Functional Consequences of a Novel Variant of PCSK1

Background Common single nucleotide polymorphisms (SNPs) in proprotein convertase subtilisin/kexin type 1 with modest effects on PC1/3 in vitro have been associated with obesity in five genome-wide association studies and with diabetes in one genome-wide association study. We here present a novel SNP and compare its biosynthesis, secretion and catalytic activity to wild-type enzyme and to SNPs that have been linked to obesity. Methodology/Principal Findings A novel PC1/3 variant introducing an Arg to Gln amino acid substitution at residue 80 (within the secondary cleavage site of the prodomain) (rs1799904) was studied. This novel variant was selected for analysis from the 1000 Genomes sequencing project based on its predicted deleterious effect on enzyme function and its comparatively more frequent allele frequency. The actual existence of the R80Q (rs1799904) variant was verified by Sanger sequencing. The effects of this novel variant on the biosynthesis, secretion, and catalytic activity were determined; the previously-described obesity risk SNPs N221D (rs6232), Q665E/S690T (rs6234/rs6235), and the Q665E and S690T SNPs (analyzed separately) were included for comparative purposes. The novel R80Q (rs1799904) variant described in this study resulted in significantly detrimental effects on both the maturation and in vitro catalytic activity of PC1/3. Conclusion/Significance Our findings that this novel R80Q (rs1799904) variant both exhibits adverse effects on PC1/3 activity and is prevalent in the population suggests that further biochemical and genetic analysis to assess its contribution to the risk of metabolic disease within the general population is warranted.


Introduction
Prohormone convertase 1/3 is a calcium-dependent serine endoprotease essential for the conversion of a variety of prohormones and neuropeptide precursors to their bioactive forms. Human prohormone convertase 1/3 (PC1/3) is encoded by the gene PCSK1, which is located on chromosome 5 and is comprised of 14 exons [1]. PC1/3 is expressed in a subset of endocrine and neuroendocrine tissues, cells equipped with a regulated secretory pathway. During transit through the secretory pathway, PC1/3 is first synthesized in the endoplasmic reticulum (ER) as an inactive 94 kDa zymogen composed of an N-terminal signal peptide, a prodomain which serves as an intramolecular chaperone and inhibitor; a catalytic domain which accomplishes substrate hydrolysis; a P (homo B) domain which contributes to enzymatic properties; and a carboxyl-terminal (CT) domain which, when removed by partial or complete in trans proteolytic processing, results in a much more active, but also less stable, enzymatic form (reviewed in [2] (Figure 1). PC1/3 is abundantly expressed in the arcuate and paraventricular nuclei of the hypothalamus [3,4], tissues that are known to mediate satiety and hunger signals [5]. Substrates of PC1/3, such as proinsulin, proglucagon, proghrelin, agouti-related protein, pro-neuropeptide Y, provasopressin and proopiomelanocortin are responsible for the regulation of absorption, metabolism and acquisition (appetite) of nutrients [6,7,8,9,10,11,12,13,14].
Deficiencies in PC1/3 frequently lead to imbalances in prohormone processing that result in an array of metabolic phenotypes, previously investigated both in mouse models and in humans. Three human subjects have been described with an autosomal recessive disorder (MIM:600955) associated with severe mutations of PC1/3 resulting in early-onset obesity, hyperphagia, hypoadrenalism, reactive hypoglycemia, malabsorptive diarrhea, and hypogonadism [15,16,17]. Interestingly, the PC1/3 null mouse model, unlike the PC1/3-deficient human, is not obese. Although of normal weight at birth, PC1/3 null mice have a high post-natal mortality rate, and those that do survive have a significant reduction in body mass as compared to wild-type animals by the age of 6 weeks. The stunted growth of PC1/3 null mice is believed to be due at least in part to reduced processing of growth hormone releasing hormone (GHRH) and thus reduced circulating levels of growth hormone (GH) [8]. In addition to a reduction in GHRH, the levels of several key neuroendocrine peptides such as ACTH, insulin and glucagon-like peptides-1 and -2 are reduced in these animals due to lack of precursor processing by PC1/3 [8].
While the PC1/3 null mouse is not obese, a mouse model of obesity has been generated via introduction of a missense mutation in PCSK1 at amino acid position 222, near the calcium-binding pocket in the catalytic domain. This hypomorph mutation resulted in obesity, hyperphagia and increased metabolic efficiency due to decreased autocatalytic maturation of the enzyme to smaller molecular weight forms [18]. Three common SNPs in PCSK1 have been identified and associated with obesity. All three SNPs (included in this study for comparison) exhibit moderate effects on catalytic activity in vitro and on natural substrate processing in rat pituitary tumor cells [19,20]. Two of the three non-deleterious SNPs (S690T [rs6235] and Q665E [rs6234]) have been associated with diabetes-related traits [20,21,22].
In the work presented below, the novel variant NP_000430.3:p.Arg80Gln (R80Q; rs1799904), identified and functionally evaluated for the first time here, was compared with previously described SNPs associated with obesity and/or diabetes [rs6234], and S690T [rs6235]) for potentially deleterious effects on the biosynthesis, secretion and catalytic activity of PC1/3. Our data suggest that this novel R80Q variant (rs1799904) deserves further analysis to assess its genetic association with metabolic diseases such as obesity and diabetes.

Databases used and protein structure/function analysis methods
Alleles that varied from the human reference genome build GRCh37 [23] were obtained from the dbSNP [24], 1000 Genomes [25], NHLBI [26], and NIEHS [27] datasets and were merged into a custom SQL database. dbSNP data were compiled from various sources, with allele frequencies available only for a subset of variants. The 1000 Genomes dataset was based on both low coverage whole genome and higher coverage exome sequencing of 1092 individuals. The NHLBI and NIEHS data were obtained from exome sequencing of 6500 and 95 individuals respectively. Population allele frequencies were calculated using the combined datasets wherever allele counts were present. Variations in PCSK1 (chr5:95726119-95769847) were identified and analyzed with the Ensembl Variant Effect Predictor version 2.6 [28] and Ensembl database homo_sapiens_variation_68_37 [29] to determine the effect of the variant on the transcript. Nonsynonymous codon substitutions were analyzed using the SIFT [23,30,31,32,33], PolyPhen [34,35,36], and Condel [37]

Expression vector construction/mutagenesis
Flag-tagged human PC1/3, a kind gift of J. W. Creemers [20] was modified by site-directed mutagenesis using the Stratagene  QuikChange method [38] to introduce the mutations shown in Figure 1. All mutations were verified by sequencing of the entire PC1/3 cDNA insert.

Transient transfection of PC1/3 variants
To assess the biosynthesis and secretion profiles of PC1/3 variants in a cell line that does not express endogenous PC1/3, Ad-293 (Stratagene) HEK cells, plated at a density of 2610 5 cells per well in 24-well plates, were transfected with plasmids encoding either wild-type or variant PC1/3s in triplicate wells. Cells were transfected with 200 ng of plasmid DNA per well using Lipofectamine (Invitrogen, Carlsbad, CA). To assess effects in a regulated neuroendocrine cell line (also lacking expression of endogenous PC1/3), Neuro-2A cells (ATCC, cat. No. CCL-131) were transfected in triplicate with the same protocol using Lipofectamine 2000 (Invitrogen, Carlsbad, CA). For both cell lines, five hours post-transfection, 1 ml of growth medium was added to each well and incubation continued for an additional 24 h. Cells were then washed with PBS and 0.3 ml of Opti-MEM (Invitrogen, Carlsbad, CA) containing 100 ug/ml bovine aprotinin (Desert Biologicals) was added to each well. Cells were incubated for an additional 18-24 h before conditioned medium and cells were harvested. Conditioned medium was analyzed first by enzyme assay; both cells and medium samples (for HEK cells) and medium samples (for Neuro-2A cells) were then subjected to SDS-PAGE followed by Western blotting using primary antiserum against the amino terminus of mature mouse PC1/3 [39]. Mouse monoclonal anti-ß-actin antiserum (Sigma-Aldrich, St. Louis, MO) was used to assess cellular actin levels as a loading control. Western blots were then probed with horseradish peroxidase-coupled secondary antiserum. Visualization of immunoreactive protein was accomplished using the SuperSignal West Femto Maximum Sensitivity Substrate kit (Thermo Scientific, Rockford, IL). Maximum rates were obtained from the linear portion of the kinetic measurement curves. Specific activities of PC1/3 proteins in the conditioned medium were determined by dividing maximum rates by band intensities of total secreted immunoreactive protein, each determined in triplicate, and quantified with an Alphaimager 3300 (Alpha Innotech Corporation, San Leandro, CA) imaging system.

Analysis of public databases; structure-function analysis
A total of 1020 allelic variants (data not shown) within the PCSK1 gene were found in the public databases, of which 54 were potentially consequential splice site or missense variants ( Table 1). Thirty-seven non-synonymous substitutions were predicted to be possibly or probably deleterious by at least one model (SIFT, PolyPhen, or Condel, where Condel represents a consensus modeling program). Two of the three previously described variants are common, with MAFs of 23.7% for S690T (rs6235) and 25.0% for Q665E (rs6234), whereas the N221D SNP (rs6232) is less common (MAF = 3.3%). None of these three variants were predicted to be deleterious using SIFT, PolyPhen, or Condel. In contrast, the novel variants that were predicted as ''possibly'' or ''probably'' deleterious were unique to one sample or were observed with very low frequency (minor allele frequencies (MAFs) of 0.008%-0.87%. In addition we considered a frameshift variant The R80Q (rs1799904) variant that differed from the human reference genome and was predicted to have a potentially consequential effect on the transcript was selected from the dbSNP 137, 1000 Genomes, NHLBI, and NIEHS public datasets.  N586TfX4 (g.5:95730696), which exhibited an unusually large MAF of 6.1% in a previous release of the 1000 Genomes data. We selected the most common novel variant R80Q (rs1799904; MAF = 0.87%), and N586TfsX4 for genomic sequencing and potential functional studies, comparing them with already described common variants of PC1/3.

SNP validation by sequencing
Genomic DNA from two individuals homozygous for the most common variants was obtained from the Coriell Institute and subjected to Sanger sequencing. The DNA sample containing rs1799904 (R80Q (g.5:95764963C.T; c.239G.A) was found to be homozygous for the R80Q mutation in exon 2 (Figure 2), while the N586Tfsx4-containing SNP (g.5:95730696TC.T; c.1755delG) was determined to be a false positive (i.e. no frameshift mutation was found in in exon 13) (data not shown).

Secretion and biosynthesis of PC1/3 variants
In order to assess whether the novel variant R80Q (rs1799904) affected the biosynthesis or secretion of PC1/3, expression vectors encoding wild-type and variant PCSK1s were transiently transfected into HEK and/or Neuro-2a cells (both lines lack expression of endogenous PC1/3). PC1/3 proteins containing the previously described S690T/Q665E (rs6234/rs6235) pair, as well as the individual S690T and Q665E SNPs, did not exhibit significantly altered expression and secretion patterns as compared to wild-type PC1/3. The N221D (rs6232) substitution resulted in reduced secretion and cleaved forms of PC1/3 in the medium (Figure 3). The secretion profile of the R80Q (rs1799904) substitution differed  from wild-type PC1/3, in that the 74 and 66 kDa lower molecular weight forms of PC1/3 were absent from the medium (in HEK cell experiments) or reduced (in Neuro-2a cell experiments), although the total level of secreted PC1/3 was not reduced.

Catalytic activity of PC1/3 variants
To determine the impact of these variations on PC1/3 catalytic activity, conditioned medium of HEK cells transfected with either empty vector, variant PC1/3s, or wild-type PC1/3 was subjected to a fluorogenic assay. Maximum rates of fluorogenic substrate cleavage were normalized using the band intensities of secreted PC1/3s in order to determine the specific activity of each variant relative to wild-type PC1/3. The S690T/Q665E (rs6234/rs6235) and S690T (rs6234) amino acid substitutions did not significantly alter specific activity (95% confidence level; p.0.13). The Q665E substitution alone resulted in a small but significant 27% decrease in specific activity as compared to wild-type (p = 0.05). The N221D (rs6232) substitution decreased specific activity by 36% (p = 0.02), and the R80Q variation resulted in a 38% decrease (p = 0.02) (Figure 4). When expressed in Neuro-2a cells, the R80Q (rs1799904) variant resulted in a 42-48% decrease (p,0.0001) in activity as compared to wild-type PC1/3 ( Figure 5).

Discussion
In studies of European populations, PCSK1 represents the third most important gene contributing to extreme obesity [40]. Functional studies of certain SNPs associated with obesity that impose modest or no significant effects on PC1/3 function in vitro have supported the idea that even slight variations in PC1/3 activity can predispose an individual to higher risk of obesity [20]. Individuals who are compound heterozygotes or are homozygous for rare severe deleterious mutations in PCSK1 suffer from multidimensional disease states, including small intestinal dysfunction, hyperphagia and obesity [15,16,17]. Even heterozygous mutations which result in functional enzymatic changes have been linked to obesity, despite the presence of a normal allele [40]. The mechanism by which modest deficiencies in PC1/3 activity can lead to such profound phenotypes when present on a single allele remains unknown. A closer look into the complex biochemistry of commonly found variations of this enzyme may provide answers to these questions. In this work, we have analyzed public databases for other less common and rare deleterious variants and identified the variant R80Q (rs1799904), and have compared the effects of this variant to those of known polymorphisms.
Consistent with previous studies [19,20], we found that the amino acid substitutions S690T/Q665E (rs6234/rs6235) did not significantly alter the specific activity or biosynthesis and secretion of PC1/3 in HEK cells. The Q665E substitution alone did result in a slight decrease in specific activity as compared to wild-type enzyme, and may represent the more detrimental of the two mutations (S690T/Q665E), which were previously identified as a paired SNP associated with a higher risk of developing obesity and diabetes [19,20,21]. In our hands, the N221D (rs6232) substitution decreased specific activity by a somewhat greater extent than previously reported, possibly due to differences in enzyme assay methods [20].
However, of all of the variants we analyzed in HEK cells, the novel R80Q (rs1799904) variant exhibited the most detrimental effects on PC1/3 maturation and specific activity. This variant yielded an 87 kDa product in the conditioned medium that did not undergo further carboxy-terminal processing to the more active 74 and 66 kDa forms, resulting in an enzyme with significantly lower specific activity, similar to the more common obesity risk N221D (rs6232) variant. This novel R80Q variant exhibited an even more pronounced decrease in specific activity when expressed in a cell line containing a regulated secretory pathway (Neuro-2a), where wild-type PC1/3 is likely able to achieve greater specific activity through more complete maturation to its lower molecular weight forms within regulated secretory vesicles. The lower molecular weight forms of PC1/3 exhibit a different substrate specificity than full-length 87 kDa PC1/3 [41]; this could be an important mechanism for SNPs to exert functional effects. Another possible functional consequence of altering the profile of active species is a change in enzyme stability, since carboxy-terminally truncated species are known to be more labile than the 87 kDa form (reviewed in [2]). Since the C-terminal region of PC1/3 has been implicated in targeting of this enzyme to secretory granules [42,43], altered C-terminal processing may also result in changes in enzyme distribution. Further studies using immunocytochemistry in transfected Neuro-2A cells will shed additional light on this question. The proPC1/3 maturation process begins with the autocatalytic intramolecular cleavage of the pro-domain in the ER at the primary cleavage site, RSKR 107-110 [44,45]. This cleavage yields an 87 kDa form of PC1/3 that, by analogy with the related enzyme furin [46] likely remains associated with its own prodomain through non-covalent interactions until its arrival at the trans-Golgi network. Although this has not yet been strictly demonstrated for PC1/3, the PC1/3 prodomain most likely assists in the folding of the catalytic domain and in enzyme inhibition during secretory pathway transport [47,48,49,50,51,52]. If prodomain processing of PC1/3 occurs similarly to that of furin, trans-Golgi network protonation of a histidine in the vicinity of the secondary cleavage site (RRSRR 77-81 ) then results in secondary site cleavage at R 81 , followed by dissociation of prodomain fragments from PC1/3 [53,54]. The inhibitory role of the prodomain is of particular interest to this study when we consider the location of the R80Q (rs1799904) substitution within the secondary cleavage site of the prodomain (Figure 1). Independent studies have shown that alteration of mouse proPC1/3 prodomain residues either within or surrounding cleavage motifs can affect propeptide processing; the in vitro proteolytic conversion of an R80A mutant propeptide (the same residue as the R80Q variant studied here) by wild-type enzyme was impaired compared to wild-type propeptide [44]. Given this finding, our lack of identification of propeptide-bearing R80Q PC1/3 is puzzling. We have previously found that a portion of newly synthesized proPC1/3 is subjected to endoplasmic reticulum-associated degradation [52]; this might represent the fate of this molecular species. Collectively, these data support the idea that residues within the secondary cleavage site, including the novel variant studied here, contribute to the proper processing of proPC1/3.
The novel R80Q (rs1799904) variant (MAF = 0.87%) is about one-third as common as the N221D (rs6232) SNP (MAF = 3.3%). Although less common, the R80Q variant should be subjected to further analysis to evaluate its influence on insulin sensitivity, proinsulin conversion and the risk of developing obesity, similarly to the effect of the N221D (rs6232) SNP [20,22]. We note that 119 individuals in the public datasets have other, less common and rare variants of PCSK1, most of which are predicted to have some detrimental effect on protein function. This mutational burden on the population is not trivial and may also play a role in susceptibility to obesity or other disorders. The importance of rare variants in common disorders is not clear at present, but advances in massively parallel sequencing and computational analysis may soon shed additional light on this question.
In conclusion, we show that the novel PCSK1 variant R80Q (rs1799904) exhibits deleterious effects on PC1/3 maturation. This PC1/3 variant exhibits decreased catalytic activity as compared to wild-type PC1/3 and to previously described obesity risk SNPs; therefore, it may contribute to a higher risk of metabolic disease in the general population. Our results suggest that further study of less common and rare variations in PCSK1 from both biochemical and genetic standpoints will be useful in elucidating the mechanisms by which variant PC1/3s contribute to metabolic diseases such as obesity and diabetes.