Human and mouse activin genes: Divergent expression of activin A protein variants and identification of a novel heparan sulfate-binding domain in activin B

Activins are members of the transforming growth factor-β (TGF-β) superfamily of signaling proteins and were originally identified as components of follicular fluid. The proteins are now known to play critical roles in numerous normal and pathological processes and conditions, but less is clear about the relationships between their gene organization and protein variant expression and structure. The four human and mouse activin (Act) genes, termed INHβA, INHβB, INHβC and INHβE, differ in exon numbers. Human INHβA is the most complex with 7 exons and elicits production of three Act A variants (Act A X1, X2 and X3) differing in their pro-region, as we showed previously. Here we further analyzed the mouse INHβA gene and found that its 4 exons encode for a single open reading frame (mouse Act A), corresponding to the shortest human Act A X3 variant. Activins are synthesized and secreted as large complexes made of a long pro-region and a short mature C- terminal ligand and are known to interact with the heparan sulfate (HS) chains of cell surface and matrix proteoglycans. Human Act A X1 and X2 variants do have a HS-binding domain (HBD) with Cardin/Weintraub traits in their pro-region, while the X3 variant does not as shown previously. We found that the mouse Act A lacks a HBD as well. However, we identified a typical HBD in the pro-region of both mouse and human Act B, and synthetic peptides containing that domain interacted with immobilized HS and cell surface with nanomolar affinity. In sum, human and mouse Act A genes elicit expression of different variant sets, while there is concordance in Act B protein expression, reflecting possible evolutionary diversity in function of, and responses to, these signaling proteins in the two species.


Introduction
Our results reveal significant differences in expression of Act A protein variants in human compared to mouse. In addition, we have identified a novel HBD in mouse and human Act B that may play an important role in regulating distribution and function of Act B dimers and Act A/Act B heterodimers within the extracellular milieu.  (Table 1), while its human counterpart spans 25 kb and contains 7 exons [19]. As in humans [19], mouse INHβB, INHβC and INHβE genes all contain 2 exons (Tables 1 and 2). Interestingly, INHβC and INHβE reside next to each other on Chr 10 in mouse ( Table 2) and Chr 12 in humans [19], suggesting that they arose via gene duplication.  1 NCBI Nucleotide Accession number. 2 NCBI Protein Accession number. For corresponding human protein data, see S1 Table. 3 NCBI Gene ID. 4 Exons were determined using the Blat function, UCSC Genome Browser.

Gene organization and RNA expression of Act A
https://doi.org/10.1371/journal.pone.0229254.t001 The human INHβA gene contains 7 exons that are thought to be alternatively spliced and to elicit 3 distinct Act A protein variants named X1, X2 and X3 [19]. Recently, we verified those findings and demonstrated that the variants are produced by human cells and tissues and that the longer X1 and X2 variants contain a typical HBD in their pro-region, while the shorter X3 variant does not [19]. Since the mouse INHβA gene contains 4 exons, it could also potentially give rise to multiple mRNAs and protein variants. To address this question, we translated the entire reference sequence of mouse INHβA mRNA Ref Seq (XM_011244285) encompassing the 4 exons and used it as input sequence. This analysis suggested that only one open reading frame of 1275 nucleotides was generated starting in exon 3 and ending in exon 4 (Fig 1). The resulting predicted protein matched the amino acid sequence of the only mouse Act A protein variant in the databanks (NP_032406) (heretofore termed mouse Act A). Interestingly, mouse Act A is nearly identical in both size and sequence (96%) to the shortest human Act A X3 variant, sharing identical signal sequences, splice junctions and furin cleavage sites with human X3 (Fig 2; S1  To directly test whether mouse Act A protein is indeed the only gene product expressed in vivo, we isolated total RNA from whole mouse embryos and liver. Full-length cDNAs were prepared and used as template in RT-PCR reactions containing forward primers from each of the 4 mouse INHβA exons and an anchored, reverse primer near the end of exon 4 ( Fig 3A; see S2 Table for primer sequence). When forward primers derived from exons 1 or 2 were included, no PCR amplicons of expected size (> 1kB) were generated, even after varying reaction conditions, annealing temperature or including DMSO that has been shown to improve heat denaturation of template DNAs and yield of PCR products [20]. In contrast, when forward primers residing in exon 3 were included, PCR products of the correct size were readily obtained ( Fig 3B). Sanger sequencing of the excised bands confirmed the identity of these transcripts that contained the predicted ATG start site ( To verify the apparent close similarity of mouse Act A and human Act A X3 variant, we used the I-TASSER server at the University of Michigan [21] to define and illustrate the 3D architecture of the two proteins. As shown in Fig 4, both proteins displayed a characteristic and nearly overlapping structure, with a large N-terminal pro-region and a C-terminal mature ligand in which the wrist helical portion was connected to the two fingers and terminal thumb via the palm [17,19]. Neither protein contains a CW motif/HBD in the pro-region or mature ligand.

Analysis of mouse and human activin B
Activins exert their biological functions either as homodimers or heterodimers [1], and it is also well established that many members of the TGFβ protein superfamily interact with HSPGs in the extracellular matrix and cell surface [22]. Thus, the absence of a typical CW motif/HBD in mouse Act A and human X3 variant suggests that intriguingly, these proteins would not utilize HS-dependent restraining mechanisms and could potentially be freer to act. An additional and not exclusionary possibility is raised by the fact that Act A can occur as a heterodimer with Act B. If Act B were to possess a HBD, then heterodimers of Act A and Act B would be able to interact with HS and exploit HS-dependent mechanisms for action. To test these possibilities, we extended our analyses to mouse and human Act B.
In both species, the INHβB gene contains 2 exons (Table 1) and displays a single open reading frame encoding a protein of 411 amino acids in mouse and 407 amino acids in human ( Fig  5). The proteins have an identical signal peptide, a consensus furin splicing site (RKR) [8], a large N-terminal pro-region and a mature ligand (Fig 5 and S1 Table). Attentive survey of their amino acid sequences quickly revealed the presence of the short amino acid segment with the sequence KGSRRKVRVK in the pro-region of each protein (residues 195-204 in mouse and 191-200 in human). This segment is rich in basic residues and closely resembles a typical HS-binding CW motif with a consensus sequence BXXBBBXBXB, where B is a basic residue (Lys or Arg) and X is an uncharged residue [12,15].
To determine whether this motif could serve as part of a HBD within Act B, we synthesized a peptide encompassing the motif and included a biotin tag on the N-terminus via a Gly linker to facilitate analysis ( Fig 6A). The peptide was next tested in solid-phase binding assays where HS was immobilized on standard 96-well plates, following procedures we have reported previously [23]. The peptide did interact with HS with high affinity and saturable binding kinetics (K d~8 nM), and binding was fully prevented by addition of soluble heparin competitor ( Fig  6C). Stereological analysis indicated that the motif exhibited an unstructured configuration ( Fig 6B). RNAs isolated from whole mouse embryos or liver were processed for cDNA synthesis and subjected to PCR analysis using primers from: exon 1 (primer 1), exon 2 (primer 5), exon 3 (primer 7) or exon 4 (primer 9) along with a common anchored reverse primer (primer 16) annealing near the 3' end of the open reading frame of INHβA (all primers are listed in S2 Table). Lane 1, base pair ladder; lanes 2-5, whole embryo RNA PCR products; lanes 6-9, liver RNA PCR products; and lanes 10-13, no template controls (NTC). Note no PCR products were generated with forward primers derived from exons 1 and 2, while amplicons of the correct size were produced when forward primers from exons 3 and 4 used (lanes 4, 5, 8 and 9). C, electrophoretic analysis of RT-PCR products derived from mouse INHβB. cDNA libraries were generated from embryo (E), liver (L), kidney (K), uterus (U) and used as template for PCR reactions using primers (see S2 Table) from: exon 1 (primer 20 or 21), lanes 2-5 and 6-9, respectively and a common anchored reverse primer (primer 22) annealing near the 3' of exon 2 of INHβB mRNA. Note that Act B is highly expressed in embryonic tissue (E) and uterus (U) and at lower levels in liver (L) and kidney (K). The identity of the PCR products was confirmed by Sanger DNA sequencing (S2 Fig). NTC, no template control. https://doi.org/10.1371/journal.pone.0229254.g003 The ability of this peptide to bind immobilized HS is an accepted test of interaction specificity, but it does not reproduce the more complex process of cell or tissue interactions. Thus, we determined whether the peptide was able to interact with the cell surface. As in previous studies, we prepared HBCs consisting of the peptide complexed with a fluorescent NA derivative (NA-488) [19]. The resulting complex was incubated with U937 cells that had been lightly prefixed with 1% buffered formalin. After incubation, the cells were washed, and the amount of bound peptide was assessed by FACs. Clearly, the peptide was able to readily interact with the cell surface, and its binding was blocked by soluble heparin competitor ( Fig 6D). Thus, mouse and human Act B contain a typical HBD that exhibits high affinity binding to HS and the cell surface.
Next, we tested the ability of full-length Act B protein (pro-region and mature ligand) to interact with HS. An expression construct containing the entire open reading frame of Act B with a His tag on the C-terminus was transfected into AD-293 cells (Fig 7) as described [19]. After 3 days, conditioned medium from the transfected cells was harvested and incubated with SP Sepharose (a strong cation exchange resin), Talon resin (interacts with the His tag) or heparin agarose. After incubation, the resins were washed, and bound proteins were eluted, size fractionated on SDS-polyacrylamide gels and analyzed by immunoblot. We found that a protein of~14 kDa (the predicted size of the mature Act B ligand) bound to SP Sepharose and Talon resins and was recognized by anti-Act B or anti-His antibodies (Fig 7C and 7D). We also observed some higher molecular weight bands on these blots (brackets, Fig 7C and 7D) that likely represented unprocessed forms of the pro-protein and were absent in conditioned medium from control non-transfected cells. Interestingly, in addition to binding to Talon, these higher mass proteins were able to interact with heparin agarose (Fig 7D, lane 2). This is consistent with the fact that the HBD resides in the pro-region (see above) of Act B. Indeed, the mature Act B ligand (Fig 7D, arrow on right) did not interact with heparin agarose ( Fig  7D, compare lanes 2 and 3).
To obtain additional insights into how Act B interacts with HS, we used the I-TASSER server [24] to interrogate the overall structure of full-length Act B and assess the location and orientation of the HBD with respect to the entire protein (Fig 8). We found that mouse and human Act B proteins exhibited a very similar three-dimensional structure (Fig 8), in close correlation with their 96% primary sequence homology (Fig 5). Interestingly, the CW motif/ HBD appeared to be located on the surface of the pro-region, presumably well exposed for interactions with HS, but displayed no obvious ordered structure or distinct secondary structure. In an earlier report, we demonstrated by RNA analyses on human cells and tissues that the INHβA gene is alternatively spliced and generates three Act A protein variants (X1, X2 and X3: longest to shortest) [19]. These variants contain a divergent N-terminal pro-region, but the same C-terminal mature ligand. We found that these proteins were not only synthesized, but also processed and secreted properly since we were able to recover mature human Act A ligand (116 residues) following transfection of expression constructs containing full-length cDNAs into AD-293 cells [19]. In contrast, we find here that only one Act A transcript can be amplified using mouse cDNA as template in RT-PCR reactions. This RNA begins at a predicted AUG codon in exon 3 and ends in exon 4 and encodes a 424 amino acid protein that is essentially identical (96%) to the shortest human X3 variant (see . Clearly the difference in INHβA gene structure and exon numbers in mouse and human genomes elicits significant differences in Act A mRNA and protein expression. It is interesting to note that the rat INHβA gene contains 5 exons (transcript variants X4, XM_008771714; X5, XM_008771715 and X6, XM_008771716), thus seemingly representing an intermediate evolutionary step between mouse with 4 exons and humans with 7 exons. It is possible that such changes in gene structure and expression reflect evolutionary modulations in the range and function by Act A and its variants in these species.

Discussion
It is generally assumed that activins are synthesized, dimerize and undergo proteolytic cleavage by furin/proprotein convertases intracellularly and are subsequently secreted as complexes containing the N-terminal pro-region non-covalently associated with the C-terminal mature ligand. Following secretion, this complex binds HSPGs or other ECM components and the mature ligand is released by an unknown mechanism [17]. In the current report, we show that mouse Act A protein, like its human X3 counterpart lack a HBD. In earlier studies, we found that the longer human Act A X1 and X2 variants do possess a stereotypic HBD near the N-terminus of their pro-region with high affinity binding capacity for HS in solid-phase and cell surface binding assays [19]. One would predict that the absence or presence of a HBD would have functional consequences since the expected interactions of human Act A X1 and X2 isoforms with HSPGs would influence protein distribution and activity. Most previous studies on human Act A have made use of full-length X3 variant expression constructs [25] and thus, some of the data and insights may have to be re-evaluated using X1 or X2 constructs. Because the only mouse Act A also lacks a HBD as we show here, how would it exist within the extracellular environment and would it have other means of interacting with the ECM? As suggested above, one possible answer is offered by the fact that Act A exists either as a homodimer or a heterodimer with Act B. Thus, it is possible that mouse Act A dimers (and human X3 dimers) do not interact with HSPGs and are free to exert their biological activity. In contrast, an ActA-ActB heterodimer could exploit the HBD in Act B reported here, to interact with HSPGs and exhibit more restrained and perhaps more selective biological action and range in vivo.
In their seminal studies, Cardin and Weintraub first described the domains present in four proteins-apolipoprotein B, apolipoprotein E, vitronectin and platelet factor-4-that mediate interactions with HS. In these studies, two binding motifs were identified-XBBXBX and XBBBXXBX-where B represents a basic residue (Arg or Lys) and X represents a non-charged amino acid [15]. Additional motifs have since been described in many proteins [12] and are frequently referred to as Cardin Weintraub (CW) motifs. The CW motif we identified in Act B-KGSRRKVRVK-has a consensus BXXBBBXBXB, and binds HS with high affinity in solidphase binding assays and binds to the cell surface (Fig 6). In silico modeling reveals that the motif has a disordered structure resembling the HBD present on the C-terminal region of mature BMP 5-7, while differing from the HBD present in the N-terminal region of mature BMP 2-4, which has a helical configuration [23]. Our modeling of full-length Act B protein shows that its HBD resides on the surface, in the middle of the pro-region, where it would be readily available to interact with HSPGs and potentially other components in the ECM. Similarly, the HBD present in human Act A X1 and X2 variants also resides in the pro-region, but differs from the domain in Act B (Fig 8) as it has a helical structure and resides closer to the Nterminus of the pro-region [19]. Clearly, different members of the TGFβ superfamily have evolved HBDs which differ in secondary structure and location either within the pro-region or mature ligand, while some lack a HBD such as mouse Act A and the human Act A X3 variant. Future studies will need to be directed at uncovering the seemingly important implications of this broad diversity. It should also be remembered that while HBDs are important mediators of protein-HS interactions and protein distribution and function, there is evidence that other individual residues or short segments have roles in these interactions, as well [26].
Activins are multi-functional cytokines which play important roles in a wide range of normal and abnormal physiological processes. One example is Fibrodysplasia Ossificans Progressiva (FOP), a severe congenital disorder characterized by formation and accumulation of extra-skeletal bone (heterotopic ossification) [27]. The majority of FOP patients carry mutations in the type I BMP receptor ALK2 encoded by ACVR1 and display a recurrent ACVR1 R206H mutation [27]. Under normal circumstances, ALK2 interacts with various BMPs and signals via canonical SMAD1/5/8 pathway, but is not responsive to activins [6,28]. Recent important studies have revealed that mutant ALK2 R206H becomes responsive to exogenous Act A and elicits pSMAD1/5/8 signaling [29]. Treatment of FOP-like mutant mice with an Act A neutralizing monoclonal antibody was found to significantly inhibit heterotopic ossification [29]. Our data here show that mouse Act A does not possess a HBD and thus, may have a broader activity range. In contrast, an Act A-Act B heterodimer could exploit the HBD present in the pro-region of Act B and interact with HSPGs. Lastly, it remains to be studied whether FOP pathogenesis is influenced by expression and/or bioavailability of Act A homodimers versus heterodimers and whether the significant differences in Act A variants in humans and mice could affect disease progression, severity and responses to treatment.

Reagents
NeutrAvidin (NA), NA-HRP, NA-DyLight 488 (NA-488) and Fisher exACTGene DNA Ladder (100-10,000 bp) were obtained from Thermo-Fisher. Human Act B Antibody (MAB659) was obtained from RD Systems. Heparin was obtained from Sagent Pharmaceuticals and HS was obtained from Millipore-Sigma (www.sigmaaldrich.com). All DNA Oligonucleotide primers were obtained from Integrated DNA Technologies (www.idtdna.com) and are listed in Table 3 and S2 Table. Peptides were synthesized and purified by Peptide 2 (www.peptide2. com).

RNA, cDNA and DNA sequencing
RNA was isolated from adult mouse tissues and embryos using TRIzol Reagent (Thermo-Fisher) following the manufacturers protocol. Five micrograms of glycogen was added to the aqueous phase prior to addition of isopropanol to facilitate RNA precipitation. RNA concentration was determined using a Nanodrop spectrophotometer. cDNA was prepared with a Versco cDNA synthesis kit, using 3 μg of RNA as template plus 200 ng of random hexamer and 200 ng oligo dT as primers, following the manufacturer's instructions. All PCR reactions were optimized for each primer set by gradient PCR, using an Applied Biosystems Verity thermal cycler. PCR reactions contained 1 μl of cDNA, primer sets (S2 Table) and GoTaq Green master mix (Promega) following the manufactures protocol. PCR products were size fractionated on 1% agarose gels and subsequently the bands were isolated, DNA was purified from the gel using Micro Bio-Spin columns (Bio-Rad) and subjected to sequencing. All Sanger DNA sequencing was performed by the Center for Applied Genomics (CAG; caglab.org) at the Children's Hospital of Philadelphia.

Solid-phase binding assays
Nunc MaxiSorp 96 well flat bottom plates were coated with HS in 50 mM carbonate buffer (pH 9.4) overnight at 4˚C as described [23]. All binding assays were carried out in PBS, 0.1% tween 20 (PBST) containing 1% bovine serum albumin. The plates were incubated with peptide tetramers consisting of biotinylated peptide complexed with NA-HRP for a minimum of 2

Activin B protein expression constructs
The entire open reading frame for human Act B was cloned into EcoR1/Xba1 sites of mammalian expression vector pcDNA3.1 as described [19]. This expression construct contains a cytomegalovirus (CMV) promoter for constitutive expression in mammalian cells. A His 8 tag was added to the C-terminus (3'end of cDNA) to facilitate purification via immobilized metal affinity chromatography (IMAC). After cloning, the resulting expression constructs were verified by Sanger DNA sequencing. For protein expression, 6 μg of plasmid DNA plus 18 μl of FuGene6 Transfection reagent (Promega) was added to 600 μl of Opti-MEM medium and incubated for 20 minutes at room temperature. Next, the transfection mixture was added to AD-293 cells growing in 10 ml RPMI medium and 5% fetal calf serum. Conditioned medium was collected after 72 hours and incubated with TALON metal affinity resin (Takara Bio USA) for purification. The resin was washed with PBS and transferred to Micro Bio-Spin columns (Bio-Rad). His-Tag bound proteins were eluted with 0.5 M imidazole and analyzed on NuPAGE 4-12% Bis-Tris Gels (Invitrogen) under reducing conditions followed by transfer to PVDF membranes. Blots were probed with primary anti-His and anti-Act B antibodies and developed with IRDye Secondary Antibodies (Li-Cor).

mRNA expression of Act B constructs
To evaluate mRNA expression of Act A expression constructs, transfected cells were washed and pelleted. RNA was isolated by combining Trizol and Direct-zol RNA MiniPrep Kit (Zymo Research #2050) following the manufacturers protocols. During isolation, RNA is treated with DNAase, to remove any contaminating plasmid or genomic DNA. RNA concentration and purity was assessed using a NanoDrop 2000 spectrophotometer. Act B mRNA expression was determined using forward primer 1 and reverse primer 3 (Table 3) by RT-PCR, which yields an amplicon of 267 bp.

Sequence alignments and in silico modeling
Alignments of primary sequences were constructed using Unipro UGENE [30]. Protein structure predictions were generated using the I-TASSER server for Protein Structure and Function at the University of Michigan (zhanglab.ccmb.med.umich.edu) [21]. The resulting structures were visualized using Chimera modeling software (www.cgl.ucsf.edu/chimera).

Statistical analysis
Statistical evaluation of data utilized Student's t-test employing GraphPad Prism software (www.graphpad.com).