Proprotein Convertase Subtilisin/Kexin Type 7 (PCSK7) Is Essential for the Zebrafish Development and Bioavailability of Transforming Growth Factor β1a (TGFβ1a)*

Background: The in vivo importance of PCSK7 in the vertebrates is currently poorly understood. Results: Inhibiting PCSK7 in zebrafish results in various developmental defects and dysregulation of gene expressions. Conclusion: PCSK7 is essential for zebrafish development and regulates the expression and proteolytic cleavage of TGFβ1a. Significance: PCSK inhibitors are considered future therapeutics for human diseases; understanding the biological role of PCSK7 is therefore critical. Proprotein convertase subtilisin/kexin (PCSK) enzymes convert proproteins into bioactive end products. Although other PCSK enzymes are known to be essential for biological processes ranging from cholesterol metabolism to host defense, the in vivo importance of the evolutionarily ancient PCSK7 has remained enigmatic. Here, we quantified the expressions of all pcsk genes during the 1st week of fish development and in several tissues. pcsk7 expression was ubiquitous and evident already during the early development. To compare mammalian and zebrafish PCSK7, we prepared homology models, which demonstrated remarkable structural conservation. When the PCSK7 function in developing larvae was inhibited, we found that PCSK7-deficient fish have defects in various organs, including the brain, eye, and otic vesicle, and these result in mortality within 7 days postfertilization. A genome-wide analysis of PCSK7-dependent gene expression showed that, in addition to developmental processes, several immune system-related pathways are also regulated by PCSK7. Specifically, the PCSK7 contributed to the mRNA expression and proteolytic cleavage of the cytokine TGFβ1a. Consequently, tgfβ1a morphant fish displayed phenotypical similarities with pcsk7 morphants, underscoring the importance of this cytokine in the zebrafish development. Targeting PCSK activity has emerged as a strategy for treating human diseases. Our results suggest that inhibiting PCSK7 might interfere with normal vertebrate development.

exerts its function mainly in the trans-Golgi network and on the cell surface (13,14). Importantly, previous biochemical studies indicate that PCSK7 operates often redundantly with other PCSKs, especially FURIN. For example, the activity of PCSK7 can be replaced in the site-specific cleavage of pro-BMP4 (15), pro-PDGF (16), pro-NGF (17,18), and pro-VEGF-C (19). However, PCSK7 seems to be solely responsible for rescuing an unstable MHC I in post-endoplasmic reticulum compartments (20) and for the proteolysis of pro-EGF (21). In addition, a genome-wide association study recently revealed that an SNP (rs236918) in the PCSK7 gene region is associated with an increase in the serum level of soluble transferrin receptor, thus implying that PCSK7 could have a role in iron metabolism in humans (22). This observation was later confirmed by showing that PCSK7 is the only convertase that sheds transferrin receptor into the medium in several cell lines (23).
Most mammalian proteins and genes have orthologs in the zebrafish, which makes them a feasible model for studying the role of PCSK enzymes in vertebrate biology in vivo. Previously, the functions of FURIN (24), PCSK5 (25), MBTPS1 (26), and PCSK9 (27) have been assessed in developing zebrafish. Inactivation of FURIN and MBTPS1 in zebrafish results in defective skeletal and cartilage formation, respectively, whereas PCSK5 and PCSK9 are important for neurological development. To decode the poorly defined PCSK7 function in vertebrate biology in vivo, we inhibited PCSK7 in zebrafish using morpholinos (MOs). The phenotypes of the pcsk7 morphants clearly demonstrate that PCSK7 has a critical and non-redundant role in early zebrafish development.
Zebrafish Lines-Embryos used in all experiments were obtained through natural crosses of wild-type AB strain individuals. Adult fish used in QRT-PCR expression analyses as well as in MO experiments were also of the wild-type AB strain. All fish used in the experiments were maintained under standard conditions at 28 QRT-PCR Expression Analyses-Expression of pcsk genes was measured both in various adult zebrafish tissues (female/ male gonads, liver, kidney, intestine, eye, gill, brain, skin, and tail) as well as in whole developing embryos of different ages (1-7 days postfertilization (dpf)). Conventional quantitative real time PCR with a reverse transcribed cDNA template from total RNA was used. ef1a gene (ENSDARG00000020850) was utilized as a housekeeping gene (28). Primers used in the QRT-PCR analyses are presented in Table 1.
Morpholinos-Prior to morpholino design 8 -10 individual fish of the AB strain were sequenced for each gene to verify the sequence identity with the published sequence and to find genomic regions homomorphic enough for MO design. The furinA and furinB MOs were designed to hit ATG sites, whereas the pcsk7 MOs target exon-intron boundaries in exon 3 (pcsk7 e3 MO) and exon 8 (pcsk7 e8 MO). These exons contain codons for amino acids of the catalytic triad of the PCSK7 enzyme. The morpholino for p53 was commercially predesigned, and the random control (RC) morpholino is a random base mixture at every position intended for use as a negative control.
Morpholino and RNA Injections-For injections, 0.25-1.0 pmol of each MO was used. A rhodamine dextran tracer was used to control the injections, and unsuccessfully injected embryos were removed at 1 dpf. All injections (1-2 nl) were administered into the yolk sac of a one-to four-cell stage zebrafish embryo. 0.2 M KCl was used as buffer in the injection solutions. The RNA used in the RNA rescue experiments was transcribed from pcsk7 cDNA with the SP6 mMESSAGE mMACHINE kit (Ambion, Austin, TX) according to the manufacturer's instructions (imaGenes, Berlin, Germany). All MO injection experiments were controlled with embryos injected with RC MO (GeneTools) as well as with non-injected embryos. Injections were carried out using a PV830 Pneumatic PicoPump (World Precision Instruments) and a Nikon SM7645 microscope. For visual analysis and live fish pictures, a SteREO Lumar V12 microscope with the AxioVision Rel. 4.8 program (Carl Zeiss) was used.

Gene
Sequence 5-3 pcsk7 Histological Staining-The larvae were fixed in 4% paraformaldehyde, PBS solution and embedded in 2% agarose. Next, the samples were dehydrated in an alcohol series (70%, 96%, absolute ethanol, and xylene; 1 h in each). The dehydrated samples were embedded in paraffin, and 5-m transverse sections were cut. These were then fixed on glass slides at 58°C, deparaffinized in xylene (2 ϫ 4 min) and in absolute, 96, and 70% ethanol (each 1-2 ϫ 3 min), and rinsed with distilled H 2 O. The hematoxylin-eosin staining was performed as follows: Mayer's hematoxylin, 2 min; running tap water, 2 min; water, 1.5 min; 70% ethanol, 15 s; eosin Y, 15 s; 96% ethanol, 30 s twice; absolute ethanol, 1 min twice; and xylene, 1-4 min. Thereafter, the slides were mounted, and pictures were taken using an Olympus SZX16 microscope and a Color View Soft Imaging System camera.
pcsk7 in Situ Hybridization-Whole mount in situ hybridization was performed for pcsk7 as described in detail in Thisse and Thisse (29) to analyze the detailed expression pattern of the gene. 2-dpf embryos were used. An antisense probe was used for specific detection of the pcsk7 mRNA, and a sense probe was used to control for unspecific background staining.
RNA Microarray-100 ng of total RNA (triplicate samples pooled from 18 -35 embryos/group; two time points of 6 and 24 hpf) isolated from fish injected with RC (0.5 pmol) or pcsk7 e3 (0.5 pmol) ϩ p53 (0.75 pmol) MO was amplified and Cy3-labeled. Altogether, 1.65 g of sample was then hybridized on the arrays (Agilent 4x44K Zebrafish GE v3; 65°C, overnight). Chips were scanned using an Agilent Technologies G2565CA Scanner, and numeric data were obtained from Agilent Feature Extraction software v10.7.1 (The Finnish Microarray and Sequencing Centre, Turku, Finland). Only features having corresponding Entrez ID available were retained. The expression value for each gene is calculated in three phases. 1) If the same probe appears in the raw data file more than once, the row from the data file that has the highest average expression across all samples is used to represent that probe. 2) If the same gene appears in the raw data file more than once, the probe that has the highest average expression across all samples is used to represent that gene. 3) The resulting gene expression values are quantile-normalized across samples. Genes with a median expression Ͼ10 on a linear scale in either the pcsk7 morphant or control samples were included in further analyses. A twosample t test was used to test the differential expression of each gene between the morphant and control samples using the log 2 gene expression values. To control for the false discovery rate, the resulting p values from the t test were used to calculate q values for each gene as described (30). Genes with a q value Ͻ0.05 were retained. A log 2 ratio between the median of pcsk7 morphant samples and controls was calculated for each gene, and genes with an absolute log 2 ratio Ն2 were considered differentially expressed. A gene ontology enrichment analysis was conducted using hypergeometric distribution testing. Here, we tested whether the number of differentially expressed genes annotated to a gene ontology (GO) term was larger than could be expected by chance. The resulting p values serve as an indication of the possible enrichment of each GO term.
To summarize the GO categories, each enriched term was annotated to a high level GO term. A high level GO term refers to terms that are directly linked to the root term biological process, molecular function, or cellular component. Finally, the number of terms under each high level GO term was calculated. Here, we included the 15 most enriched terms into the analysis. The original raw data of the microarray are available in the Gene Expression Omnibus database.

Quantification of pcsk Gene Expression in Zebrafish-
The zebrafish genome contains orthologs for most mammalian genes. We blasted the human PCSK protein sequences against zebrafish genome databases and found unambiguous orthologs for all but two members (PCSK4 and PCSK6) of the mammalian PCSK family. Two isoforms for both human FURIN (furinA and furinB) and PCSK5 (pcsk5a and pcsk5b),which probably arose from a genome duplication that occurred early in the evolution of teleost fish, were present in the zebrafish genome (31).
We first analyzed the expression of pcsk genes in the whole developing zebrafish using a QRT-PCR approach (1-7 dpf). In these experiments, different pcsk genes showed variable expression patterns. For example, the pcsk family members with reported implications in neuronal development (pcsk1, pcsk2, pcsk5a, and pcsk9) were significantly up-regulated during the early development (Fig. 1A). In contrast, the expression of other pcsk genes remained relatively constant during the first 7 days of life. However, a subtle but significant up-regulation of the furinB, pcsk5b, and pcsk7 mRNAs could be observed at 2.5 dpf.
In accordance with previous studies in mammals, our experiments assessing tissue-specific pcsk expression in adult fish showed particularly high expression of pcsk1 and pcsk2 in eye and brain (32,33) (Fig. 1B). Apart from the lack of pcsk2 in gill tissue, variable degrees of expression of all proprotein convertase enzymes could be detected with QRT-PCR in the tested tissues. Interestingly, two orthologs of FURIN and PCSK5 showed (to some extent) complementary tissue expression profiles. The pcsk7 expression was highest in female gonads, brain, and eye in adult zebrafish. The magnitude of pcsk5a/b and pcsk7 expression levels was generally lower than that of the furin genes. These data corroborate that pcsk7 is present in FIGURE 1. Expression of pcsk genes in zebrafish. QRT-PCR analysis was used to analyze the relative expression of the pcsk genes in developing zebrafish larvae (1-7 dpf) (A) and various adult zebrafish tissues (B). A, pcsk gene expression levels were normalized to the housekeeping gene ef1a, and the normalized gene expression on 1 dpf was given a value of 1. Other time points are shown as relative to this. Asterisks denote statistical significance for differences in comparisons between 1 dpf and other time points: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001 in Welch-corrected two-tailed Student's t tests. Experiments in A were performed with three biological replicates (each sample consisted of 15-30 individual larvae depending on the age of larvae). B, in adult tissue analyses, gene expression levels related to that of ef1a are shown. Experiments in B were performed twice in technical replicates with essentially similar results. Error bars represent S.D. Note the diverse scales on the y axis. DECEMBER 20, 2013 • VOLUME 288 • NUMBER 51 both developing larvae and multiple adult zebrafish tissues together with other convertase enzymes. This allows the reliable assessment of the specific role of the pcsk7 in vertebrate biology.

PCSK7 in Zebrafish
Homology Modeling of PCSK7-To assess putative structural and electrostatic differences between mammalian and zebrafish PCSK7, we prepared a homology model of the protein with Modeler 9v10 (34) using published crystallographic structures of mouse FURIN and yeast Kexin as templates. Our modeling efforts showed evident structural conservation in catalytic and P domains of PCSK7 between the human and zebrafish (Fig. 2). We then analyzed the binding of a substrate to PCSK7 by superimposing the crystallographic structure of subtilisin (Protein Data Bank code 1CSE (35)) with the PCSK7 models. These data demonstrated that most of the potential substratebinding residues are also conserved between human and zebrafish. The alignment of the PCSK7 catalytic and P domains of several species further demonstrated that the active sites are close to identical except for a non-conserved loop preceding the catalytic histidine (Fig. 3). The human sequence VENG in this loop is replaced by GPSD in zebrafish, which potentially affects substrate specificity at the P2 or P1Ј site (Figs. 2 and 3). Notably also, a little further away from the active site, differences in the charge distribution can be observed (Figs. 2, lower part, and 3).
In conclusion, PCSK7s in less developed organisms display remarkable structural similarities but subtle electrostatic differences with mammalian counterparts. This indicates that studying the zebrafish PCSK7 function can also give insights into other vertebrate homologs.
Inactivation of PCSK7 in Zebrafish Larvae Results in Developmental Lethality-Previous in vitro experimental data show that FURIN can often replace PCSK7 (19,36,37). To compare and contrast how these enzymes regulate zebrafish development, we first interfered with furinA and furinB translation using morpholino technology. In accordance with a previous report (24), the simultaneous blocking of furinA and furinB with MOs resulted in a reduction in ventral jaw structures and consequently an open mouth phenotype (data not shown). Disruption of furinA/B translation also reduced the survival of fish larvae; only 30.6% (22 of 72 fish) of MO-injected fish were alive on 7 dpf. We next designed two distinct morpholinos that block the pre-messenger RNA splicing at exon-intron boundaries around the catalytic site containing exons 3 and 8 in the zebrafish pcsk7 gene (e3 and e8 MOs). Injecting the pcsk7-targeting MOs alone or in combination resulted in numerous defects that reduced the survival of fish larvae dramatically and resulted in 100% mortality by 7 dpf (Fig. 4). The survival rate and gross phenotypes observed were MO dose-dependent, and the development of several organs, including the brain, eye, heart, otolith, and tail, was severely affected.
Because p53-dependent off-target neural toxicity has been estimated to affect 15-20% of all morpholino injections (38,39), we co-injected a p53 MO together with the pcsk7-silencing MOs. The phenotypes and survival of morphant fish remained similar compared with fish not injected with the p53 MO (Fig.  4). This indicates that the observed lethality of the PCSK7 morphant was not due to p53-mediated toxicity. To further validate the specificity of the pcsk7 MO phenotypes, we set up an RNA rescue experiment. When in vitro transcribed pcsk7 RNA was co-injected together with e3 MO into the developing larvae, the severity of the MO phenotypes was reduced, and the survival improved (p Ͻ 0.0001, log rank 2 ϭ 38.94, df ϭ 1) (Fig. 5, A and  B). To verify that the pcsk7 MOs specifically disrupt the splicing of the pcsk7 pre-mRNA molecules, we amplified and sequenced the affected exon regions from the morphant fish (Fig. 6, A and B, and data not shown). Our results demonstrated that injecting the e3 MO results in three mRNA products with different lengths, whereas the e8 MO deletes a 58-bp fragment at the end of exon 8. These data show that pcsk7 morpholinos target pcsk7 pre-mRNA and that pcsk7 has a critical and non-redundant role in zebrafish development.
PCSK7 Regulates Brain, Otolith, and Eye Development-Our expression analyses implicated pcsk7 as prominently present in adult zebrafish eye and brain tissues. To investigate the pcsk7 expression pattern in larvae, we used RNA in situ hybridization on wild-type fish at 2 days postfertilization. These experiments demonstrated that pcsk7 is most abundant in the entire cranial area and the eye already in developing wild-type zebrafish (Fig.  7, A and B).
The observed gross anatomical differences and location of early pcsk7 expression prompted us to analyze in more detail the head region phenotypes seen in the developing zebrafish. A dramatic underdevelopment of the entire brain and eyes was observed in histological sections (Fig. 7, C and D). The cranium of pcsk7 morphants contained mostly a non-cellular, meshlike substance, whereas the brain remained underdeveloped and was located abnormally downward between the eyes. In addi- tion, the cellular structure in the eyes of MO-injected fish was unorganized, and photoreceptor, outer and inner plexiform, and inner nuclear and ganglion cell layers were undefined. However, the retinal pigment epithelium and lens structures were also evident in morphant fish. The zebrafish with a nonfunctional PCSK7 also had an abnormal number of otoliths.
Injection of the e3 MO resulted in a reduction in otolith number (one otolith per ear in Ͼ92% of fish, p ϭ 7.9eϪ63), whereas the fish injected with the e8 MO in contrast had an increased amount of otoliths per ear (three otoliths per ear in Ͼ26% of fish, p ϭ 2.4eϪ11). All evaluated 149 control fish had the normal two otoliths per ear (Fig. 8, A-E). These data show that in  DECEMBER 20, 2013 • VOLUME 288 • NUMBER 51 accordance with its spatiotemporal expression PCSK7 is important for several developmental processes of the vertebrate cranial organs.

PCSK7 in Zebrafish
Analysis of PCSK7-dependent Gene Networks-Genomewide expression analyses can be used to clarify the biological relevance of gene-specific studies. Therefore, to comprehend the gene networks that are affected in the pcsk7 morphant fish, we performed a genome-wide mRNA expression study in whole fish larvae after PCSK7 inhibition. We chose to evaluate gene expression at two different time points, 6 and 24 hpf. At the gastrula stage (6 hpf), most of the zygotic genes have commenced transcript accumulation and often peak in their expression. In contrast, after 24 hpf, the genome-wide gene expression profile remains relatively steady as was demonstrated in a previous analysis with wild-type zebrafish (40).
Conventional PCSK enzymes are highly redundant in substrate processing. Therefore, it is important to evaluate whether disrupting PCSK7 function causes any significant changes in the expression of other proprotein convertases (41), which might contribute to or even compensate for the observed phenotype. At both 6 and 24 hpf, the other conventional pcsk genes were not found to be significantly up-regulated in pcsk7 morphant fish. pcsk7 expression, however, was significantly reduced in the morphant zebrafish (log ratio of Ϫ1.95 and p ϭ 5.77eϪ6 at 24 hpf). The decrease in the pcsk7 mRNA level is presumably due to the cellular deletion of the defectively spliced pcsk7 pre-mRNA molecule.
To understand which major biological, molecular, and cellular processes are dependent on the intact PCSK7 enzyme, we first performed a GO enrichment analysis using hypergeometric distribution testing. At 6 hpf, as expected from the observed morphant phenotypes, terms related to development (for example otic placode formation) and transcription regulation were abundantly enriched. Consequently, at 24 hpf, genes related to metabolism were also commonly up-regulated. Strikingly, at both 6 and 24 hpf, genome-wide mRNA expression samples showed a clear enrichment in immune system-related terms with immune system being the most enriched (p ϭ 0.0004) term at 24 hpf. Several cytokine signaling-affiliated terms could also be found among the most significantly enriched terms at both time points (supplemental Table S1 and Fig. 9).  A more detailed analysis of the PCSK7-dependent genes in the microarray revealed the dysregulation of several immune system-, neurological system-, and otolith/otic vesicle-related genes (supplemental Table S2 and Fig. 9). For example, the expression levels of important regulators of both the adaptive and innate immunity, stat4, tgf␤1a, and csfra, were greatly reduced at 6 hpf in the pcsk7 morphants, whereas otolith/earrelated foxi1 (down-regulated at 6 hpf) and pax2a, msxc, and fsta (up-regulated at 6 hpf) were also dysregulated. In addition, a multitude of differentially expressed neurological genes, including several protocadherin and fibroblast growth factor genes, were found to be dependent on a functional PCSK7.
We have demonstrated previously that another PCSK family member, FURIN, is a key regulator of both T helper 1 type immune responses and T regulatory cell-mediated peripheral immune tolerance (42,43). These crucial events of adaptive immunity are dependent on transcription factor STAT4 function and cytokine TGF␤1 signaling, respectively. We noticed that in our microarray experiment both of the aforementioned genes were significantly repressed when pcsk7 expression was blocked. To confirm this finding, we performed a QRT-PCR analysis that revealed a down-regulation of both tgf␤1a and stat4 by 11.7-and 5.1-fold, respectively, at 6 h postfertilization in an RC versus pcsk7 ϩ p53 comparison (three biological replicates, one-tailed p ϭ 0.037 for both genes in Student's t test with Welch correction; data not shown). These findings imply that, in addition to FURIN, PCSK7 may also have an important regulatory role in developing adaptive immune responses.
PCSK7 Regulates the Expression of tgf␤1a in Zebrafish-In mammals, FURIN is the major proprotein convertase enzyme that regulates the bioavailability of the anti-inflammatory TGF␤1 cytokine (43,44). From previous experiments, it is also noteworthy that TGF␤1 can directly up-regulate the expression of furin (45). However, PCSK7 can also proteolytically process the TGF␤ superfamily cytokines. Mammalian PCSK7 is sug-   DECEMBER 20, 2013 • VOLUME 288 • NUMBER 51 gested to cleave pro-BMP4 in a developmentally regulated fashion, and it possesses up to one-third of the capacity of FURIN in pro-TGF␤1 processing in vitro (15,44). Furthermore, the mRNA expression correlation of human PCSK7 with TGF␤1 is very strong, which may indicate a physiological role for PCSK7 in pro-TGF␤1 processing (46). As described above, we observed that tgf␤1a (ENSDARG00000041502), the zebrafish counterpart for the mammalian TGF␤1, was markedly downregulated in the microarray analysis at 6 hpf in pcsk7 morphant fish (supplemental Table S2 and Fig. 9). This coordinated expression of pcsk7 and tgf␤1a prompted us to address the importance of PCSK7 for tgf␤1a expression and activation and to assess whether the lack of TGF␤1a could contribute to the PCSK7-dependent phenotype.

PCSK7 in Zebrafish
We first wanted to investigate whether zfPCSK7 can directly proteolytically process zfpro-TGF␤1a. To this end, the cDNA encoding TGF␤1a was amplified from wild-type zebrafish and subcloned into a myc-His expression plasmid. After verification by sequencing (data not shown), we then co-expressed tgf␤1a together with zebrafish furinA, furinB, or pcsk7 in FURIN-deficient RPE.40 cells. For comparison, we also co-expressed human TGF␤1 together with human FURIN and PCSK7 cDNAs. Our results clearly demonstrate that in addition to FURIN both zebrafish and human PCSK7s are able to process and promote the release of bioactive TGF␤1 (zebrafish, 16 kDa; human, 14 kDa) into cell culture supernatants (Fig. 10A). In line with previous reports, the PCSK7s possessed approximately one-third of the activity of FURIN in the TGF␤1 maturation (44).
We then assessed whether the tgf␤1a mRNA expression also remains down-regulated in pcsk7 morphant fish in a later developmental stage and found that tgf␤1a gene expression is significantly repressed also at 48 hpf (p ϭ 0.0052; Fig. 10B). To investigate how the lack of TGF␤1a contributes to the pcsk7 morphant phenotype, we injected zebrafish with a TGF␤1ablocking MO. In these experiments, we observed clear similarities between the tgf␤1a and pcsk7 morphants; both showed tail abnormalities, pericardial swelling, and an abnormal number of otoliths (24.4% of tgf␤1a morphants (21 of 86) had three otoliths per ear (p ϭ 2.16eϪ9 in Fisher 2 ϫ 2 test; Fig. 10, C and

D)).
Taking these observations together, our data suggest that in zebrafish the intact PCSK7 enzyme is important for the expression and bioavailability of mature TGF␤1a and that a defect in this contributes to the observed phenotype of the pcsk7 morphant fish.

DISCUSSION
Regardless of extensive studies on proprotein convertase enzymes in vertebrate biology, the function and significance of the evolutionarily ancient PCSK7 has remained largely unclear.
In an effort to fill this gap, we studied the function of PCSK7 in zebrafish and observed that it is critical for the development of zebrafish larvae. The pcsk7 morphant fish display severe developmental defects that lead to 100% mortality within the first 7 days of life. The lack of functional PCSK7 enzyme interferes with the organogenesis of several key elements, including the brain, eyes, and otic vesicles. In addition, our genome-wide gene expression and biochemical analyses demonstrate that PCSK7 regulates genes important for organogenesis and immunology and that it is specifically capable of contributing to the function of cytokine TGF␤1a in developing fish larvae.
To first validate the feasibility of zebrafish as a model for the analysis of the PCSK7 function in vertebrate biology, we surveyed the expression of all identifiable proprotein convertases. In accordance with previous reports in mammals, the zebrafish PCSK enzymes also show variation in their developmental and tissue-specific expression. Zebrafish homologs for the neuroendocrine system-specific PCSK1 and PCSK2 were particularly highly expressed in the fish neural tissues, and the homologs of previously reported ubiquitous enzymes, such as FURIN, PCSK5, and PCSK7, were found to be relatively widely expressed similarly to earlier reports using mammals (17,18,47). PCSK7 shares several common substrate molecules with PCSK5 and FURIN at least in in vitro analyses (48). We found that pcsk7 was co-expressed in different tissues and developing fish together with other PCSK enzymes that have been reported FIGURE 9. Heat map of the most differentially expressed genes and high level summary of gene ontology enrichments between control and pcsk7 morphant fish. Samples were prepared, and data were analyzed as described under "Experimental Procedures." A, zebrafish genes that have an identifiable human homolog are shown in the heat maps. B, summaries from the enriched gene ontology terms. Pie charts show enriched high level categories for cellular component, biological process, and molecular function at 6 and 24 hpf. In each chart, the size of the wedge corresponds to the number of terms enriched under the given high level category. DECEMBER 20, 2013 • VOLUME 288 • NUMBER 51 to compensate its biological function. Taken together, our quantitative pcsk expression data demonstrated obvious similarities between the mammalian and fish pcsk expression. Consequently, deleting the PCSK7 function in developing fish can also give important insights into the specific biological function of this poorly defined proprotein convertase in other vertebrates.

PCSK7 in Zebrafish
We and others have previously reported notable evolutional conservation of the catalytic and P domains in the PCSK enzymes (46,49). To specifically address structural and electrostatic properties in the PCSK7 enzymes, we generated homology models of mammalian and fish PCSK7s and investigated the PCSK7 sequences in several species. In conclusion, our modeling data support the idea that the catalytic and P domains in zebrafish PCSK7 share most structural features with the human counterpart. Therefore again, investigating the fish PCSK7 function is also likely to generate novel insights into the PCSK7 function in other vertebrates.
The crucial role of many of the PCSK enzymes in vertebrate development is indisputable. For example, FURIN-deficient mouse embryos show defective ventral closure and axial rotation and die during the 2nd week of embryonic development (2). In addition, a mutation in furinA causes significant embryonic lethality in zebrafish despite the duplication of the furin gene (24). These fundamental phenotypes can be explained by a lack of processing of PCSK substrate molecules; in early mouse development, the significance of the proper activation of the TGF␤ family cytokines, such as BMPs and NODAL (50), is particularly emphasized. In contrast, a thorough functional analysis of a mammalian model for PCSK7 is not available in the literature. A few scattered references to Pcsk7 knock-out mice, however, suggest either complete redundancy or at least noncritical functions in mammalian development (10 -12). Our analysis of the pcsk7 morphant fish and a recent publication by Senturker et al. (51) using a Xenopus model system demonstrate that PCSK7 is indispensable at least in lower vertebrates. Both studies come to the conclusion that a lack of PCSK7 function leads to severe defects especially in the neural system and eye. In zebrafish, these defects lead to 100% mortality within the 1st week postfertilization.
A complete knock-out or an inactive mutation of the gene would give the definitive answer for the biological relevance of PCSK7 in zebrafish. However, to our knowledge, PCSK7-null zebrafish have not been produced. To overcome this limitation, we analyzed how the lack of an active PCSK7 protein affects the larva development. Using morpholinos to assess the biological function of a protein has provoked some controversy. In some cases, morpholinos can cause unspecific phenotypes due to p53-dependent off-target neural toxicity. However, in our experiments, silencing the p53 pathway did not alter the conclusive role of PCSK7. In addition, sequencing the MO-truncated pcsk7 mRNA showed that the e8 morpholino efficiently removes the end of exon 8 (58 bp) and shifts the reading frame in the catalytic domain. This replaces roughly 50 C-terminal residues of the catalytic domain by 23 residues of non-native sequence followed by a stop codon. In contrast, the e3 morpholino results in the complete deletion of exon 3 of PCSK7. This removes a codon encoding one of the amino acids of the catalytic triad. In addition, the correct reading frame is lost for the remaining protein. As a consequence, PCSK7 proteins translated in both morphants lack the entire P domain, which is needed for correct folding of the enzyme (52). A PCSK7 without the P domain is unlikely to traverse to the secretory pathway, and it might eventually be degraded in the cell.
Co-injecting pcsk7 mRNA with morpholino significantly improved the survival of morphant larvae and partially restored the defected phenotypes, but a complete rescue could not be achieved. The partial improvement of phenotype by RNA res- FIGURE 10. PCSK7 regulates TGF␤1a, which affects the zebrafish larva development and otolith formation. A, FURIN-deficient RPE.40 cells were transiently transfected with zebrafish tgf␤1a-myc or human TGF␤1-myc together with zffurinA, zffurinB, zfpcsk7, huFURIN, or huPCSK7. Pro-TGF␤1 (45 kDa) and mature TGF␤1 (16 kDa in zebrafish and 14 kDa in human) expressions were detected with Western blotting (WB). Mature/pro-TGF␤1 ratios were quantified using NIH ImageJ software, and ratios in cells transfected only with tgf␤1 cDNAs were given an arbitrary value of 1. Equal loading of cell lysates and supernatants was verified by Ponceau S staining (data not shown). The experiment was repeated twice with similar results. B, tgf␤1a mRNA expression was measured by QRT-PCR from pcsk7 e3 ϩ p53 and RC morphant embryos at 48 hpf (**, p ϭ 0.0052, two-tailed Student's t test). C and D depict the phenotypes (4 dpf) of zebrafish injected with tgf␤1a ϩ p53 MOs (tgf␤1a MO, 1.0 pmol; p53 MO, 1.5 pmol). tgf␤1a morphants had incorrect otolith numbers: tgf␤1a MO, 65 fish with two otoliths per ear, 21 fish with three otoliths per ear; RC MO, 120 fish with two otoliths per ear, 0 fish with three otoliths per ear (p ϭ 2.16eϪ9 in Fisher 2 ϫ 2 test).
cue has also been reported by others (53). Therefore, it is probable that a more natural spatiotemporal RNA expression of pcsk7 than what could be achieved using RNA injections into the yolk sac would be needed to cancel out all the effects of pcsk7 MO.
Little information has been available on PCSK7-dependent biological processes at the genomic level. Our GO enrichment analysis revealed several biological, molecular, and cellular functions that were significantly altered if PCSK7 function was inhibited. These findings corroborated the observed gross phenotype of the pcsk7 morphants by highlighting the enrichment of several development-related GO terms. Noteworthy, the enriched GO terms also included the otic placode formation, which is a novel finding and suggests, together with the abnormal amount of otoliths in the pcsk7 morphants, a specific and non-redundant role for PCSK7 in ear development. In the future, it will be interesting to assess whether genetic alterations in human PCSK7 (22) play a role in hearing-or balancerelated phenotypes.
One of the few specific PCSK7 functions in vitro is the rescue of an unstable MHC I-peptide complex (20). Analyses of microarray data on our pcsk7 morphant fish revealed that PCSK7 in addition to having a role in several developmental pathways is indeed linked strongly to immunological processes. Differentially expressed genes included several genes that are key regulators of both the adaptive and innate immunity, such as tgf␤1a, stat4, csf1a, ccr7, and various MHC genes. The host defense-linked GO terms containing these genes were among the most enriched categories already at 6 hpf and remained significantly overrepresented until 24 hpf. It has been shown earlier that TGF␤1 up-regulates the convertase FURIN, which is the major proteolytic activator of this central anti-inflammatory cytokine (44). Interestingly, our genome-wide expression analysis demonstrated that when functional PCSK7 is not available during early development tgf␤1a was among the most down-regulated genes at 6 hpf. The strength of down-regulation was weaker at later phases of development, but tgf␤1a expression remained significantly down-regulated up to 2 days postfertilization.
Importantly, tgf␤1a and pcsk7 morphant fish shared several phenotypic similarities, further suggesting that PCSK7 has a role in the regulation of the TGF␤1a cytokine in developing zebrafish. The detailed mechanisms behind the PCSK7-dependent regulation of the TGF␤1a bioavailability are not fully understood. Mature TGF␤1 cytokine is known to promote its own function by up-regulating the expressions of its own mRNA and the converting enzyme FURIN. Accordingly, because we and others have shown that PCSK7 can also process and activate pro-TGF␤1 at one-third of FURIN proteolytic capacity, the reduced PCSK7 expression can interfere with this feed-forward loop and result in reduced tgf␤1a expression (Fig.  10) (44). However, it is equally probable that in vivo PCSK7 promotes the TGF␤1a function in an indirect manner that does not involve direct proteolysis (21).
In addition to its fundamental inhibitory role in immunity (54,55), TGF␤1 directly controls cell differentiation and proliferation. It is noteworthy that TGF␤1 deficiency in mice causes significant intrauterine lethality (56). Our experiments concur with this multifunctionality of TGF␤1 and suggest that in zebrafish TGF␤1a plays a role in various developmental processes, including otolith formation.
Proprotein convertases have a fundamental role in both health and disease. Interfering with PCSK activity holds promise for future treatment of a plethora of diseases ranging from infections to atherosclerosis. These efforts are often compromised by the lack of specificity of inhibitors of PCSK family members. If the inhibitors are considered for clinical use, it is of utmost importance to also understand the biological significance of PCSK7. Our data presented here underscore the importance of PCSK7 in zebrafish neural development and more specifically genomic processes associated with immunity. These can also be important factors to take into account when extrapolating the unwanted effects of general PCSK inhibitors in disease settings.