Structure-based activity prediction of cyp21a2 stability variants: a survey of available gene variations

ABSTRACT Congenital adrenal hyperplasia due to 21-hydroxylase deficiency accounts for 90–95% of CAH cases. In this work we performed an extensive survey of mutations and SNPs modifying the

coding sequence of the _CYP21A2_ gene. Using bioinformatic tools and two plausible CYP21A2 structures as templates, we initially classified all known mutants (n = 343) according to their

putative functional impacts, which were either reported in the literature or inferred from structural models. We then performed a detailed analysis on the subset of mutations believed to

exclusively impact protein stability. For those mutants, the predicted stability was calculated and correlated with the variant’s expected activity. A high concordance was obtained when

comparing our predictions with available _in vitro_ residual activities and/or the patient’s phenotype. The predicted stability and derived activity of all reported mutations and SNPs

lacking functional assays (n = 108) were assessed. As expected, most of the SNPs (52/76) showed no biological implications. Moreover, this approach was applied to evaluate the putative

synergy that could emerge when two mutations occurred _in cis._ In addition, we propose a putative pathogenic effect of five novel mutations, p.L107Q, p.L122R, p.R132H, p.P335L and p.H466fs,

found in 21-hydroxylase deficient patients of our cohort. SIMILAR CONTENT BEING VIEWED BY OTHERS COMPUTATIONAL ANALYSIS OF ANDROGEN RECEPTOR (AR) VARIANTS TO DECIPHER THE RELATIONSHIP

BETWEEN PROTEIN STABILITY AND RELATED-DISEASES Article Open access 21 July 2020 NONSENSE VARIANT OF _NR0B1_ CAUSES HORMONE DISORDERS ASSOCIATED WITH CONGENITAL ADRENAL HYPERPLASIA Article

Open access 09 August 2021 GENOMIC AND SEQUENCE VARIANTS OF PROTEIN KINASE A REGULATORY SUBUNIT TYPE 1Β (PRKAR1B) IN PATIENTS WITH ADRENOCORTICAL DISEASE AND CUSHING SYNDROME Article 08

September 2020 INTRODUCTION Congenital adrenal hyperplasia (CAH) due to 21-hydroxylase deficiency (OMIM 201910) represents 90–95% of all CAH cases1,2,3. This autosomal recessive disorder,

which is the most frequent inborn error of metabolism, has a broad spectrum of clinical forms, ranging from severe or classical, to mild late onset or non-classical (NC). The classical form

includes the salt-wasting (SW) and simple virilizing (SV) of early onset. Girls with classical CAH are typically born with ambiguous genitalia. Patients with NC CAH exhibit clinical

manifestations of hyperandrogenism. The gene encoding 21-hydroxylase, _CYP21A2_ is mapped to the short arm of chromosome 6 (6p21.3), within the human leukocyte antigen (HLA) complex, in the

so-called RCCX module. Approximately two thirds of the chromosomes analyzed have a duplicated RCCX module that includes a genomic DNA segment composed of the pseudogenes _RP2, CYP21A1P,

TNXA_, and a second active copy of the _C4_ (long or short) gene4,5. _CYP21A1P_ shares 98% sequence identity with _CYP21A2_. Due to the high degree of sequence identity between the gene and

its pseudogene, most of the disease-causing mutations described in 21-hydroxylase deficiency are likely to be the consequence of non-homologous recombination or gene conversion events6,7.

Although most of the patients carry _CYP21A1P_-derived mutations, an increasing number of naturally occurring mutations have been found in disease-causing alleles (see:

http://www.hgmd.cf.ac.uk for details). Mutations in the _CYP21A2_ gene cause varying degrees of 21-hydroxylase activity loss. _In vitro_ studies revealed that mutations leading to a complete

inactivation of 21-hydroxylase are usually associated with the SW phenotype. Mutations that reduce enzyme activity close to 2% cause the SV phenotype, whereas those with a residual

enzymatic activity in the range of 10% to 60% result in the mild NC CAH phenotype. In addition, a great number of patients are compound heterozygotes carrying different _CYP21A2_ mutations

on each allele, and their phenotypes depend on the milder gene defect8. 21-hydroxylase belongs to the cytochrome P450 protein family, a huge and diverse family found in bacteria, archaea and

eukaryotes. In humans, there are 57 genes and more than 59 pseudogenes grouped in 18 families and 43 subfamilies, all with a high sequence identity9. 21-hydroxylase displays microsomal

localization10 and, like other microsomal P450s, this enzyme accepts electrons provided by a NADPH-dependent P450 oxidoreductase (POR), reducing molecular oxygen and hydrolyzing substrates.

This enzyme has 494–495 aminoacids with a molecular weight of 52 kDa11,12. Over the last few years, much progress has been made towards predicting protein stabilities and correlating them to

protein activities13,14,15,16,17. Homology modeling and fast energetic calculations have emerged as useful tools to evaluate, through structure-based methods, the impairment of protein

stability. Human 21-hydroxylase models have been built based on the available low homology CYP protein families13,16. With the aim of predicting the effect of newly uncharacterized mutations

with improved accuracy, we have developed and evaluated a procedure based on the high identity bovine and human templates18,19. Using bioinformatic tools and either the human crystal

structure or a model based on the bovine CYP21A2 counterpart, we initially classified all mutants in coding regions according to their putative role in protein dysfunction and/or location in

the structure and focused our analysis on those affecting protein stability. Using this approach, we estimated _in silico_ the residual activity of mutants that lack functional assays.

Furthermore, we estimated the effect of double mutations/SNPs, located _in cis,_ on P450CYP21 protein stability. In addition, we propose the putative pathogenic effect of five novel

mutations, p.L107Q, p.L122R, p.R132H, p.P335L and p.H466fs, found in 21-hydroxylase deficient patients of our cohort. RESULTS SURVEY OF CYP21A2 REPORTED VARIANTS With the aim of predicting

the effect of uncharacterized mutations, we initially performed an extensive survey of mutations and SNPs modifying the coding sequence of the gene (n = 343). Using either the human crystal

structure (PDB ID:4Y8W) or a model based on the high identity crystal structure from the bovine protein (PDB ID:3QZ1), the variants were classified according to their proposed effect on

protein dysfunction and/or location in the structure (Supplementary Information, Table S1). CORRELATION OF PROTEIN STABILITY AND CYP21A2 ACTIVITY We focused our analysis on mutations assumed

to be involved in protein stability (148 variants), under the hypothesis that protein destabilization affects enzymatic activity. Of these, we initially selected variants with experimental

enzymatic activity reported until 2013 (n = 30, see Table S1). Using the FoldX algorithm, the predicted free energy of each of the mutants relative to the wild type counterpart (∆∆G) was

plotted against the natural logarithm (ln) of the _in vitro_ activity as previously reported16. As shown in Fig. 1, a good correlation between FoldX’s predictions and experimental activity

was obtained. Correlation was higher when using the bovine-based model (R2 = 0, 79) than for the human crystal structure (R2 = 0, 60). As a reference, we compared the _in vitro_ activity for

an overlapping set of mutants with the predicted stability reported by two previous models, obtaining a R2 = 0, 48 (n = 13) with the rabbit CYP2C513 and a R2 = 0, 68 (n = 7) with the bovine

CYP21A220. To validate our method, we estimated the _in silico_ enzymatic activity of mutants with experimental activities reported from 2014 up to the present21,22,23,24, excluding

residues known to impair protein function independently of protein stability. To accomplish this, the predicted residual activity after calculation of ∆∆G was compared with the activity

reported in functional assays and with the patients’ phenotypes. As shown in Table 1, in 5 out of 10 mutations the predicted _in silico_ residual activity was similar to the experimental

results. Nevertheless, a close look at the remaining 5 mutations revealed that, contrary to the _in vitro_ assay results, the _in_ silico predicted p.P45L activity may correlate better with

the patient’s phenotype (Table 1). For the p.K102R variant, the predicted activity using the bovine-based template might be related to a NC CAH allele. Nevertheless, the _in silico_ activity

predicted using the human crystal structure is consistent with the fact that p.K102R has long been considered a common polymorphism25. For the p.M150R variant, both computed _in silico_

activities may predict severe alleles, whereas discordant results were reported _in vitro (_17%, NC; 4%, SV). Finally, for the p.M283V variant the values of the _in silico_ activities were

twice the residual enzymatic activities found _in vitro_. Nevertheless, both are activities found in mild alleles. _IN SILICO_ PREDICTION OF RESIDUAL ENZYMATIC ACTIVITY Using this procedure

we predicted the _in silico_ residual enzymatic activity of 32 CYP21A2 mutations20,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43 that lack functional assays and are putatively

involved in protein destabilization. The predicted activities were calculated from the linear fit based on the bovine template, which showed a better correlation with the experimental

activities (see above). As shown in Table 2, in 10 mutations, the _in silico_ activity was in accordance with the allele type expected from the observed phenotype. However, in 9 of 32

mutations such correlation could not be assessed due to the lack of information of the mutation in the homologous allele and/or the patient’s phenotype. We extended our approach to the

_CYP21A2_ variants (n = 76) deposited as SNPs in public databases (Supplementary Information, Table S2). All the predicted enzymatic activities above 75% were considered to be

non-pathogenic. Fifty-two SNPs disclosed no biological implications. STABILITY PREDICTION OF DOUBLE MUTANTS/SNPS LOCATED _IN CIS_ We aimed to identify CYP21A2 variants in which two mutations

can combine to generate a severe effect and to identify variants presenting a synergistic effect, namely, improving or impairing the enzymatic activity to a greater extent than expected by

the sum of each individual mutation. To this end, we analyzed the effect of two allelic variants (mutation-mutation, mutation-SNP or SNP-SNP) found _in cis_ on protein stability. Of all

possible combinations for mutant-mutant or mutant-SNPs, we considered the following scenarios: 1) there is a trivial situation in which both mutations, as well as their addition, are above

the cut-off value (1.6 kcal/mol), and consequently we expect a pathogenic effect (most cases reported, not shown); 2) another possibility is that neither of the mutations nor SNPs exceeds

the cut off value but their sum indeed does, in which case we predict pathogenicity derived from impaired enzymatic activity (Supplementary Information, Tables S3A and S4A); 3) and finally,

when only one of the mutations exceeds the cut off value but combined with another mutation or SNP their sum drops below the cut-off value, in which case we predict a non-pathogenic effect

(Supplementary Information, Tables S3B and S4B). Interestingly, FoldX allowed us to identify synergistic effects (either positive or negative) that may suggest a change in the classification

of effect of the mutation. For example, a negative synergistic effect is one in which the sum of the effects of two mutations/SNPs exceeds the cut off, but their combined analysis by FoldX

results in a lower value (Supplementary Information, Tables S3C,D and S4C,D), whereas a positive synergistic effect is one in which the sum of two mutations or SNPs does not exceed the

cut-off value, but their combined analysis with FoldX does (Supplementary Information, Table S4E). Strikingly, there are cases in which neither of the mutations, nor their sum, exceeds the

cut-off value, but FoldX nevertheless predicts a synergistic effect and thus, pathogenicity (Supplementary Information, Table S4F). When SNP-SNP double mutants were analyzed, none of them

presented values above the cut-off (not shown). The combination of p.K102R with p.S268T, each with a small contribution to destabilization (0.81 and 0.74 kcal/mol, respectively), did however

result in a value close to that of the cut-off (1.55 kcal/mol). STRUCTURE-BASED PREDICTED EFFECT OF NOVEL MUTANTS We analyzed the putative structural and functional effects of 5 novel

mutations, p.L107Q, p.L122R, p.R132H, p.P335L and p.H466fs, found in patients from our cohort (see Supplementary Information Table S5 for details on patients phenotypes and genotypes). None

of these novel variants were found in the 1000 Genome Database. In addition, all but one, p.P335L, are located in CYP21 protein residues that are highly conserved throughout mammalian

species (Supplementary Information, Figure S1). Figure 2 shows the structural analysis of the novel point mutations on the human protein. As shown, the side chain of L107 is located very

close (4.31Å) to the propionate moiety of the heme group (Fig. 2D). The introduction of glutamine’s positive charge might disrupt hydrophobic interactions that stabilize the heme group. The

mutation in residue 122 replaces a hydrophobic leucine with a bulky and positively charged arginine residue, thus modifying the electrostatic potential surface of the protein. Conversely,

the change of a positively charged arginine to a histidine residue in position 132 might cause a decrease in the density of positive charge on the protein’s surface (Fig. 2B,C). Both changes

are positioned in regions where several amino acids have been suggested to interact with the POR13. Thus, we expect changes in electrostatic potential to significantly affect

protein-protein interactions. Residue P335 (Fig. 2E) is located in a loop between helices K and L19. The change from a proline residue to a leucine does not introduce a charge modification

in the region and neither heme/ligand nor POR interactions are involved, although a spatial displacement of the loop upon mutation cannot be ruled out. We classified this mutation as being

putatively involved in protein stability. We found a ∆∆G of −2.16 ± 0.15 kcal/mol for the bovine-based template and consequently the _in silico_ predicted activity is ≥100%. Similar results

were obtained when modeling residues involved in novel point mutations using the bovine template (Supplementary Information, Figure S2). Lastly, the p.H466fs mutation causes a frameshift in

the carboxy-terminus of the protein, resulting in a completely different tract of 59 residues and introducing 27 additional amino acids to the protein. Consequently, this variant cannot be

accurately modeled by the present approach; notwithstanding, a nonfunctional protein could be expected. DISCUSSION The adrenocortical 21-hydroxylase is one of the key enzymes in

glucocorticoid and mineralocorticoid biosynthesis, and mutations in the _CYP21A2_ gene cause the CAH as a result of 21-hydroxylase deficiency. Most of the reported mutations in the coding

region result in aminoacid substitutions that may disturb essential functional and structural motifs of the protein (http://www.hgmd.cf.ac.uk). Activity impairment will also be evident when

the mutation affects the correct folding and stability of the protein and thus its availability in the cell. Furthermore, an even more subtle case can be envisioned in which the protein’s

activity is impaired by mutations that alter protein dynamics and thus, its behavior in the cell. There are several examples of structure-based studies correlating specific aminoacidic

change in CYP21A2 and other proteins with the severity of the encoded allele. CYP21A2 studies were initially based on low-identity templates13,16, and then repeated when a high-identity

bovine protein18 was made available20. Recent publication of the human CYP21A2 crystal structure19 encouraged us to improve and expand our analysis by including this structure as template.

It is worth noting that most of our test data sets are obviously biased towards mutations with a deleterious effect as they come from clinical cases. For a subset of these mutations,

functional assays have been performed, demonstrating their involvement in the pathogenicity of the disease. Nevertheless, for some of the reported variants, no information is available.

21-hydroxylase deficiency is a recessive disorder and most of the patients are compound heterozygotes with different mutations in each allele. Thus, a detailed description of the putative

severity of a mutant protein must be carefully considered within the context of the mutation on the homologous chromosome and the patient’s phenotype. Considering a protein length of 494–495

amino acids for the human CYP21 and only 20 different residues, the number of possible mutants is between 49420 and 49520. So far, only 210 of them have been seen in patients and, among

these, approximately 33% are assumed to be involved in protein stability and could be evaluated by our method. In addition, 76 allelic variants presumably involved in protein destabilization

were found in individuals from the general population, and their putative implications on protein activity are not known. We believe that activity prediction for variants could be useful to

partially understand their pathophysiological implications. This is particularly relevant in the case of mild double mutants _in cis_ in which the combined effect is unknown but may be

predicted. In the future, this method could also be extended to several excluded positions depending on the development of algorithms capable of readily predicting heme and ligand

interactions, and the availability of templates with different ligands (including interacting partners, such as POR). This sort of extension has been implemented to successfully assess

protein-DNA interactions15,44. It is expected that both the sequence identity and structural resolution improve energetic predictions. Surprisingly, in our analysis, stability calculations

based on the bovine crystal structure resulted in a better correlation to experimental measurements than those based on the human counterpart. We hypothesize that these unexpected results

could be related to the experimental conditions in which each structure was obtained. X-Ray diffraction crystallography provides a picture of the lowest energy conformation for a given

protein, usually biasing our interpretation to a unique and rigid entity. Furthermore, proteins that interact with several ligands often adopt different conformations depending on the

binding partner. In such cases, it is useful to have several structures with the different ligands or NMR data to further understand the conformational states of the system. In our study we

have worked with only two structures: the bovine protein bound to 17-hydroxyprogesterone (17-OHP) and the human CYP21A2 bound to progesterone, each of which may represent a biased

conformation that may in turn affect the structure-based energetic calculations. Thus, considering that proteins are dynamic entities that fluctuate and interact with several ligands in the

cell, the overall behavior may be better grasped by a structure that could represent protein fluctuations around the energetic minimum rather than a structure representing the minimum itself

as observed in the static crystal. Remarkably, when we validated our method taking into account the most recent mutations with functional assays reported, half of them were in agreement

with _in vitro_ residual activities. Only 1 out of the 10 variants, p.R149C, was found to be a completely discordant result, and contrary to the functional assays, some of the _in silico_

results may better represent the patient’s phenotype. We then proceeded to predict stability and associated activity for all reported mutations or SNPs not expected to be involved in other

processes except for protein stability. As expected, most of the SNPs described in population-based studies were found to have no biological implications. Moreover, when a correlation of the

_in silico_ results and the expected activity from the observed phenotype could be assessed, we found consistency in 10 of the variants analyzed. These results reinforce our approach as a

useful tool for predicting residual activities of uncharacterized allelic variants. A number of 21-hydroxylase deficient alleles have been reported with two mutations occurring _in cis_.

Thus, we extended our method to analyze CYP21A2 variants that can combine to generate a severe and/or synergistic effect, including those reported to have no biological effect on protein

activity (SNPs). Though there are very few experimental results of the final residual activity of _in-cis_ double mutants to compare with, we predicted several combinations to have a severe

and/or synergistic effect of a rather small magnitude. Strikingly, the combination of p.K102R with p.S268T, two variants classified individually as non-pathogenic in experimental assays24,45

reaches a destabilization value close to the cut-off. This is particularly important, since both variants are described in a great number of individuals from the general population46.

Though the absence of mutations in patients diagnosed with 21-hydroxylase deficiency has been described previously, the putative pathogenic effect of two SNPs _in cis_ has yet to be

considered. Indeed, several authors have suggested that p.S268T could result in a decreased enzymatic activity when presented _in cis_ with another polymorphic variant47,48. The

structure-based approach enables us the prediction of the effect of mutations that modify protein stability. For mutations impairing activity by other means, it is necessary to perform _in

vitro_ activities or develop computational tools that can explicitly model interactions with ligands (e.g. substrates, the heme group or other proteins). From the set of newly described

mutants, only p.P335L was expected to affect stability. Nevertheless, we found that the predicted activity of this variant should be close to that of the wild-type protein since no

destabilization change was found. Sequence comparisons demonstrated variability in the 335 residue throughout CYP21 proteins of different mammalian species. Indeed, a leucine residue is

located at this position in the mouse and the rat proteins. In addition, _in vitr_o studies have suggested that the presence of two mild mutations _in cis_ is generally associated with a

severe impairment of enzymatic activity23,49,50,51. However, the patient with the p.P335L variant presented a NC phenotype, carrying a large gene conversion/deletion of the _CYP21A2_ gene on

the homologous allele (null allele) and the mild p.V281L mutation _in cis._ Taken together, these observations prompted us to suggest that this allelic variant may not influence the

residual activity of the protein, in agreement with the _in silico_ predictions. However, caution should be taken considering that a significant stabilization was found for this variant and

a large stabilization could also affect protein degradation or its dynamics, and thus protein function. In the post-genomic and personalized medicine era a large amount of genetic

information is expected to accumulate. The development of an efficient tool to analyze this information is of utmost importance. In particular, connecting sequence information to phenotypic

effects is an ongoing effort that could assist physicians in the near future. So far, most of the information on mutants is biased towards pathological effects but sooner or later an immense

amount of uncharacterized variants will be described with the massive sequencing approaches currently underway. Until statistics or other methods can be used with enough confidence, we

propose that _in silico_ activity prediction using structure-based analysis could be a valuable tool, being particularly relevant in the case of double variants occurring _in cis_. MATERIALS

AND METHODS ETHICAL APPROVAL All the procedures performed in this study were in accordance with the ethical standards of the institutional and/or national research committee and with the

1964 Helsinki declaration and its later amendments or comparable ethical standards. Written informed consent was obtained from all patients and parents involved in this work. The study was

approved by the ethics committee of the Administración Nacional de Laboratorios e Institutos de Salud (ANLIS), Buenos Aires, Argentina. _CYP21A2_ GENOTYPING Details on methods in mutation

genotyping and sequence alignments are presented in Supplementary Information. SURVEY OF ALLELIC VARIANTS AND MUTATIONS IN HUMAN CYP21A2 Mutations and allelic variants in the coding region

of the CYP21A2 were extracted from the Human Cytochrome P450 Allele Nomenclature database (www.cypalleles.ki.se), and from the bibliography. In addition, database of single nucleotide

polymorphisms (SNPs) and multiple small-scale variations that include insertions/deletions, microsatellites and non-polymorphic variants (http://www.ncbi.nlm.nih.gov/projects/SNP/)52, as

well as the 1000 Genome Database53 were also consulted. When available, the _in vitro_ activity for progesterone (P) and/or 17-hydroxyprogesterone (17-OHP) were included. When more than one

activity was reported for the same mutation, we considered references that provided measurements of residual activity exclusively in _ex vivo_ systems (in COS1 or COS7 cells). In addition,

when possible, the most accurate, newest and better related to patient’s phenotype was preferred. The full list of mutations/SNPs considered is listed in Supplementary Information Table S1.

MODEL BUILDING AND ASSESSMENT In addition to the human crystal (PDB ID: 4Y8W), a model of the human CYP21A2 protein based on the structure of the bovine CYP21A2 (PDB ID: 3QZ1) was generated

using MODELLER version 9.1154 including heme cofactor explicitly. Amino acid sequence was taken from UniProt (NP_000491.4) while alignments were performed with MEGA 4 software55. A

significant improvement of the model was obtained by manually displacing L129 in the automatic alignment and iterative loop refinement. The use of multiple templates of other proteins of the

CYP family did not improve the model (not shown). Model’s quality was assessed by DOPE56, QMEAN Z-scores57 and Ramachandran plots58. Protein model is available at

www.modelarchive.org/project/index/doi/ma-anifj. _IN SILICO_ MUTAGENESIS Mutations were generated and analyzed using FoldX 3.0 Beta 5.1 (foldx.crg.es)59. Repair PDB FoldX command was used to

optimize the total energy of the protein to FoldX’s force field before mutations were done. Mutagenesis was carried out using the BuildModel FoldX command, and each mutation was calculated

five times. VARIANT CLASSIFICATION Variants were classified according to the following categories: nonsense, indel, deletion or duplication, heme or ligand interaction, POR interaction,

protein degradation, Meander and the ERR-Triad, or stability (Table S1). For those in the latter group, we proceeded to calculate protein stabilities and correlated them with reported _in

vitro_ activities (see below). Some residues were not classified into the former categories, but were excluded instead due to a poor structure resolution of the corresponding templates.

Residues were considered to be involved in heme or ligand binding when the amino acid is within 5 Angstroms and pointing towards heme/ligand. Residues were considered to interact with POR

according to Robins _et al_.13, or when charged residues (Arginine and Glutamic acid) were found in a close proximity and exposed to the same surface than those reported by Robins _et

al_.13, namely residues R124, E140, E320, R341, R356, R366, R369, R431 and R444. STABILITY CALCULATIONS Protein stabilities were calculated using FoldX’s Stability command, and ∆∆G values

were estimated as the difference between the energy of the wild type protein and the average of five replicas for each point mutation. A threshold of 1, 6 kcal/mol was considered, as it

corresponds to twice the standard deviation calculated with FoldX. We considered values above this threshold to significantly destabilize a protein. Mutations located in functional residues

(e.g. those involved in disulfide bridge or in heme, substrate or ligand binding) were excluded from the stability analysis as their effects on protein activity are influenced by other

variables besides protein stability. To favor wild type conformation, all residues involved in ligand or heme interactions where fixed when optimizing the structures to FoldX force field.

STATISTICAL ANALYSES The predicted free energy of each of the mutants relative to the wild type counterpart (∆∆G) was plotted against the natural logarithm (ln) of the _in vitro_ activity.

Mutants with experimental activities around 0% of the wild type presented destabilization values around 5, 5 kcal/mol. Thus, this value was considered as the maximum one for the fitting,

even when FoldX predicts larger values (mutants L142P and L166P). The predicted activities of those mutants lacking functional assays were derived from the fitting of the abovementioned

correlation. Statistical analyses were performed using the Infostat V.11 (http://www.infostat.com.ar/?lang=en) and GraphPad Prism V. 5.01 Softwares

(http://www.graphpad.com/scientific-software/prism/). Spearman’s correlation coefficient was used to determine a monotonous relation between ∆∆G and the ln of the activity. Permutation test

was applied to evaluate its statistical significance. The parameters of the linear regression were calculated by the least squares method and the statistical significance of the regression

line’s slope by a permutation test. A p < 0.05 was considered significant. DOUBLE MUTANT/SNP EFFECT Only those variants with a reported _in vitro_ activity and classified as putatively

involved in protein stability were analyzed. Double mutants/SNPs _in cis_ were generated as described above. Synergistic effect in those cases was assessed by comparing the estimated ∆∆G of

the double mutant to the sum of both single mutants. When there is no synergy, the difference between them tends to zero. ADDITIONAL INFORMATION HOW TO CITE THIS ARTICLE: Bruque, C. D. _et

al_. Structure-based activity prediction of CYP21A2 stability variants: A survey of available gene variations. _Sci. Rep._ 6, 39082; doi: 10.1038/srep39082 (2016). PUBLISHER'S NOTE:

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. REFERENCES * New, M. I., White, P. C., Pang, S. Dupont, B. &

Speiser, P. W. The Adrenal Hyperplasias in The Metabolic Basis of Inherited Disease (eds Scriver, C. R., Beaudet, A. L., Sly, S. & Valle, D. ) 1881–1917 (McGraw-Hill, 1989). * Miller, W.

L. Clinical review 54: Genetics, diagnosis, and management of 21-hydroxylase deficiency. J. Clin. Endocrinol. Metab. 78, 241–246 (1994). CAS  PubMed  Google Scholar  * Pang, S. & Shook,

M. K. Current status of neonatal screening for congenital adrenal hyperplasia. Curr. Opin. Pediatr. 9, 419–423 (1997). CAS  PubMed  Google Scholar  * Koppens, P. F. et al. Family studies of

the steroid 21-hydroxylase and complement C4 genes define 11 haplotypes in classical congenital adrenal hyperplasia in The Netherlands. Eur. J. Pediatr. 151, 885–892 (1992). CAS  PubMed 

Google Scholar  * Blanchong, C. A. et al. Deficiencies of human complement component C4A and C4B and heterozygosity in length variants of RP-C4-CYP21-TNX (RCCX) modules in caucasians. The

load of RCCX genetic diversity on major histocompatibility complex-associated disease. J. Exp. Med. 191, 2183–2196 (2000). CAS  PubMed  PubMed Central  Google Scholar  * Donohoue, P. A. et

al. Gene conversion in salt-losing congenital adrenal hyperplasia with absent complement C4B protein. J. Clin. Endocrinol. Metab. 62, 995–1002 (1986). CAS  PubMed  Google Scholar  * Higashi,

Y., Tanae, A., Inoue, H. & Fujii-Kuriyama, Y. Evidence for frequent gene conversion in the steroid 21-hydroxylase P-450(C21) gene: implications for steroid 21-hydroxylase deficiency.

Am. J. Hum. Genet. 42, 17–25 (1988). CAS  PubMed  PubMed Central  Google Scholar  * White, P. C. & Speiser, P. W. Congenital adrenal hyperplasia due to 21-hydroxylase deficiency. Endocr.

Rev. 21, 245–291 (2000). CAS  PubMed  Google Scholar  * Nelson, D. R. et al. Comparison of cytochrome P450 (CYP) genes from the mouse and human genomes, including nomenclature

recommendations for genes, pseudogenes and alternative-splice variants. Pharmacogenetics 14, 1–18 (2004). CAS  PubMed  Google Scholar  * Kominami, S., Ochi, H., Kobayashi, Y. & Takemori,

S. Studies on the steroid hydroxylation system in adrenal cortex microsomes. Purification and characterization of cytochrome P-450 specific for steroid C-21 hydroxylation. J. Biol. Chem.

255, 3386–3394 (1980). CAS  PubMed  Google Scholar  * Higashi, Y., Yoshioka, H., Yamane, M., Gotoh, O. & Fujii-Kuriyama, Y. Complete nucleotide sequence of two steroid 21-hydroxylase

genes tandemly arranged in human chromosome: a pseudogene and a genuine gene. Proc. Natl. Acad. Sci. USA 83, 2841–2845 (1986). ADS  CAS  PubMed  PubMed Central  Google Scholar  * White, P.

C., New, M. I. & Dupont, B. Structure of human steroid 21-hydroxylase genes. Proc. Natl. Acad. Sci. USA 83, 5111–5115 (1986). ADS  CAS  PubMed  PubMed Central  Google Scholar  * Robins,

T., Carlsson, J., Sunnerhagen, M., Wedell, A. & Persson, B. Molecular model of human CYP21 based on mammalian CYP2C5: structural features correlate with clinical severity of mutations

causing congenital adrenal hyperplasia. Mol. Endocrinol. 20, 2946–2964 (2006). CAS  PubMed  Google Scholar  * Pey, A. L., Stricher, F., Serrano, L. & Martinez, A. Predicted effects of

missense mutations on native-state stability account for phenotypic outcome in phenylketonuria, a paradigm of misfolding diseases. Am. J. Hum. Genet. 81, 1006–1024 (2007). CAS  PubMed 

PubMed Central  Google Scholar  * Alibés, A. et al. Using protein design algorithms to understand the molecular basis of disease caused by protein-DNA interactions: the Pax6 example. Nucleic

Acids Res. 38, 7422–7431 (2010). PubMed  PubMed Central  Google Scholar  * Minutolo, C. et al. Structure-based analysis of five novel disease-causing mutations in 21-hydroxylase-deficient

patients. PLoS One 6, e15899 (2011). ADS  CAS  PubMed  PubMed Central  Google Scholar  * Worth, C. L., Preissner, R. & Blundell, T. L. SDM–a server for predicting effects of mutations on

protein stability and malfunction. Nucleic Acids Res. 39, W215–222 (2011). CAS  PubMed  PubMed Central  Google Scholar  * Zhao, B. et al. Three-dimensional structure of steroid

21-hydroxylase (cytochrome P450 21A2) with two substrates reveals locations of disease-associated variants. J. Biol. Chem. 287, 10613–10622 (2012). CAS  PubMed  PubMed Central  Google

Scholar  * Pallan, P. S. et al. Human Cytochrome P450 21A2, the Major Steroid 21-Hydroxylase: Structure of the enzyme progesterone substrate complex and rate-limiting C-H bond cleavage. J.

Biol. Chem. 290, 13128–13143 (2015). CAS  PubMed  PubMed Central  Google Scholar  * Haider, S. et al. Structure-phenotype correlations of human CYP21A2 mutations in congenital adrenal

hyperplasia. Proc. Natl. Acad. Sci. USA 110, 2605–2610 (2013). ADS  CAS  PubMed  PubMed Central  Google Scholar  * Brønstad, I. et al. Functional studies of novel CYP21A2 mutations detected

in Norwegian patients with congenital adrenal hyperplasia. Endocr. Connect. 3, 67–74 (2014). PubMed  PubMed Central  Google Scholar  * Massimi, A. et al. Functional and Structural Analysis

of Four Novel Mutations of CYP21A2 Gene in Italian Patients with 21-Hydroxylase Deficiency. Horm. Metab. Res. 46, 515–520 (2014). CAS  PubMed  Google Scholar  * Taboas, M. et al. Functional

studies of p.R132C, p.R149C, p.M283V, p.E431K, and a novel c.652-2A&gt;G mutations of the CYP21A2 gene. PLoS One 9, e92181 (2014). ADS  PubMed  PubMed Central  Google Scholar  * Barbaro,

M. et al. _In vitro_ functional studies of rare CYP21A2 mutations and establishment of an activity gradient for nonclassic mutations improve phenotype predictions in congenital adrenal

hyperplasia. Clin. Endocrinol. (Oxf). 82, 37–44 (2014). Google Scholar  * Rodrigues, N. R. et al. Molecular characterization of the HLA-linked steroid 21-hydroxylase B gene from an

individual with congenital adrenal hyperplasia. EMBO J. 6, 1653–1661 (1987). CAS  PubMed  PubMed Central  Google Scholar  * Tardy, V. Gene symbol: CYP21A2. Disease: steroid 21-hydroxylase

deficiency. Hum. Genet. 119, 363 (2006). PubMed  Google Scholar  * Tardy, V. T. V. & Morel, Y. Gene symbol: CYP21A2. Hum. Genet. 121, 293 (2007). PubMed  Google Scholar  * Tardy, V. T.

V. Gene symbol: CYP21A2. Hum. Genet. 121, 292–293 (2007). PubMed  Google Scholar  * Wang, R. et al. 21-Hydroxylase deficiency-induced congenital adrenal hyperplasia in 230 Chinese patients:

Genotype–phenotype correlation and identification of nine novel mutations. Steroids 108, 47–55 (2016). CAS  PubMed  Google Scholar  * New, M. I. et al. Genotype-phenotype correlation in

1,507 families with congenital adrenal hyperplasia owing to 21-hydroxylase deficiency. Proc. Natl. Acad. Sci. USA 110, 2611–2616 (2013). ADS  CAS  PubMed  PubMed Central  Google Scholar  *

Milacic, I. et al. Molecular genetic study of congenital adrenal hyperplasia in Serbia: novel p.Leu129Pro and p.Ser165Pro CYP21A2 gene mutations. J. Endocrinol. Invest. 38, 1199–1210 (2015).

CAS  PubMed  Google Scholar  * Vrzalová, Z. et al. Identification of CYP21A2 mutant alleles in Czech patients with 21-hydroxylase deficiency. Int. J. Mol. Med. 26, 595–603 (2010). PubMed 

Google Scholar  * Speiser, P. W., New, M. I. & White, P. C. Molecular Genetic Analysis of Nonclassic Steroid 21-Hydroxylase Deficiency Associated with HLA-B14, DR1. N. Engl. J. Med. 319,

19–23 (1988). CAS  PubMed  Google Scholar  * Kirac, D. et al. The Frequency and the Effects of 21-Hydroxylase Gene Defects in Congenital Adrenal Hyperplasia Patients. Ann. Hum. Genet. 78,

399–409 (2014). CAS  PubMed  Google Scholar  * Concolino, P., Mello, E., Zuppi, C. & Capoluongo, E. Molecular diagnosis of congenital adrenal hyperplasia due to 21-hydroxylase

deficiency: an update of new CYP21A2 mutations. Clin. Chem. Lab. Med. 48, 1057–1062 (2010). CAS  PubMed  Google Scholar  * Loke, K. Y., Lee, Y. S., Lee, W. W. & Poh, L. K. Molecular

analysis of CYP-21 mutations for congenital adrenal hyperplasia in Singapore. Horm. Res. 55, 179–84 (2001). CAS  PubMed  Google Scholar  * Ezquieta, B. et al. Non-classical 21-hydroxylase

deficiency in children: association of adrenocorticotropic hormone-stimulated 17-hydroxyprogesterone with the risk of compound heterozygosity with severe mutations. Acta Paediatr. 91,

892–898 (2002). CAS  PubMed  Google Scholar  * Stikkelbroeck, N. M. M. L. et al. CYP21 gene mutation analysis in 198 patients with 21-hydroxylase deficiency in The Netherlands: six novel

mutations and a specific cluster of four mutations. J. Clin. Endocrinol. Metab. 88, 3852–3859 (2003). CAS  PubMed  Google Scholar  * Deneux, C. et al. Phenotype-genotype correlation in 56

women with nonclassical congenital adrenal hyperplasia due to 21-hydroxylase deficiency. J. Clin. Endocrinol. Metab. 86, 207–213 (2001). CAS  PubMed  Google Scholar  * Bojunga, J. et al.

Structural and functional analysis of a novel mutation of CYP21B in a heterozygote carrier of 21-hydroxylase deficiency. Hum. Genet. 117, 558–564 (2005). CAS  PubMed  Google Scholar  *

Wasniewska, M. et al. Novel mutation of CYP21A2 gene (N387K) affecting a non-conserved amino acid residue in exon 9. J Endocrinol Invest 32, 633 (2009). CAS  PubMed  Google Scholar  *

Baradaran-Heravi, A. et al. Three novel CYP21A2 mutations and their protein modelling in patients with classical 21-hydroxylase deficiency from northeastern Iran. Clin. Endocrinol. (Oxf).

67, 335–41 (2007). CAS  Google Scholar  * Jiang, L. et al. Identification and functional characterization of a novel mutation P459H and a rare mutation R483W in the CYP21A2 gene in two

Chinese patients with simple virilizing form of congenital adrenal hyperplasia. J. Endocrinol. Invest. 35, 485–9 (2012). CAS  PubMed  Google Scholar  * Nadra, A. D., Serrano, L. &

Alibés, A. DNA-binding specificity prediction with FoldX. Methods in Enzymology 498, 3–18 (2011). CAS  PubMed  Google Scholar  * Wu, D. A. & Chung, B. Mutations of P45Oc21 (Steroid

21-Hydroxylase) at Cys428, Val281, and Serf” Result in Complete, Partial, or No Loss of Enzymatic Activity, Respectively. J. Clin. Invest. 88, 519–523 (1991). CAS  PubMed  PubMed Central 

Google Scholar  * Ozturk, I. C., Wei, W.-L., Palaniappan, L., Rubenfire, M. & Killeenas, A. A. Analysis of CYP21 Coding Polymorphisms in Three Ethnic Populations: Further Evidence of

Nonamplifying CYP21 Alleles Among Whites. Mol. Diagnosis 5, 47–52 (2000). CAS  Google Scholar  * Asanuma, A. et al. Molecular analysis of Japanese patients with steroid 21-hydroxylase

deficiency. J. Hum. Genet. 44, 312–317 (1999). CAS  PubMed  Google Scholar  * Dolzan, V. et al. Mutational spectrum of steroid 21-hydroxylase and the genotype-phenotype association in Middle

European patients with congenital adrenal hyperplasia. Eur. J. Endocrinol. 153, 99–106 (2005). CAS  PubMed  Google Scholar  * Nikoshkov, A., Lajic, S., Holst, M., Wedell, A. & Luthman,

H. Synergistic effect of partially inactivating mutations in steroid 21-hydroxylase deficiency. J. Clin. Endocrinol. Metab. 82, 194–199 (1997). CAS  PubMed  Google Scholar  * Menassa, R. et

al. p.H62L, a rare mutation of the CYP21 gene identified in two forms of 21-hydroxylase deficiency. J. Clin. Endocrinol. Metab. 93, 1901–1908 (2008). CAS  PubMed  Google Scholar  * Tardy, V.

et al. Phenotype-genotype correlations of 13 rare CYP21A2 mutations detected in 46 patients affected with 21-hydroxylase deficiency and in one carrier. J. Clin. Endocrinol. Metab. 95,

1288–1300 (2010). CAS  PubMed  Google Scholar  * Sherry, S. T. et al. dbSNP: the NCBI database of genetic variation. Nucleic Acids Res. 29, 308–311 (2001). CAS  PubMed  PubMed Central 

Google Scholar  * Sudmant, P. H. et al. An integrated map of structural variation in 2,504 human genomes. Nature 526, 75–81 (2015). CAS  PubMed  PubMed Central  Google Scholar  * Sali, A.

& Blundell, T. L. Comparative protein modelling by satisfaction of spatial restraints. J. Mol. Biol. 234, 779–815 (1993). CAS  PubMed  Google Scholar  * Tamura, K., Dudley, J., Nei, M.

& Kumar, S. MEGA4: Molecular Evolutionary Genetics Analysis (MEGA) software version 4.0. Mol. Biol. Evol. 24, 1596–1599 (2007). CAS  PubMed  Google Scholar  * Shen, M.-Y. & Sali, A.

Statistical potential for assessment and prediction of protein structures. Protein Sci. 15, 2507–2524 (2006). CAS  PubMed  PubMed Central  Google Scholar  * Benkert, P., Biasini, M. &

Schwede, T. Toward the estimation of the absolute quality of individual protein structure models. Bioinformatics 27, 343–350 (2011). CAS  PubMed  Google Scholar  * Ramachandran, G. N.,

Ramakrishnan, C. & Sasisekharan, V. Stereochemistry of polypeptide chain configurations. J. Mol. Biol. 7, 95–99 (1963). CAS  PubMed  Google Scholar  * Schymkowitz, J. et al. The FoldX

web server: an online force field. Nucleic Acids Res. 33, W382–388 (2005). CAS  PubMed  PubMed Central  Google Scholar  * Soardi F. C. et al. Inhibition of CYP21A2 enzyme activity caused by

novel missense mutations identified in Brazilian and Scandinavian patients. J Clin Endocrinol Metab 93, 2416–2420 (2008). CAS  PubMed  Google Scholar  * Speiser P. W. et al. Congenital

adrenal hyperplasia due to steroid 21-hydroxylase deficiency: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab 95, 4133–4160 (2010). CAS  PubMed  PubMed Central 

Google Scholar  Download references ACKNOWLEDGEMENTS We thank Dr. F. Pisciottano, MSc. G. Acevedo, Msc. L. Simonetti and Dr. J. C. Calvo for critical revision of the manuscript. We also

specially thank Dr. Mehrnoosh Arrar for the revision of the English clarity and readability of the manuscript. CDB, CSF, NB, LDE, AS, LA and LD are professional staff from the Administración

Nacional de Laboratorios e Institutos de Salud (ANLIS). ADN and LD are staff researchers from the Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET). MD is a fellow from

the Agencia Nacional de Promoción Científica y Tecnológica (ANPCyT). This research was supported by grants from ANPCyT: PICT 2013-1417; CONICET: PIP 11220130100542; Fondos Concursables ANLIS

(FOCANLIS) 2013, and Universidad Nacional de Buenos Aires: UBACyT 20020110100005BA. AUTHOR INFORMATION Author notes * Present address: Laboratorio de Cultivo Celular y Medicina

Regenerativa, Servicio de Ortopedia y Traumatología, Hospital de Agudos Juan A. Fernández, Buenos Aires, Argentina. * Delea Marisol and Fernández Cecilia S. contributed equally to this work.

AUTHORS AND AFFILIATIONS * Centro Nacional de Genética Médica, ANLIS, Buenos Aires, Argentina Carlos D. Bruque, Marisol Delea, Cecilia S. Fernández, Juan V. Orza, Melisa Taboas, Noemí

Buzzalino, Lucía D. Espeche, Andrea Solari, Liliana Alba & Liliana Dain * Instituto de Biología y Medicina Experimental, CONICET, Buenos Aires, Argentina Carlos D. Bruque & Liliana

Dain * Consultorio y Laboratorio de Genética, Rosario, Argentina Verónica Luccerini * Departamento de Química Biológica Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires,

IQUIBICEN-CONICET, Buenos Aires, Argentina Alejandro D. Nadra Authors * Carlos D. Bruque View author publications You can also search for this author inPubMed Google Scholar * Marisol Delea

View author publications You can also search for this author inPubMed Google Scholar * Cecilia S. Fernández View author publications You can also search for this author inPubMed Google

Scholar * Juan V. Orza View author publications You can also search for this author inPubMed Google Scholar * Melisa Taboas View author publications You can also search for this author

inPubMed Google Scholar * Noemí Buzzalino View author publications You can also search for this author inPubMed Google Scholar * Lucía D. Espeche View author publications You can also search

for this author inPubMed Google Scholar * Andrea Solari View author publications You can also search for this author inPubMed Google Scholar * Verónica Luccerini View author publications

You can also search for this author inPubMed Google Scholar * Liliana Alba View author publications You can also search for this author inPubMed Google Scholar * Alejandro D. Nadra View

author publications You can also search for this author inPubMed Google Scholar * Liliana Dain View author publications You can also search for this author inPubMed Google Scholar

CONTRIBUTIONS Conceived and designed the experiments: A.D.N. and L.D. Survey of reported mutations, molecular modeling, structure-based analyses, and stability calculations: C.D.B., M.D. and

C.S.F. Statistical analyses: J.V.O. _CYP21A2_ genotyping and MLPA analyses: C.S.F., M.T., N.B., and L.D.E. Analyzed the data: C.D.B., A.D.N. and L.D. Contributed reagents/materials/analysis

tools: A.D.N., C.S.F. and L.D. Wrote the paper: C.D.B., A.D.N. and L.D. All authors reviewed the manuscript. ETHICS DECLARATIONS COMPETING INTERESTS The authors declare no competing

financial interests. ELECTRONIC SUPPLEMENTARY MATERIAL SUPPLEMENTARY INFORMATION RIGHTS AND PERMISSIONS This work is licensed under a Creative Commons Attribution 4.0 International License.

The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not

included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit

http://creativecommons.org/licenses/by/4.0/ Reprints and permissions ABOUT THIS ARTICLE CITE THIS ARTICLE Bruque, C., Delea, M., Fernández, C. _et al._ Structure-based activity prediction of

CYP21A2 stability variants: A survey of available gene variations. _Sci Rep_ 6, 39082 (2016). https://doi.org/10.1038/srep39082 Download citation * Received: 24 July 2016 * Accepted: 16

November 2016 * Published: 14 December 2016 * DOI: https://doi.org/10.1038/srep39082 SHARE THIS ARTICLE Anyone you share the following link with will be able to read this content: Get

shareable link Sorry, a shareable link is not currently available for this article. Copy to clipboard Provided by the Springer Nature SharedIt content-sharing initiative