, through HDAC6, maintains Sam68 in a deacetylated state.To prove that HDAC6 plays a decisive role in the modulation of

Sam68 acetylation and in turn CD44 gene AS, we investigated theconsequences upon HDAC6 knockdown. Semiquantitative RT-PCR analysis of CD44 splice variants in HeLa cells that werestably knocked down with HDAC6–shRNA lentivirus revealed anincrease in the incorporation of CD44 splice variants v3, v5, v6,and v7 (Fig. 6D). Convincingly, we observed enhanced acetylationof Sam68 upon HDAC6 knockdown, indicating that HDAC6maintains Sam68 in a deacetylated state (Fig. 6E). Results ofSam68 acetylation (SI Appendix, Fig. S6H) and CD44 gene AS(SI Appendix, Fig. S6I) upon knockdown of HDAC1 showed nochange in Sam68 acetylation or AS, indicating that HDAC6 isthe specific deacetylase that coordinates with SMAR1 for Sam68deacetylation. Nevertheless, to further extrapolate and establish thebiological significance of SMAR1 in regulating HDAC6-mediateddeacetylation of Sam68 (and subsequent AS of the CD44), we in-vestigated Sam68 acetylation in MDA-MB-468, T47D, and MCF-7breast cancer cells. Of significance, we observed acetylation ofSam68 in LOH-harboring MDA-MB-468 and T47D cells (Fig. 6F,lanes 2 and 3), which in the earlier results were shown to exhibitenhanced incorporation of CD44 variable exons (Figs. 2F and 2G).Sam68 remained deacetylated in MCF-7 cells that do not harborLOH. Strikingly, MDA-MB-468 and T47D cells expressed higherHDAC6 levels than MCF-7 cells (Fig. 6F). This finding highlightsthe role of SMAR1 in the maintenance of Sam68 in a deacetylatedstate. IP assays in MCF-7 cells, which express SMAR1, maintainHDAC6–SMAR1–Sam68 complex; however, the interaction wasfound to be perturbed in MDA-MB-468 and T47D (Fig. 6G, Left).Notably, the interaction of HDAC6 with Sam68 was restored upon

E3380 | www.pnas.org/cgi/doi/10.1073/pnas.1418603112 Nakka et al.

Dow

nloa

ded

by g

uest

on

June

20,

202

0

SMAR1 expression in LOH-harboring MDA-MB-468 and T47Dcells (Fig. 6G,Middle and Bottom Right). Contrastingly, MCF-7 cellsthat were knocked down for SMAR1 failed to show the interactionbetween HDAC6 and Sam68 (Fig. 6G, Upper Right). Together,these data convincingly establish that SMAR1 is a critical mediatorof the HDAC6–Sam68 interaction and that SMAR1 coordinatesHDAC6-mediated deacetylation of Sam68.

Inactivation of HDAC6 Facilitates Sam68 Acetylation and AS. Tofurther ascertain our observations of HDAC6-mediated deacetyla-tion of Sam68 and CD44 gene AS, we used an HDAC6-specificinhibitor, tubacin. We observed that treatment of HeLa cells withtubacin resulted in the enhanced incorporation of CD44 alternateexons v3–v7 (Fig. 6H). Furthermore, analysis of nuclear extractsfrom tubacin-treated HeLa cells revealed distinct acetylation ofSam68 (Fig. 6I). Considering that HDAC6 in complex with SMAR1deacetylates Sam68 and to establish that LOH-mediated Sam68acetylation is key to enhance CD44 variant exon inclusion (Figs. 6Fand 2G), we treated MCF-7 cells with tubacin and examined Sam68acetylation, as well as the CD44 variant exon inclusion. Consistentwith our assumption, we observed a favorable switch in the in-corporation of CD44 alternate exons v3–v7 (Fig. 6J), as well asSam68 acetylation (Fig. 6K). However, the variables v5 and v6showed a decrease in incorporation after 6 h of tubacin treatment.This result can be attributed to either the toxicity or altered stabilityof tubacin in these cells. Together, these results ascertain thatHDAC6 deacetylates Sam68, and inactivation of HDAC6 favorsSam68 acetylation, which in turn facilitates AS of the CD44mRNA.

DiscussionOur present study delineates a complex regulatory network under-lying the signal-dependent and -independent regulation of cotran-

scriptional AS by SMAR1 (Fig. 7). Endogenously, SMAR1 incomplex with HDAC6 associates with Sam68 and maintains itin a deacetylated state, consecutively inhibiting CD44 variant exoninclusion. Alternatively, activation of the ERK–MAPK pathway, whichis considered to be indispensable for CD44 gene AS, disrupts theHDAC6–SMAR1–Sam68 complex, enhancing Sam68 acetylation,which results in cotranscriptional inclusion of CD44 gene variantexons. Similarly, knockdown of SMAR1, which also favors Sam68acetylation independent of ERK–MAPK pathway signaling, enhancesCD44 gene AS. Together, these results emphasize the significanceof Sam68 acetylation in the context of regulated splicing.SMAR1 is a scaffold/matrix attachment region-binding protein

that regulates gene expression by binding to various promoterelements (11, 12). Here, we show that SMAR1 is enriched in thenuclear speckles, which are the sites for the association of ac-tively transcribing genes, wherein chromatin modulators, splicingfactors, snRNPs, and hnRNPs are aggregated (24). Associationof SMAR1 with the spliceosome components U1, U2, U4, U5,and U6 snRNAs provided our initial insights into the role ofSMAR1 in splice site selection. It is noted that certain other NMfactors, such as SAFB1 and SAFB2, are also reported to mod-ulate the alternative exon inclusion (36, 37).Cancer cells often express aberrant splice variants due to

altered expression of trans-factors. For instance, enhanced ex-pression of the hnRNP A1/A2 and PTB represses the inclusionof exon 9 in the pyruvate kinase M gene, which contributes to theWarburg effect in cancer cells (16). Similarly, CD44 gene ASdistinctly illustrates the role of variable exons during the me-tastasis of various tumors (22, 28). Our findings underscore thesignificance of SMAR1-mediated CD44 gene splicing in thecontext of breast cancer metastasis.

Fig. 6. HDAC6 in complex with SMAR1 deacetylates Sam68. (A) HeLa cell nuclear extracts were immunoprecipitated (IP) and immunoblotted (IB) with theindicated antibodies. (B) Sequential IP analysis of HeLa cell nuclear extracts, which were first immunoprecipitated with α-SMAR1 antibody and subsequentlywith α-HDAC6 antibody. Eluates were immunoblotted with the indicated antibodies. PARP served as a negative control. (C) Schematic representation of insilico docking analysis of HDAC6 (orange)–SMAR1 (green)–Sam68 (gray) in the presence of RNA, indicating a stable complex. (D and E) Semiquantitative RT-PCR analysis of CD44 splice variants (D) and analysis of Sam68 acetylation (E) in HeLa cells that were stably knocked down with HDAC6–shRNA. (F) Analysis ofSam68 acetylation in the nuclear extracts of the mentioned cell lines. (G) Analysis of HDAC6–SMAR1–Sam68 associations in the mentioned cell lines that weretransduced with either SMAR1-sh3 or Ad-SM and immunoprecipitated with α-Sam68 antibody. Eluates were further blotted with the indicated antibodies.(H and I) RT-PCR analysis of CD44 variant exon inclusion (H) and Sam68 acetylation (I) in tubacin-treated (5 μM) HeLa cells, for the indicated time points. (J and K)RT-PCR analysis of CD44 variant exon inclusion (J) and Sam68 acetylation (K) in tubacin-treated (5 μM) MCF-7 cells, for the indicated time points. *IgG.

Nakka et al. PNAS | Published online June 15, 2015 | E3381

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

20,

202

0

The ERK–MAPK pathway substantially enhances the inclu-sion of CD44 variant exons (23). Proteins such as Sam68,SRm160, Brm, and HP1γ that regulate CD44 pre-mRNA splic-ing are substrates of ERK kinase (18, 20, 21). Given that SMAR1knockdown enhances CD44 variant exon inclusion, we observedphosphorylation-mediated translocation of SMAR1 from nucleusto the cytoplasm upon ERK activation, thus mimicking SMAR1knockdown. As such, our data provide further insights into themechanism through which the ERK–MAPK pathway inducesCD44 variant exon inclusion. Mutation of FSM at T345A/T360A,which is phosphorylation-deficient, failed to translocate to thecytoplasm upon ERK activation, inhibiting CD44 alternate exoninclusion. ChIP and RIP assays that indicated association ofSMAR1 with the chromatin and CD44 pre-mRNA highlighted aninverse correlation in the occupancy of Brm, HP1γ, and RNA PolII upon ERK activation (19, 21). Because SMAR1 knockdownfavors CD44 alternate exon inclusion, the possibility of increasedoccupancy of Brm and HP1γ under such a scenario, which in turnenhances the occupancy of serine 5 phospho-RNA Pol II, cannotbe ruled out. This possibility might be one of the contributingfactors that enhance CD44 gene AS upon SMAR1 knockdown,because RNA Pol II with decreased elongation rate facilitates theuse of weaker splice sites (38).Despite failed activation of ERK-1/2 upon SMAR1 knock-

down, increased inclusion of CD44 alternate exons indicatesthe existence of a noncanonical pathway for CD44 gene AS.Because PTMs are known to play a key regulatory role during amajority of the cellular events, we investigated possible PTMsthat Sam68 could acquire through SMAR1. ERK-1/2 is knownto induce phosphorylation of Sam68 (18); however, failed ac-tivation of ERK-1/2 upon SMAR1 knockdown highlighted thedispensability of Sam68 phosphorylation. The association of SMAR1with HDAC1 was earlier reported to regulate p53 transactivationpotential via deacetylation/acetylation (12), which is in agreementwith the concept that PTMs such as acetylation provide a regu-latory layer during many cellular events (39). Our present workshows that endogenous Sam68 is maintained in a deacetylatedstate and knockdown of SMAR1 enhances Sam68 acetylation. A

stable association of SMAR1 with the lysine-enriched KH domainof Sam68 strongly supports the concept that these lysines aremasked from being acetylated during the interaction. ERK–MAPK signal-induced disruption of SMAR1–Sam68 associationfacilitates Sam68 acetylation that was abated in the presence ofERK inhibitor U0126. It is noted that extracellular stimuli, whichinduce protein phosphorylation, mark a first wave of PTMs thatcan be further converted to acetylation-based events (38). Thus,our study is in agreement with this hypothesis and depicts thatERK-1/2–mediated phosphorylation of SMAR1 and Sam68 (18) isa prerequisite to disrupt their association, resulting in Sam68acetylation during signal-dependent alternate exon inclusion. Astudy using various deletion and mutant constructs of Sam68distinctly highlighted the role of Sam68 KH-domain lysine resi-dues and the requirement for Sam68 acetylation during AS ofCD44. Because the pattern of acetylation alters the mechanisticimpact, acetylation at a cluster of lysine residues results in theformation of charged patches (39), which in turn might enhancethe affinity of Sam68 toward RNA. Together, Sam68 acetyla-tion favors the inclusions of exons with weak splice sites andthus might be offering a fail-safe mechanism of CD44 gene ASduring embryogenesis.Deacetylation of Sam68 by SMAR1 in cooperation with

HDAC6 highlights the fact that the SMAR1–HDAC1 pool,which is mostly localized at gene promoters, is distinct fromSMAR1–HDAC6–Sam68 pool that localizes to the CD44 genebody and regulates AS. The basis for the specificity of HDAC6association with Sam68 can be attributed to its differential oc-cupancy. HDAC1 and HDAC3 are mostly localized to pro-moters, whereas HDAC2 and HDAC6 are enriched on the genebody as well as promoters (35). The association of HDAC6 withRNA Pol II (35) further contributes to the HDAC6-mediateddeacetylation of Sam68. In this case, it is tempting to speculatethat HDAC6 could also modulate the elongation rate of RNAPol II via deacetylation of additional transcriptional elongationand splicing factors that are bound to the CTD tail, resulting inregulated splicing through an alternative mechanism.This previously unidentified molecular mechanism of regu-

lated AS via Sam68 acetylation adds an extra dimension in un-derstanding cancer prognosis, especially in breast cancer. Ourdata, to our knowledge, for the first time show that LOH at16q24.3 locus in breast cancer cells is associated with low levelsof SMAR1 and enhanced Sam68 acetylation, which in turn fa-vors CD44 variant exon inclusion and metastasis. MCF-7 cells,without LOH, express SMAR1 and maintain Sam68 in adeacetylated state, resulting in significantly low levels of CD44variable exons. Despite the presence of relatively high levels ofHDAC6 in MDA-MB-468 and T47D compared with MCF-7, ourwork suggests that Sam68 is maintained in an acetylated state inthese cells due to their lack of SMAR1, which is indispensablefor the colocalization of HDAC6 and Sam68. Despite decreasedincorporation of proximal variant exons upon restoration ofSMAR1 expression in LOH-harboring cells, the distal exonsremain unaltered. This observation highlights that there aresome other contributing factors encoded from the 16q24 locus,which might modulate the incorporation of CD44 distal variantexons. Alternatively, enhanced interaction of Sam68 with the U5snRNA upon ERK activation (19) and endogenous association ofSMAR1 with U5 snRNA indicates that in addition to modulationof Sam68 acetylation, cooperation between SMAR1 and HDAC6might act as “roadblock” to sequester U5 snRNA from Sam68,thus preventing the splice site selection.In summary, our study elucidated a previously unappreciated

regulatory mechanism used to control alternate exon use, wherebythe NM-associated protein SMAR1 and chromatin-modifyingenzyme HDAC6 modulate the acetylation of splicing factorslike Sam68 to modulate AS in response to extracellular cues(Fig. 7). To our knowledge, we show for the first time that

Fig. 7. A model representing SMAR1 phosphorylation and Sam68 acety-lation mediated regulation of AS. (Upper) Endogenously, SMAR1 in com-plex with HDAC6 maintains Sam68 in a deacetylated state. (Lower Left) In asignal-dependent manner of regulation, ERK-1/2 kinase phosphorylatesSMAR1 and Sam68, thus disrupting HDAC6–SMAR1–Sam68 complex,resulting in SMAR1 export to the cytoplasm. SMAR1 translocation tocytoplasm induces Sam68 acetylation and enhances CD44 gene AS. (LowerRight) Conversely, in a signal-independent manner, as observed uponknockdown of SMAR1 or LOH at Chr.16q24.3 locus, HDAC6 fails to main-tain Sam68 in deacetylated state with a concomitant increase in CD44 genevariant exon inclusion.

E3382 | www.pnas.org/cgi/doi/10.1073/pnas.1418603112 Nakka et al.

Dow

nloa

ded

by g

uest

on

June

20,

202

0

HDAC6 plays a pivotal role in regulated splicing of CD44 viaSam68 deacetylation. Our findings reveal that knockdown orreduced levels of SMAR1 promotes cancer cell invasion andmetastasis, whereas overexpression restricts cell invasion. Thisfinding suggests that targeting Sam68 acetylation may providealternative therapeutic strategies during breast cancer metastasis.Thus, we provide a previously unidentified mechanism of regula-tion of AS in the etiology of human breast cancer.

Materials and MethodsAdditional materials and methods on plasmids, antibodies, chromatinimmunoprecipitation, and RNA immunoprecipitation, as well as on themodeling and docking studies, are provided in SI Appendix, SI Materialsand Methods.

Cells and Cell Culture. HEK 293T, HeLa, and MCF-7 cells were cultured inDMEM supplemented with 10% (vol/vol) FBS, 100 U/mL penicillin, and 100μg/mL streptomycin (Invitrogen) at 37 °C and 5% CO2. MDA-MB-468 andT47D cells were grown in L-15 and RPMI medium under similar conditions.

Adenovirus and Lentivirus. SMAR1 adenovirus was used as described (12).For SMAR1, HDAC6, and Sam68 lentivirus generation, HEK 293T cellswere cotransfected with pSPAX, pMD2.G, SMAR1–shRNA (Clone ID:V2LHS_174233; V3LHS_374011; V3LHS_374008; RHS4346 for nonsilencing;Open Biosystems), HDAC6–shRNA [Clone ID: TRCN0000004839 (sh1);TRCN0000004840 (sh2)], or Sam68–shRNA (Clone ID: TRCN0000000044–TRCN0000000047; TRCN0000000104). Indicated cell lines were transducedwith a 1:1 mix of viral supernatant and growth medium. Stable cell lineswere selected with 1.5 μg/mL puromycin (Sigma).

RNA Isolation and Semiquantitative RT-PCRs. RNA was isolated by using TRIzol(Invitrogen). Reverse-transcription was performed with 2 μg of RNA by usingM-MuLV reverse transcriptase (Invitrogen), Oligo dT (IDT), and RNase OUT

(Invitrogen). PCRs were performed in 25 μL of reaction mixture by using TaqDNA polymerase (Chromus Biotech).

In Situ NM Staining. Nuclear matrices were prepared as mentioned in a well-established protocol from Nickerson et al. (25). IF analysis of the isolatedmatrices was performed by using a standard protocol after fixation withparaformaldehyde (12).

Whole-Cell Lysates and Nuclear and Cytoplasmic Fractionations. Indicated cellswere lysed in whole-cell lysis buffer (50 mM Tris·Cl, pH 7.4, 1% Nonidet P-40,250 mM NaCl, 0.5 mM Na3VO4) supplemented with complete protease-inhibitor mixture (Roche). For nuclear and cytoplasmic fractionations, cellwere incubated in cytosolic buffer (10 mM Hepes, pH 7.6, 10 mM KCl, 1 mMDTT, 1 mM LiCl, 0.5 mM MnCl2, and 1 mM PMSF) containing 1% NonidetP-40 for 15 min on ice. Cells were vortexed vigorously for 10 s and spun at12,000 × g for 30 s. Supernatant was stored for cytosolic extract, pellets werewashed twice with cytosolic buffer, and nuclei were lysed in buffer containing20 mM Hepes (pH 7.9), 300 mM NaCl, and 1.5 mM MgCl2.

α-Acetyl-Lysine Acetylation Assay. For assessing the acetylation on indicatedproteins, 500 μg of nuclear extract was incubated with 0.5 μg of α-acetyl-lysine antibody (Millipore) at 4 °C with slight mixing for 4 h. Subsequently,protein A/G beads were added and further incubated for 2 h. Immuno-complexes were washed three times with 500 μL of IP buffer and furthereluted at 95 °C in IP buffer with SDS loading dye. The eluates were loaded onto the SDS/PAGE and immunoblotted with indicated antibodies.

ACKNOWLEDGMENTS. We thank Prof. Harald König for providing CD44 pET-v5-Luc plasmid and pcDNA3.1 myc-Sam68 plasmids; and Prof. Juan ValcárcelJuárez for the FAS minigene construct. We also thank Dr. Shekhar C. Mande,Director, National Centre for Cell Science. This work was supported by theCouncil of Scientific and Industrial Research, India; the Department of Biotech-nology, India; and the Indian Council of Medical Research, Government of India(S. Chattopadhyay); and Canadian Institutes of Health Research (F.J.D.).

1. Wang ET, et al. (2008) Alternative isoform regulation in human tissue transcriptomes.Nature 456(7221):470–476.

2. Delsert C, Morin N, Klessig DF (1989) Cis-acting elements and a trans-acting factoraffecting alternative splicing of adenovirus L1 transcripts. Mol Cell Biol 9(10):4364–4371.

3. Allemand E, Batsché E, Muchardt C (2008) Splicing, transcription, and chromatin: Aménage à trois. Curr Opin Genet Dev 18(2):145–151.

4. Zeitlin S, Parent A, Silverstein S, Efstratiadis A (1987) Pre-mRNA splicing and the nu-clear matrix. Mol Cell Biol 7(1):111–120.

5. Berezney R, Mortillaro MJ, Ma H, Wei X, Samarabandu J (1995) The nuclear matrix: Astructural milieu for genomic function. Int Rev Cytol 162A:1–65.

6. Chattopadhyay S, Kaul R, Charest A, Housman D, Chen J (2000) SMAR1, a novel, al-ternatively spliced gene product, binds the Scaffold/Matrix-associated region at the Tcell receptor beta locus. Genomics 68(1):93–96.

7. Powell JA, et al. (2002) Sequencing, transcript identification, and quantitative geneexpression profiling in the breast cancer loss of heterozygosity region 16q24.3 revealthree potential tumor-suppressor genes. Genomics 80(3):303–310.

8. Birot A, et al. (2000) Identification and molecular analysis of BANP. Gene 253(2):189–196.

9. Kaul R, et al. (2003) Direct interaction with and activation of p53 by SMAR1 retardscell-cycle progression at G2/M phase and delays tumor growth in mice. Int J Cancer103(5):606–615.

10. Jalota-Badhwar A, et al. (2007) SMAR1-derived P44 peptide retains its tumor sup-pressor function through modulation of p53. J Biol Chem 282(13):9902–9913.

11. Rampalli S, Pavithra L, Bhatt A, Kundu TK, Chattopadhyay S (2005) Tumor suppressorSMAR1 mediates cyclin D1 repression by recruitment of the SIN3/histone deacetylase1 complex. Mol Cell Biol 25(19):8415–8429.

12. Sinha S, et al. (2010) Coordinated regulation of p53 apoptotic targets BAX and PUMAby SMAR1 through an identical MAR element. EMBO J 29(4):830–842.

13. Singh K, et al. (2007) p53 target gene SMAR1 is dysregulated in breast cancer: Its rolein cancer cell migration and invasion. PLoS ONE 2(7):e660.

14. David CJ, Manley JL (2010) Alternative pre-mRNA splicing regulation in cancer:Pathways and programs unhinged. Genes Dev 24(21):2343–2364.

15. Christofk HR, et al. (2008) The M2 splice isoform of pyruvate kinase is important forcancer metabolism and tumour growth. Nature 452(7184):230–233.

16. Chen M, David CJ, Manley JL (2012) Concentration-dependent control of pyruvatekinase M mutually exclusive splicing by hnRNP proteins. Nat Struct Mol Biol 19(3):346–354.

17. Blaustein M, Pelisch F, Srebrow A (2007) Signals, pathways and splicing regulation. IntJ Biochem Cell Biol 39(11):2031–2048.

18. Matter N, Herrlich P, König H (2002) Signal-dependent regulation of splicing viaphosphorylation of Sam68. Nature 420(6916):691–695.

19. Batsché E, Yaniv M, Muchardt C (2006) The human SWI/SNF subunit Brm is a regulatorof alternative splicing. Nat Struct Mol Biol 13(1):22–29.

20. Cheng C, Sharp PA (2006) Regulation of CD44 alternative splicing by SRm160 and itspotential role in tumor cell invasion. Mol Cell Biol 26(1):362–370.

21. Saint-André V, Batsché E, Rachez C, Muchardt C (2011) Histone H3 lysine 9 trime-thylation and HP1γ favor inclusion of alternative exons. Nat Struct Mol Biol 18(3):337–344.

22. Günthert U, et al. (1991) A new variant of glycoprotein CD44 confers metastaticpotential to rat carcinoma cells. Cell 65(1):13–24.

23. Weg-Remers S, Ponta H, Herrlich P, König H (2001) Regulation of alternative pre-mRNA splicing by the ERK MAP-kinase pathway. EMBO J 20(15):4194–4203.

24. Zhong XY, Wang P, Han J, Rosenfeld MG, Fu XD (2009) SR proteins in vertical in-tegration of gene expression from transcription to RNA processing to translation.MolCell 35(1):1–10.

25. Nickerson JA, Krockmalnic G, Wan KM, Penman S (1997) The nuclear matrix revealedby eluting chromatin from a cross-linked nucleus. Proc Natl Acad Sci USA 94(9):4446–4450.

26. Izquierdo JM, et al. (2005) Regulation of Fas alternative splicing by antagonistic ef-fects of TIA-1 and PTB on exon definition. Mol Cell 19(4):475–484.

27. Kaufmann M, et al. (1995) CD44 variant exon epitopes in primary breast cancer andlength of survival. Lancet 345(8950):615–619.

28. Yae T, et al. (2012) Alternative splicing of CD44 mRNA by ESRP1 enhances lung col-onization of metastatic cancer cell. Nat Commun 3:883.

29. Callen DF, et al. (2002) Defining regions of loss of heterozygosity of 16q in breastcancer cell lines. Cancer Genet Cytogenet 133(1):76–82.

30. Lewis TS, Shapiro PS, Ahn NG (1998) Signal transduction through MAP kinase cas-cades. Adv Cancer Res 74:49–139.

31. Shin C, Feng Y, Manley JL (2004) Dephosphorylated SRp38 acts as a splicing repressorin response to heat shock. Nature 427(6974):553–558.

32. Heyd F, Lynch KW (2010) Phosphorylation-dependent regulation of PSF by GSK3controls CD45 alternative splicing. Mol Cell 40(1):126–137.

33. Paronetto MP, Achsel T, Massiello A, Chalfant CE, Sette C (2007) The RNA-bindingprotein Sam68 modulates the alternative splicing of Bcl-x. J Cell Biol 176(7):929–939.

34. Wang Z, et al. (2009) Genome-wide mapping of HATs and HDACs reveals distinctfunctions in active and inactive genes. Cell 138(5):1019–1031.

35. Chaudhary N, et al. (2014) SMAR1 coordinates HDAC6-induced deacetylation of Ku70and dictates cell fate upon irradiation. Cell Death Dis 5:e1447, 10.1038/cddis.2014.397.

36. Nayler O, et al. (1998) SAF-B protein couples transcription and pre-mRNA splicing toSAR/MAR elements. Nucleic Acids Res 26(15):3542–3549.

37. Sergeant KA, et al. (2007) Alternative RNA splicing complexes containing the scaffoldattachment factor SAFB2. J Cell Sci 120(Pt 2):309–319.

38. de la Mata M, et al. (2003) A slow RNA polymerase II affects alternative splicing invivo. Mol Cell 12(2):525–532.

39. Yang XJ, Seto E (2008) Lysine acetylation: Codified crosstalk with other post-translational modifications. Mol Cell 31(4):449–461.

Nakka et al. PNAS | Published online June 15, 2015 | E3383

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

20,

202

0