Tropomyosin isoforms have specific effects on the transcriptome of undifferentiated and differentiated B35 neuroblastoma cells

Tropomyosins, a family of actin‐associated proteins, bestow actin filaments with distinct biochemical and physical properties which are important for determining cell shape and regulating many cellular processes in eukaryotic cells. Here, we used RNA‐seq to investigate the effect of four tropomyosin isoforms on gene expression in undifferentiated and differentiated rat B35 neuroblastoma cells. In undifferentiated cells, overexpression of tropomyosin isoforms Tpm1.12, Tpm2.1, Tpm3.1, and Tpm4.2 differentially regulates a vast number of genes, clustering into several gene ontology terms. In differentiated cells, tropomyosin overexpression exerts a much weaker influence on overall gene expression. Our findings are particularly compelling because they demonstrate that tropomyosin‐dependent changes are attenuated once the cells are induced to follow a defined path of differentiation. Database Sequence data for public availability are deposited in the European Nucleotide Archive under the accession number PRJEB24136.

Tropomyosins, a family of actin-associated proteins, bestow actin filaments with distinct biochemical and physical properties which are important for determining cell shape and regulating many cellular processes in eukaryotic cells. Here, we used RNA-seq to investigate the effect of four tropomyosin isoforms on gene expression in undifferentiated and differentiated rat B35 neuroblastoma cells. In undifferentiated cells, overexpression of tropomyosin isoforms Tpm1.12, Tpm2.1, Tpm3.1, and Tpm4.2 differentially regulates a vast number of genes, clustering into several gene ontology terms. In differentiated cells, tropomyosin overexpression exerts a much weaker influence on overall gene expression. Our findings are particularly compelling because they demonstrate that tropomyosin-dependent changes are attenuated once the cells are induced to follow a defined path of differentiation.

Database
Sequence data for public availability are deposited in the European Nucleotide Archive under the accession number PRJEB24136.
The actin cytoskeleton provides structural support and is essential for the cellular morphogenesis of eukaryotic cells and many physiological processes including cell motility, endocytosis, transport of organelles, apoptosis, and the maintenance of correct directional signal transmission in neurons [1,2]. Several actin-binding proteins are known to bestow the actin cytoskeleton with a high degree of structural and functional diversity by regulating actin filament turnover, filament branching, bundling, and cross-linking of individual filaments [3]. Tropomyosins (Tpms) are a large family of actin-binding proteins that are generated by alternative splicing from four different genes (Tpm1, Tpm2, Tpm3, and Tpm4). They not only define distinct filamentous actin populations in different cell populations and subcellular compartments, but also regulate the access of other actin-binding proteins to the actin filament [4,5]. Tpms have been identified to play an important role in a range of cellular processes with changes in protein expression, including the regulation of cell transformation [6,7], ERK-mediated proliferation [8], insulin-stimulated GLUT4 transport [9], and anoikis [10]. The regulation of these diverse cellular processes suggests that an altered composition of Tpm expression in eukaryotic cells will lead to changes in cellular pathways at a global level rather than impacting on interactions with only a select number of established interaction partners. In this study, we aimed to test whether the level of expression of Tpms determines the cell transcriptome in an isoform-specific manner and whether these changes are dependent on the differentiation stage of these cells. Rat B35 neuroblastoma cells [11,12] have extensively been used to study cellular processes of eukaryotic cells, including neuronal morphogenesis [11,13], cell motility [14,15], vesicular trafficking [16], and apoptosis [17,18]. Our group has used the B35 cell system previously to study the role of different Tpm isoforms in neuronal cell morphogenesis [13,19]. Stable rat B35 neuroblastoma cell lines overexpressing the Tpm isoforms Tpm1.12, Tpm2.1, Tpm3.1, and Tpm4.2, from each of the Tpm genes 1-4, were previously generated [13,17,19]. Neuronal differentiation requires the coordinated reorganization of the actin cytoskeleton to facilitate the sprouting and elongation of neurites, which ultimately form axons and dendrites. Our previous studies showed that the overexpression of Tpms in B35 cells was not only sufficient to induce the formation of neurites, but also differentially influenced neurite branching and extension in differentiating B35 cells [13].
In this study, we wanted to understand how the transcriptome of B35 cells is altered in response to the overexpression of different Tpm isoforms. The analysis of the transcriptome does not only provide insight into which genes are being expressed, and at what level, but also sheds light onto how the distinct expression of genes could alter essential cellular pathways and mechanisms of cellular morphogenesis.
We employed Illumina RNA-seq which has been demonstrated to yield higher sensitivity, deeper resolution, and greater reproducibility when compared to conventional genomic methods such as microarray analysis [20]. Another advantage of RNA-seq is the ability to identify novel transcripts and splice variants, which is not possible using microarray analysis. To identify gene clusters of biological pathways, which may reveal hidden patterns that regulate specific biological processes, differentially expressed genes (DEGs) were further analyzed via clusterProfiler. The main advantage of clusterProfiler is the application of both biological term classification and enrichment analysis to gene cluster analysis, thereby providing greater insight into understanding higher order functions in biological systems [21]. We found a large number of changes in genes in undifferentiated B35 cells, compared with differentiated cells, where Tpm overexpression appeared to have less influence on gene expression. Differentially expressed genes (DEGs) found in undifferentiated cells could be grouped into a range of different pathways, including pronounced changes in actin-binding pathways. However, in differentiated cells, only one isoform, Tpm3.1, had DEGs that generated gene ontology (GO) terms, suggesting limited pathway commonality in differentiated Tpm1.12, Tpm2.1, and Tpm4.2 B35 cells. Our results are consistent with the overall hypothesis that different Tpm isoforms generate distinct actin filament populations that are controlling key pathways of cellular function in eukaryotic cells.

RNA preparation and sequencing
Total RNA was isolated from three biological replicates from each transgenic cell line and empty vector controltransfected cells, using RNeasy Mini Kit (Qiagen, Hilden, Germany) followed by RNase-free DNase treatment to remove traces of genomic DNA. The Agilent 2100 Bioanalyzer RNA Nano Chip was used to assess the RNA quality of the total RNA. The RNA integrity number (RIN) values ranged between 6.0 and 7.0. RNA was generated using Illumina TruSeq RNA sample preparation for poly(A) RNA and sequenced using paired-end 100-bp reads on Illumina HiSeq2500.
Read mapping, transcript assembly, and comparative analysis rat reference genome (Rnor_6.0) using TopHat (version 2.1.1) as described previously [22]. The BAM files from TopHat were then fed into Cufflinks (version: 2.2.1) on Galaxy server for transcript assembly and expression level calculation. Annotation files from Ensembl rat genome assembly Rnor_6.0 were used as reference annotation. Next, Cufflinks-assembled transcripts were merged together using Cuffmerge on Galaxy server using uploaded reference annotation. Differential expression analysis was performed using Cuffdiff. Cuffdiff utilizes the merged files from Cuffmerge along with the original alignment files produced from TopHat to calculate expression levels and determine their statistical significance and whether the transcripts are differentially expressed. Genes with an FPKM ≥ 1 in at least one condition were considered as expressed.

Pathway analysis
The ClusterProfiler program [21] was used through the platform R studio (https://www.rstudio.com/) version 1.0.143 to undertake the pathway analysis of the significant differentially expressed linear RNA. Pathway analysis included matching the input list of annotated genes expressing linear RNA to their ENTREZ ID. These genes were then matched to their respective GO terms, and an overrepresentation test, based on hypergeometric distribution of the GO terms, was performed to identify enriched GO terms and subsequently measure the statistical significance of each enriched GO term. For this pathway analysis, the allocated ontology term option used was biological processes. The scripts used for the pathway analysis were obtained from https://bioconductor.org/packages/devel/ bioc/vignettes/clusterProfiler/inst/doc/clusterProfiler.html.

Results and Discussion
Overexpression of tropomyosin isoforms differentially affects gene expression in undifferentiated and differentiated B35 neuroblastoma cells In undifferentiated cells, the overexpression of specific Tpm isoforms alters the expression of thousands of genes, with a surprising degree of difference in effected genes between isoforms (Fig. 1A). Cells with increased levels of Tpm1.12 had changes in the expression of over 4000 genes, the highest number of observed DEGs out of the four isoforms investigated. Approximately 45% of these DEGs were specific to Tpm1.12overexpressing cells. Tpm2.1, Tpm3.1 and Tpm4.2 each had between 2000 and 2400 DEGs with isoform specificity ranging between 15 and 20%.
In differentiated cells, numbers of DEGs were overall greatly reduced compared with undifferentiated cells (Fig. 1B). The highest number of DEGs observed was 160 and again associated with Tpm1.12 overexpression. Differentiated cells, overexpressing Tpm3.1 and Tpm4.2, had similar numbers of DEGs, 120 and 136, respectively. Tpm2.1 had, with 89, the fewest number of DEGs. In terms of specificity, once differentiated, there was also reduced overlap between isoforms. Tpm3.1 showed relatively high specificity, with over 60% of observed DEGs being unique to cells, overexpressing this isoform. Tpm2.1 and Tpm4.2 isoforms had specificity ranging between 49% and 43%, respectively, much greater than observed in undifferentiated cells. Tpm1.12 was the only isoform that showed similar specificity of approximately 45%, regardless of whether the cells were differentiated or not. Overall, these results indicate that Tpms exert a broad influence on gene expression in undifferentiated cells. However, once cells undergo differentiation, gene expression appears more tightly regulated with tropomyosin having a weaker but more isoform-specific influence.
Undifferentiated Tpm1.12-, Tpm3.1-, and Tpm4.2overexpressing B35 cells display some similarity in gene regulation In undifferentiated cells, there was some commonality between the different cell lines (Table 1). In undifferentiated Tpm1.12-overexpressing cells, Ntm, which encodes for a protein that inhibits neurite outgrowth [23], was the most downregulated gene. Within undifferentiated Tpm2.1 cells, Bnc1 was among the top upregulated genes. Bnc1 is a transcription factor that plays a role in the expression of genes involved in cellular differentiation and proliferation [24]. In Tpm3.1-and Tpm4.2-overexpressing B35 cells, the gene encoding for neurofilament light chain, Nefl, was among the top downregulated genes, suggesting crosstalk between the actin and intermediate filament systems. Together, this indicates the greatest commonality between Tpm3.1-and Tpm4.2-overexpressing cells, both in the undifferentiated and in differentiated states. The similarity in effects of Tpm3.1 and Tpm4.2 has also been observed in other systems. In vitro single filament assays have shown that Tpm3.1 and Tpm4.2 often localize to the same F-actin populations and have similar, rapid association with less cooperative binding to F-actin [25]. This was analyzed using FRAP experiments where the fluorescence recovery of sfGFP (super-folder GFP) fusions of Tpm3.1 and Tpm4.2 was much more rapid, when compared to higher molecular weight Tpm isoforms. These isoforms also stimulate the ATPase activity of nonmuscle myosin IIa and, in vitro, do not efficiently protect filaments from the severing action of ADF/cofilin [25].

2overexpressing B35 cells exhibit upregulation of similar genes
Arhgap25 was among the top upregulated genes in differentiated Tpm1.12-, Tpm3.1-, and Tpm4.2-overexpressing cells, but not in Tpm2.1-overexpressing cells (Table 2). Interestingly, Arhgap25 encodes negative regulators of Rho-GTPases, which are involved in actin remodeling, cell polarity, and migration [26]. The upregulation of Arhgap25, only in response to cellular differentiation, suggests that in the differentiated Tpm1.12, Tpm3.1, and Tpm4.2 cells, Rho-GTPases such as Rac or Cdc42 need to be silenced in order to restrict lamellipodia and filopodia formation [27]. By contrast, differentiation of Tpm2.1 cells resulted in increases in genes that encode for proteins, such as Kcnn4 and Qprt, which alter cell membrane polarization (Table 2).

DEGs arising from the overexpression of tropomyosin isoforms cluster into various pathways
In undifferentiated cells, cluster analysis of DEGs from the overexpression of the four tropomyosin isoforms, used in this study, results in the emergence of multiple pathways. In differentiated cells, Tpm3.1 is the only isoform where DEGs cluster into pathways.

Tpm1.12
The DEGs arising from Tpm1.12 overexpression in undifferentiated cells cluster into 15 GO terms ( Fig. 2A). Of these GO terms, ribosomal, RNA, and ubiquitin related are among the pathways holding the highest statistical significance. Ribosomal proteins are most commonly known to be involved in protein synthesis but have also been shown to exert extraribosomal functions including immune signaling and development of various cell types [28] as well as being implicated in various cancers including glioblastoma, gastrointestinal, prostate, and lung [29][30][31][32]. In this study, expression levels of many genes, encoding ribosomal proteins (Rp), were found to be differentially regulated in both Tpm1.12-and Tpm2.1-overexpressing cells including Rpl9, Rpl13, Rpl19, and Rpl22.
In the GTPase binding GO term, the Pfn1 (profilin 1) is found to be upregulated. Pfn1 is involved in actin nucleation, mediating the exchange of ADP to ATP on monomeric actin. This process primes the actin monomer to be incorporated into the growing 'barbed end' of actin filaments, resulting in actin filament polymerization [33]. Overexpression of Tpm1.12 has previously been shown to promote neurite branching and filopodia formation [13], two processes that rely on polymerization of actin filaments. The upregulation of Pfn1 is a potential mechanism by which Tpm1.12 is able to enhance neurite branching and filopodia formation in the cell.

Tpm2.1
Differentially expressed genes from undifferentiated cells overexpressing Tpm2.1 cluster into 7 GO terms (Fig. 2B), rRNA binding, structural constituent of ribosome, threonine-type endopeptidase activity, threonine-type peptidase activity, large ribosomal subunit rRNA binding, protein C-terminal binding, and growth factor binding. Within the protein  C-terminal-binding protein pathway are genes, associated with apoptosis. Tpm2.1 is known to have tumorsuppressing properties in breast and urinary bladder cancers [10,34,35] and has recently been shown to increase cell sensitivity to apoptosis by detachment from the extracellular matrix, referred to as anoikis, and through the modulation of various apoptosisinducing proteins [17].
Cdc37 is a cochaperone protein to heat-shock protein 90 (HSP90). Cdc37 facilitates the interaction of protein kinases with HSP90 by arresting the ATPase cycle of HSP90 and inducing an open conformational state that promotes client protein interaction [43]. HSP90 in collaboration with Cdc37 has been suggested to promote the proliferation and survival of cancer cells through dysregulation of oncogenes [44], and silencing Cdc37 has been shown to enhance cell cycle arrest and apoptosis [45][46][47].
In our study, Tpm2.1 overexpression results in the upregulation of Dapk3 and the downregulation of Cdc37 genes. The ability of Tpm2.1 to modulate the expression of these genes may help to shed light onto the tumor-suppressing and proapoptotic characteristics of Tpm2.1.

Tpm4.2
Differentially expressed genes that were observed in response to the overexpression of Tpm4.2 cluster into six GO terms, protein C-terminal binding, growth factor binding, actin binding, extracellular matrix binding, glycosaminoglycan binding, and integrin binding (Fig. 2C).
The integrin-binding pathway includes genes for integrin subunits. Integrins function in cell surface adhesion and signaling, acting as a mediator between the intracellular actin cytoskeleton and the extracellular matrix [48]. Gene expression of Itga7 (Integrin subunit a7) is found to be upregulated in response to Tpm4.2 overexpression in undifferentiated B35 cells. Itga7 and b-subunits form heterodimeric integrin receptors that bind laminin and regulate cell adhesion [49]. Interestingly, Itga7 has been found to play a role in lamellipodia formation [50], neuritogenesis of cortical neurons [51], and the regeneration of subpopulations of injured sensory neurons [52]. In a previous study by Curthoys et al. [13], overexpression of Tpm4.2 in undifferentiated B35 neuroblastoma cells led to an increase in neurite branching, filopodia formation, and growth cone size. It is plausible that this observed phenotype may be promoted by the ability of Tpm4.2 to modulate genetic expression of genes such as Itga7. Furthermore, overexpression of Tpm4.2 increases protein levels of fascin [13], an actin-binding protein recruited to actin bundles during filopodia and lamellipodia formation [53].
Fbln5 is an extracellular matrix protein involved cell adhesion. In endothelial cells, Fbln5 has been shown to operate via integrin binding [54] enhancing cell attachment and adhesion as well as decreasing proliferation [55]. Interestingly, Fbln5 is reported to have context-dependent oncogenic and tumor-suppressing roles [56]. Increased levels of Fbln5 reduce cell migration and invasion in ovarian and breast cancer [57,58].
In the current study, overexpression of Tpm3.1 upregulates the expression levels of both Fbnl5 and Itga4. Previously, Tpm3.1 has been shown to recruit myosin IIa into stress fibers, stabilize actin filaments, and slow cell migration [9,14,19]. It is plausible then that Tpm3.1 works in conjunction with Fbln5, Itga4, and myosin IIa to enhance cell stability and adhesion.

Tpm3.1 overexpression in differentiated B35 cells
Interestingly, in differentiated cells, Tpm3.1 was the only isoform that generated GO terms from cluster analysis of DEGs (Fig. 2E). The GO terms observed include growth factor-binding, glycosaminoglycanbinding, water transmembrane transporter activity, RNA polymerase II transcription coactivator activity GO terms (Fig. 2E). The greatest difference was observed in the GO terms growth factor binding and glycosaminoglycan binding. Differentiated Tpm3.1 cells have increased levels of fibroblast growth factor receptors 1 and 2 (Fgfr1 and Fgfr2, respectively). Increased expression of Fgfr1 is associated with parathyroid carcinoma [61]. Tpm3.1 is the predominant Tpm isoform in numerous cancers [7]. Taken together, Fgfr1 and Tpm3.1 could be working synergistically during cancer development. Furthermore, Fgfr2 expression is upregulated in differentiating mouse podocytes in vitro, due to the reorganization of the actin cytoskeleton and extension of their cellular processes [62]. Therefore, as B35 cells overexpressing Tpm3.1 undergo differentiation, they may require increased levels of Fgfr1 and Fgfr2 to accommodate the reorganization of the actin cytoskeleton.

Actin-binding pathway
Undifferentiated B35 cells overexpressing either Tpm3.1 or Tpm4.2 are found to differentially regulate genes that cluster into the actin-binding GO term ( described in immune cells, with recent evidence also suggesting a role in neuronal cells [63]. Interestingly, CORO1A is associated with polarized cells, suggesting that it is required for active actin cytoskeleton rearrangement and protein synthesis [63]. Furthermore, members of the type 1 Coronin family typically localize to protrusions of cell membranes, where they modulate actin dynamics [64]. Therefore, the increased neurite outgrowth and axonal extension known to be associated with increased Tpm3.1 expression [65][66][67] may be assisted by increased Coro1a levels. Further investigations are needed to elucidate the interaction between Coro1a and Tpm3.1/Tpm4.2.
Despite their commonalities, Tpm3.1 overexpression and Tpm4.2 overexpression also result in differential changes in genes from the GO terms actin-binding pathway. Whereas the overexpression of Tpm4.2 results in a decrease in Marcksl1, this is not observed for Tpm3.1. Marcksl1 is an actin cross-linking protein, which undergoes phosphorylation to bundle and stabilize F-actin [68]. The inhibition of Marcksl1 phosphorylation causes an increase in actin mobility, compromised filopodia formation, enhanced lamellipodium formation, and cell migration [68]. Therefore, the increase in filopodia number previously observed in Tpm4.2-overexpressing cells [13] may be partly attributed to a concomitant decrease in Marcksl1.
Although the overexpression of Tpm1.12 did not lead to DEGs clustering to form an actin-binding pathway, both Coro1a and Marcksl1 genes are found to be differentially expressed. As observed in Tpm4.2 overexpression, Tpm1.12 overexpression upregulates Coro1a and downregulates Marcksl1 ( Table 3). The similar changes, induced by Tpm4.2 and Tpm1.12, are consistent with the study by Curthoys et al. [13], where overexpression of Tpm1.12 was found to have similar effects on cell morphology as Tpm4.2, with an increase in filopodia and neurite branching.

Potential mechanisms for transcriptional changes caused by Tpm expression
Our data demonstrate that Tpm expression leads to isoform-dependent transcriptional changes in eukaryotic cells. The mechanisms by which Tpms lead to these transcriptional changes are still unknown. A role for actin in transcriptional regulation has been well established and reviewed previously [69,70]. A potential mechanism of transcriptional regulation by Tpms may be via Tpm isoform-dependent regulation of actin turnover which is important for the localization of transcriptional regulators. Maintaining the balance between the globular (G) and the filamentous (F) pool of cytoplasmic actin has been implicated in the translocation several transcriptional regulators, including the homeobox transcription factor PREP2 [71], the transcriptional repressor YY1 [72], and the transcriptional coactivator MAL [73]. Regulation of the cytoplasmic pool of G-and F-actin by the expression levels of different Tpms could therefore impact the transcriptome as a consequence of altered translocation of these transcriptional regulators. Tpms have also been found in the nucleus [74,75], where they may directly regulate transcription of the genes that group in various pathways identified in our study.

Conclusion
In conclusion, overexpression of Tpm isoforms in undifferentiated B35 neuroblastoma cells leads to the differential expression of a plethora of genes. However, once differentiated, Tpm isoforms have a weaker influence on gene expression. In undifferentiated cells, DEGs cluster into various pathways that show some similarity between isoforms. In differentiated cells, the overexpression of Tpm3.1 was the only isoform to generate GO terms. Many of the observed DEGs are involved in cellular activities that relate to Tpm functions, suggesting that Tpms can modulate cell dynamics and properties by regulating specific genes. Overall, this study highlights the ability of Tpm isoforms to regulate patterns of gene expression in an isoform-specific manner and aligns with their capacity to similarly control actin filament function.