Transcriptome analysis reveals TMPRSS6 isoforms with distinct functionalities

Abstract TMPRSS6 (matriptase‐2) is a type II transmembrane serine protease involved in iron homoeostasis. At the cell surface of hepatocytes, TMPRSS6 cleaves haemojuvelin (HJV) and regulates the BMP/SMAD signalling pathway leading to production of hepcidin, a key regulator of iron absorption. Although four TMPRSS6 human isoforms and three mice Tmprss6 isoforms are annotated in databases (Ensembl and RefSeq), their relative expression or activity has not been studied. Analyses of RNA‐seq data and RT‐PCR from human tissues reveal that TMPRSS6 isoform 1 (TMPRSS6‐1) and 3 are mostly expressed in human testis while TMPRSS6‐2 and TMPRSS6‐4 are the main transcripts expressed in human liver, testis and pituitary. Furthermore, we confirm the existence and analyse the relative expression of three annotated mice Tmprss6 isoforms. Using heterologous expression in HEK293 and Hep3B cells, we show that all human TMPRSS6 isoforms reach the cell surface but only TMPRSS6‐1 undergoes internalization. Moreover, truncated TMPRSS6‐3 or catalytically altered TMPRSS6‐4 interact with HJV and prevent its cleavage by TMPRSS6‐2, suggesting their potential role as dominant negative isoforms. Taken together, our results highlight the importance of understanding the precise function of each TMPRSS6 isoforms both in human and in mouse.

TMPRSS6 mutations have been shown to prevent HJV cleavage either directly by altering TMPRSS6 enzymatic activity or by preventing the protease from reaching the cell surface. 15,16 Conceivably, because TMPRSS6 inhibition could potentially lower circulatory iron levels by elevating hepcidin levels, TMPRSS6 has become an appealing therapeutic target for diseases characterized by iron overload. 17,18 As proof of this, genetic knockdown of TMPRSS6 in mouse models of b-thalassaemia and haemochromatosis reduces iron overload-related characteristics and symptoms. 19,20 Velasco et al. 5 first described TMPRSS6 as an 802-amino acid (aa) protein mostly expressed in the liver, but isoforms of various lengths have been annotated in different databases with some discrepancies. In fact, it is the 802aa-form (TMPRSS6 isoform 2, according to the UniProt Consortium nomenclature 21 ) that has been most commonly used. 5,16,[22][23][24][25] However, other groups, including ours, have focussed on TMPRSS6-1 (811 aa), which is known as the "TMPRSS6 canonical isoform". 13,15,[26][27][28][29] The difference between the two isoforms is the expression of TMPRSS6-1 coding exon 1, which encodes for residues 1-9 in the N-terminal, cytoplasmic portion of the protein. 30 Although two other well-supported TMPRSS6 isoforms are annotated in UniProt 21 and Ensembl 30 databases, neither their relative expression nor their respective functionalities have been studied. In mice, Tmprss6 isoforms annotation is not constant in the different databases but three distinct coding transcripts have been annotated in NCBI Reference Sequence Database (RefSeq). 31 Using transcriptome analysis and heterologous expression, we confirm the existence and relative abundance of the different isoforms in both species. More importantly, we found revealing differences in the functionality of human TMPRSS6 isoforms. Because TMPRSS6 is such a critical player in iron regulation and a promising therapeutic target, we wanted to highlight the importance of knowing precisely which isoforms are expressed in human tissues and to characterize the distinct functional properties of these isoforms.

| Mouse primary hepatocytes
Mouse primary hepatocytes were obtained from C57BL/6 mice. Animals were anesthetized with ketamine/xylazine. The liver was perfused with 70 mL of Liver Perfusion Medium prior to the perfusion with Liver Digest Medium. Liver was then cut into small pieces and dissociated in Hepatocyte Wash Medium. Viable cells were plated in 6-well plates and were washed with PBS 24 hours post-plating. Cells were treated with TRIzol reagent to further isolate and analyse RNA.
The use of animals in the context of this project was approved by the Universit e de Sherbrooke Animal Ethic Committee. Additional details are provided in the supplemental data.

| RT-PCR
Human liver cDNA was from Applied Biosystems (Foster City, CA).
Human liver RNA pool from 5 healthy donors was from BioChain DION ET AL.
| 2499 (Newark, CA). XpressRef Universal Total RNA (total human RNA) was from Qiagen (Germantown, MD). Mouse primary hepatocytes RNA were extracted from a 6-well plate covered by adherent cells using TRIzol with chloroform, following the manufacturer's protocol.
The aqueous layer was recovered, mixed with one volume of 70% ethanol and applied directly to an RNeasy Mini Kit column (Qiagen, Germantown, MD). DNAse treatment on the column and total RNA recovery were performed as per the manufacturer's protocol. RNA quality and presence of contaminating genomic DNA were verified as previously described. 34 Figure S1 and Table S4 and for the mice primer list see Figure S2 and Table S5.

| Plasmid construction
The cDNAs encoding TMPRSS6-1 and HJV were obtained and cloned as previously described. 26 TMPRSS6-2 construct was obtained using the QuikChange site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA). TMPRSS6-3 and TMPRSS6-4 constructs were obtained by inserting a synthetic double-stranded DNA block coding for the isoform sequence into a modified form of pcDNA6/V5-His (Invitrogen) that was previously described. 26 Additional details are provided in the supplemental data.

| Immunofluorescence
Cells were seeded on poly-L-lysine-coated coverslips. Cells were transfected with appropriate plasmids. Twenty-four hours later, cell surface TMPRSS6 was labelled for 1 hour at 4°C. Cells were washed and incubated at 37°C for 15 or 30 minutes. Cells were then prepared as previously described. 35 Cells were examined on an inverted spectral scanning confocal microscope FV1000 (Olympus, Tokyo, Japan). Additional details are provided in the supplemental data.

| Expression and detection of TMPRSS6
Cells were transfected with TMPRSS6 isoform cDNAs performed with Lipofectamine 3000 in 6-well plates. Twenty-four hours later, the cell media were replaced with HCELL-100 media for 24 hours. Cell media were collected and concentrated, and cells were lysed. Samples were loaded on 12% SDS-PAGE and analysed by immunoblotting. Additional details are provided in the supplemental data.

| Proteolytic activity measurements in the cell media
At 24 hours post-transfection, the cell media were replaced with HCELL-100 media for another 24 hours. The media were collected, and activity was measured by the release of fluorescence from Boc-QAR-AMC cleavage. 26 Additional details are provided in the supplemental data.

| Membrane isolation and proteolytic activity measurements of isolated membrane fractions
Hep3B cells were transfected with TMPRSS6 isoform cDNAs performed with Lipofectamine 3000 in 10 cm culture dishes. Membranes were then isolated using ProteoExtract Native Membrane Extraction Kit as previously described. 36 The activity in the membrane fractions was measured using Boc-QAR-AMC fluorogenic substrate in assay buffer as previously described. 37 Samples were loaded on 12% SDS-PAGE and analysed by immunoblotting.

| Statistical analysis
Statistical analyses were conducted using GraphPad Prism version 7.0c (GraphPad Software, La Jolla, CA). Outliers were removed using the ROUT method (Q = 1%). One-sample t test analysis (hypothetical mean fixed at 1) was used to compare the activity of isoforms relative to mock transfected cells (fold induction). P values <.05 were considered statistically significant ( †). Normality was assessed using the D'Agostino-Pearson omnibus normality test before using nonparametric Kruskal-Wallis test to compare the activity between the isoforms. P values <.05 were considered statistically significant (*).

| Expression of TMPRSS6 transcripts in human
The human TMPRSS6 gene is located on chromosome 22 (22q12.3) and is expressed as 7 known different transcripts (Ensembl database), but only 4 of them have a well-supported annotation and are predicted to be expressed as proteins ( Figure 1A). 30 These transcripts encode 4 TMPRSS6 isoforms that lead to the production of different proteins ( Figure 1B). To facilitate reading, a 1-4 numbering nomenclature (according to the UniProt Consortium) 21 is used in this study and has been linked to the Ensembl transcript annotation. Residues 1-9 from the N-terminal, cytoplasmic tail, are present in isoform 1 but absent in isoform 2. TMPRSS6-3 has two different annotations, including one coding for a protein of 461 aa (ENST00000442782) that expresses the first 9 coding exons of TMPRSS6-1 but has an alternatively spliced form of exon 10 that drives the expression of a truncated isoform in the second CUB domain (Ensembl annotation). 30 The second annotation for TMPRSS6-3 describes a protein of 452 aa (UCSC annotation uc003aqu.3) that has the same non-coding exon 1 than TMPRSS6-2.
To the best of our knowledge, functional differences between TMPRSS6 isoforms have not been characterized. In fact, research reports have routinely been using TMPRSS6-1 or TMPRSS6-2 in   transfection experiments, while no clear function for TMPRSS6-3 or TMPRSS6-4 has been elucidated. To gain insight on the expression of TMPRSS6 isoforms and their abundance in human tissues, we analysed publicly available RNA-seq datasets of the Genotype-Tissue Expression (GTEx) project (Figure 2A). 32 We found that TMPRSS6 is mainly expressed in the liver with lower levels  Figure S3). 33 HepG2 cells expressed the highest levels of TMPRSS6-3, and we observed that elongated exon 10 is present when analysing read alignments compared to liver samples, but TMPRSS6-1 coding exon 1 is absent ( Figure S3). This suggests that TMPRSS6-3 is expressed as a 452 aa protein in liver-derived cells, in accordance with the UCSC annotation. 38  We analysed the expression of these isoforms within healthy mice liver samples performed with RNA-seq data analysis. (Figure 3C). We found that Tmprss6-1 is the main expressed isoform (79%), with Tmprss6-2 in lower abundance (21%). Tmprss6-X2 was not detected. To confirm the presence of Tmprss6 isoforms, we performed RT-PCR on mice primary hepatocytes. All transcripts were detected using this technique ( Figure 3D). These results confirm the existence and reveal the differential expression of annotated Tmprss6 isoforms in mouse liver. Notwithstanding the fact that the mouse is an interesting and essential model to study TMPRSS6 function, mouse isoforms do differ from those found in human. However, we nonetheless focused on characterizing the human isoforms of this protease.

| Human TMPRSS6 isoforms have different functionalities
We previously showed that TMPRSS6-1 contained a cell surface internalization motif in its cytoplasmic tail (aa 2-11), 26  We next examined the catalytic-related functionalities of TMPRSS6-V5-tagged isoforms in transiently transfected Hep3B cells.
Because this cell line is derived from liver (where TMPRSS6 is mostly expressed), it may be a better model to mimic the physiological conditions of the liver. We addressed the ability of TMPRSS6 to be shed from the cell surface, which is an event dependent on TMPRSS6 catalytic activity. 24,26 Immunoblotting shows that all isoforms are expressed in cell lysates as zymogen forms migrating at >100 kDa for TMPRSS6-1, TMPRSS6-2 and TMPRSS6-4 or at >50 kDa for the truncated TMPRSS6-3 ( Figure 5A, upper panel).

| Human TMPRSS6 isoforms interact with HJV
Even though TMPRSS6-3 does not have LDLRA or catalytic domains and TMPRSS6-4 has altered catalytic activity, we examined whether these isoforms interact with HJV. 13,29 Proteins from Hep3B cells cotransfected with TMPRSS6 isoforms and HJV were immunoprecipitated using an anti-V5 antibody, and the ability to interact with HJV was verified by immunoblotting with HJV antibodies. Interestingly, all four TMPRSS6 isoforms interacted with HJV as seen by the detected 50 kDa form ( Figure 6A). These results indicate that neither TMPRSS6 catalytic domain, as previously suggested, 13 nor its LDLRAs domains are required for interaction with HJV.
We next assessed the ability of these isoforms to cleave HJV.
Immunoblotting of media proteins reveals that when transfected alone, HJV is shed in the media and immunoreactive proteins, ranging from 50 to 30 kDa, can be detected ( Figure 6B, lower panel).
Finally, as TMPRSS6-3 and TMPRSS6-4 interact with but do not cleave HJV, we assessed their ability to act as dominant negative

| DISCUSSION
TMPRSS6 is a type II serine protease mainly expressed at the cell surface of hepatocytes and plays an important role in iron regulation, most notably, through HJV cleavage, thus regulating the BMP/ SMAD signalling pathway, leading to hepcidin production. 13,14 Our findings reveal the existence of 4 distinct TMPRSS6 isoforms and highlight different relative abundance in human tissues. By analysing publicly accessible RNA-seq data, we demonstrate that TMPRSS6-2 has the highest expression in human liver and should be considered the main liver isoform. We also show that TMPRSS6-1, which is considered the canonical variant according to UniProt Consortium, 21 is expressed in testis at low levels and not detected in the liver using RNA-seq and RT-PCR. Interestingly, using a combination of techniques, we show that TMPRSS6-3 (lacking LDLRA and catalytic domains) is expressed at very low levels in liver but is expressed in testis. Importantly, we report that TMPRSS6-4 is expressed in human liver which suggests that this isoform could have a significant role in hepatic functions.
We also reveal the existence of three Tmprss6 mouse isoforms and studied their expression performed with RNA-seq data analysis of mouse livers. We also confirmed the expression of these isoforms using PCR amplification on mice primary hepatocytes. The differences between mice Tmprss6 isoforms are subtle but will ultimately need to be investigated. Moreover, it is important to note that at the present time, there is less data available for mice than for humans and therefore mice annotations may not be as reliable as human annotations in databases. Thus, the possibility that other Tmprss6 isoforms exist cannot be ruled out.

Using heterologous expression in HEK293 and Hep3B cells, we
show that all human TMPRSS6 isoforms reach the cell surface.
While TMPRSS6-1 undergoes constitutive internalization, as F I G U R E 6 TMPRSS6 isoforms interaction with haemojuvelin. (A) Hep3B cells were cotransfected with TMPRSS6V5-tagged and haemojuvelin (HJV). Immunoprecipitation was performed in cell lysate using an anti-V5 antibody. Samples were loaded on 10% SDSpolyacrylamide gels and immunoblotting was performed using anti-HJV or anti-V5 antibodies (n = 3). (B) Hep3B cells were cotransfected with TMPRSS6V5-tagged isoforms and HJV. HJV cleavage in cell media was detected by immunoblotting with anti-HJV antibody. Equal amounts of cell lysate (CL) and concentrated cell medium (CM) were loaded on 12% SDS-polyacrylamide gels. Cell lysate GAPDH was blotted as a loading control (n = 3). (C) Hep3B cells were cotransfected either with HJV alone or in combination with one or two TMPRSS6V5-tagged isoform. Equal amounts of cell lysate (CL) and concentrated cell medium (CM) were loaded on 12% SDS-polyacrylamide gels. Cell lysate GAPDH was blotted as a loading control (n = 3) previously described by our group, 26 TMPRSS6-2, TMPRSS6-3 and TMPRSS6-4 remain at the plasma membrane. We also describe the inability of TMPRSS6-4 to undergo auto-activation, which leads to the production of a catalytically altered protein unable to cleave HJV in the cell media compared to TMPRSS6-1 and TMPRSS6-2. Importantly and regardless of the presence of catalytic domains or activity, we show that TMPRSS6-3 and TMPRSS6-4 interact with HJV. This finding correlates with the results of Silvestri et al. 13 and suggests that TMPRSS6 ectodomains are involved in interactions with HJV.
Moreover, we show that TMPRSS6-3 and TMPRSS6-4 act as dominant negative regulators of HJV cleavage by TMPRSS6. We believe that, similar to the TMPRSS6 mask mutant, 13 TMPRSS6-3 and TMPRSS6-4 could represent a novel mechanism of iron regulation.
Indeed, similar to TMPRSS6-3, the catalytically truncated TMPRSS6 mask mutant interacts with HJV 13 but does not repress hepcidin promoter activation. 40 Moreover, as TMPRSS6 proteolytic activity has been previously described as protective towards prostate cancer in vitro, 41  Of note, we had previously demonstrated 26 (and validated herein) that only TMPRSS6-1 undergoes internalization when expressed in two heterologous model, one of which being primary hepatocytes. Therefore, we believed isoform 1 was expressed in the liver. Because our present results reveal that isoform 1 is not detected in liver, we believe our previous data, 26 which used high concentrations of IgG antibodies (800 nmol/L) to label cell surface TMPRSS6, might therefore be a consequence of internalization by the IgG receptor FcRn (FCGRT), known to be expressed in both hepatocytes 43 and HepG2 cells. 33 On the other hand, it is clear that residues 1-9 encoded by exon 1 are responsible for the internalization of that isoform as validated when the isoform is expressed using heterologous expression in HEK293 and Hep3B cell lines.
Taken together, our results highlight the importance of identifying which TMPRSS6 isoforms are expressed in human tissues as well as the properties these isoforms possess. Considering the important role of TMPRSS6 in iron regulation, the protease isoforms herein described and studied should be taken into account in future studies, especially for TMPRSS6-3 and TMPRSS6-4, for which physiological functions are still to be elucidated.