An Updated PAH Mutational Spectrum of Phenylketonuria in Mexican Patients Attending a Single Center: Biochemical, Clinical-Genotyping Correlations

Establishing the genotypes of patients with hyperphenylalaninemia (HPA)/phenylketonuria (PKU, MIM#261600) has been considered a cornerstone for rational medical management. However, knowledge of the phenylalanine hydroxylase gene (PAH) mutational spectrum in Latin American populations is still limited. Herein, we aim to update the mutational PAH spectrum in the largest cohort of HPA/PKU Mexican patients (N = 124) reported to date. The biallelic PAH genotype was investigated by Sanger automated sequencing, and genotypes were correlated with documented biochemical phenotypes and theoretical tetrahydrobiopterin (BH4) responsiveness. Patients were biochemically classified as having classic PKU (50%, 62/124), mild PKU (20.2%, 25/124) and mild HPA (29.8%, 37/124). Furthermore, 78.2% of the included patients (97/124) were identified by newborn screening. A total of 60 different pathogenic variants were identified, including three novel ones (c. 23del, c. 625_626insC and c. 1315 + 5_1315 + 6insGTGTAACAG), the main categories being missense changes (58%, 35/60) and those affecting the catalytic domain (56.6%, 34/60), and c. 60 + 5G > T was the most frequent variant (14.5%, 36/248) mainly restricted (69.2%) to patients from the central and western parts of Mexico. These 60 types of variants constituted 100 different biallelic PAH genotypes, with the predominance of compound-heterozygous ones (96/124, 77%). The expected BH4 responsiveness based on the PAH genotype was estimated in 52% of patients (65/124), mainly due to the p. (Val388Met) (rs62516101) allele. Instead, our study identified 27 null variants with an allelic phenotype value of zero, with a predominance of c. 60 + 5G > T, which predicts the absence of BH4 responsiveness. An identical genotype reported in BIOPKUdb was found in 92/124 (74%) of our patients, leading to a genotype–phenotype concordance in 80/92 (86.9%) of them. The high number of variants found confirms the heterogeneous and complex mutational landscape of HPA/PKU in Mexico.

. Workflow scheme for inclusion in the present study of 124 patients bearing PAH biallelic genotypes. The biochemical characteristics of the four identified patients bearing monoallelic PAH genotypes are shown. The 14 patients with normal PAH genotypes are currently under study for BH4 defects. Abbreviations: HPA: hyperphenylalaninemia; PAH: phenylalanine hydroxylase gene; Phe: phenylalanine; Tyr: tyrosine; mPKU: mild phenylketonuria; MHP: mild hyperphenylalaninemia.

Amino Acid Quantification
The quantification of Phe and Tyr blood concentrations was performed by high-performance liquid chromatography (HPLC) according to a previously described methodology by Qureshi et al. [10] or by tandem mass spectrometry (MS/MS). The determination of blood amino acid profiles by tandem mass spectrometry was performed using a commercial kit from Perkin Elmer (Neobase kit, Wallac Oy, Turku, Finland), according to the manufacturer's instructions.

Genotype Analysis
Genomic DNA samples were obtained from dried blood spots by the saline precipitation method (Gentra Puregene Blood Kit, Gentra Systems, Minneapolis, MN, USA). Polymerase chain reaction (PCR) amplification and direct automated Sanger sequencing were applied to the thirteen PAH exons (NG_008690.2 RefSeqGene, NM_000277.3) in addition to their exon-intron borders. PCR oligonucleotides and sequencing conditions are presented in Supplementary Materials Appendix A and Table S1. Identified variants were classified according to BIOPKUdb and/or the literature. Novel variants were classified according to the scoring proposed by the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP) [11]. Due to limited available information regarding their functional or clinical relevance, these novel and missense changes were subjected to in silico analysis using the PolyPhen (http://genetics.bwh.harvard.edu/pph2, accessed on 20 December 2020), PROVEAN (http://provean.jcvi.org/index.php, accessed on 20 December 2020), and Pmut (http://mmb.irbbarcelona.org/PMut, accessed on 20 December 2020) programs.
Novel intronic variants were assessed by the splicing module integrated in Alamut ® Visual version 2.15 software (SOPHiA GENETICS, Lausanne, Switzerland), which include SpliceSiteFinder-like, MaxEntScan, GeneSplicer and NNSPLICE in silico evaluations for

Amino Acid Quantification
The quantification of Phe and Tyr blood concentrations was performed by highperformance liquid chromatography (HPLC) according to a previously described methodology by Qureshi et al. [10] or by tandem mass spectrometry (MS/MS). The determination of blood amino acid profiles by tandem mass spectrometry was performed using a commercial kit from Perkin Elmer (Neobase kit, Wallac Oy, Turku, Finland), according to the manufacturer's instructions.

Genotype Analysis
Genomic DNA samples were obtained from dried blood spots by the saline precipitation method (Gentra Puregene Blood Kit, Gentra Systems, Minneapolis, MN, USA). Polymerase chain reaction (PCR) amplification and direct automated Sanger sequencing were applied to the thirteen PAH exons (NG_008690.2 RefSeqGene, NM_000277.3) in addition to their exon-intron borders. PCR oligonucleotides and sequencing conditions are presented in Supplementary Materials Table S1 and Appendix A. Identified variants were classified according to BIOPKUdb and/or the literature. Novel variants were classified according to the scoring proposed by the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP) [11]. Due to limited available information regarding their functional or clinical relevance, these novel and missense changes were subjected to in silico analysis using the PolyPhen (http://genetics.bwh.harvard.edu/ pph2, accessed on 20 December 2020), PROVEAN (http://provean.jcvi.org/index.php, accessed on 20 December 2020), and Pmut (http://mmb.irbbarcelona.org/PMut, accessed on 20 December 2020) programs.
Novel intronic variants were assessed by the splicing module integrated in Alamut ® Visual version 2.15 software (SOPHiA GENETICS, Lausanne, Switzerland), which include SpliceSiteFinder-like, MaxEntScan, GeneSplicer and NNSPLICE in silico evaluations for donor and acceptor splice sites, as well as ESEFinder, RESCUE-ESE and EX-SKIP, to predict potentially deleterious effects on Exonic Splicing Enhancer (ESE) binding sites.

Protein Modeling and Mutagenesis in Silico
The crystallographic structure of human PAH with and without ligands (Protein Data Bank (PDB) codes: 1KW0 and 2PAH, respectively) was used to localize and analyze the potential pathogenic effect of the recently reported p. (His264Arg) variant (BIOPKUdb) by mutagenesis in silico. Pymol software, version 2.3.5, was used for protein analyses and figure construction [12]. 1KW0 crystal was obtained in the presence of BH 4 cofactor and the substrate analogue 3-(2-thienyl)-L-alanine (THA) [13].

Genotype-Phenotype Correlation with GPV
Our observed biochemical phenotype (cPKU, mPKU and MHP) was compared with the Genotype Phenotype Value (GPV) calculated from Allelic Phenotype Value (APV) [14] reported in BIOPKUdb. A genotype-phenotype correlation was considered concordant when the obtained GPV corresponded with the reported cut-off.

Genotype-Phenotype Correlation with Identical Genotypes
In the case in which identical genotypes were available in the BIOPKUdb, their associated phenotypes were compared with those found in the patients of the present work. For new variants not reported in BIOPKUdb, the category "not reported" or "still undetermined" was used.

Theoretical BH 4 Responsiveness and Recommendation to Test
BIOPKUdb was used to analyze the BH 4 responsiveness related to the genotypes found in the present work. To that end, identical genotypes were investigated. In cases in which genotypes or their BH 4 responsiveness had not yet been reported in BIOPKUdb, the analysis was performed based on each allele in homozygous state. Since it is well known that genotypes with a GPV equal to zero are non-BH 4 responders [15,16], genotypes with these characteristics were excluded from the analysis. If the patient contained any allele that had been classified as a responder or a slow responder in BIOPKUdb, the recommendation was to test for BH 4 responsiveness. This same recommendation to test was also made for all the genotypes that had not yet been reported in BIOPKUdb.

Statistical Analysis
Descriptive statistics (average, minimum and maximum) were determined for clinical data of patients. Allelic frequencies were calculated as the number of alleles of each variant per 100/total number of studied alleles. Comparisons of frequencies were performed using a Chi-squared test with MedCalc software version 20.010 (MedCalc Software Ltd., Ostend, Belgium) considering a p value < 0.05 as significant.

Genotype/Phenotype Concordance with GPV
Concordance with the GPV was established in 118/124 cases (95.2%). In 6/124 (4.8%) cases, concordance could not be established as the allele was reported as undetermined in the BIOPKUdb. From the remaining 118 cases with information available, 85 were concordant (72%) and 33/118 cases were discordant (28%).

Genotype/Phenotype Concordance with Identical Genotypes Reported in BIOPKUdb
An identical genotype reported in BIOPKUdb was found in 92 out of 124 patients (74%). From those 92 subjects, 80 (86.9%) were concordant with the reported phenotype, and in 12 cases, the phenotype was discordant. The disease severity was under-or overestimated in eight and four cases, respectively ( Table 2).

Clinical Description of Patients Bearing PAH Genotypes Containing c. 60 + 5G > T Variant
This pathogenic variant alters the consensus sequence of the 5' donor splice site of intron 1 of the PAH gene [17], possibly leading to the abolition of its proper recognition by splicing machinery. A total of 26 patients carried the c. 60 + 6G > T variant, either in homozygous (10/26) or in compound heterozygous (16/26) state (Table 3), and 18/26 (69.2%) of them came from the Mexican Middle West region called El Bajío (Jalisco, Guanajuato, Michoacán and Aguascalientes States). Sixty five percent of these patients (17/26) were detected early by NBS, while the remaining patients belonged to the CD group. Brain NMRI imaging of two c. 60 + 5G > T-homozygous patients from the CD group demonstrated damage in the basal ganglia, white matter, and cerebral atrophy ( Figure 4). The first patient was a one-year-old male who was detected at 9 months old; his Phe diagnostic value was 1654 µM, he was clinically diagnosed with autism and severe neurodevelopmental delay, and he had poor outcome after nutritional therapy. The other patient was a female who had a diagnostic Phe value of 1830 µM, at 10 years old, with severe neurodevelopmental delay and behavioral issues, such as auto-aggression and incontrollable tantrums. She was misdiagnosed with autism. After PKU diagnosis, nutritional therapy was started with good compliance and biochemical Phe controls; however, only a few clinical benefits were observed after treatment, such as a decrease in auto-aggression and tantrums.    Novel pathogenic PAH variants are highlighted in red bold type. GPV is equal to the higher APV of the two alleles (APV max). The GPV cut-off values are defined as cPKU (0-2.7), mPKU (2.8-6.6) and MHP (6.7-10.0). Abbreviations: CD, clinical diagnosis; cPKU, classic phenylketonuria; F, female; M, male; MHP, mild hyperphenylalaninemia; mPKU, mild phenylketonuria; NBS, newborn screening; NR, not reported or still undetermined. Symbols represent genotype/phenotype discordance, ∆ underestimation of disease severity, and • overestimation of disease severity compared with BIOPKUdb. These patients were described in our previous report [6].

Novel PAH Variants and Description of Resulting Phenotypes
The frameshift c. 23del or p. (Asn8Thrfs * 30) variant was found in a patient categorized as cPKU and bearing the p. (Val388Met) in the second allele (Patient 46, Table 2).
Clinically, this patient showed a maximum blood Phe concentration of 2798 μmol/L, a Tyr blood concentration of 133.29 μmol/L, a Phe/Tyr ratio of 20.9 and a dietary Phe tolerance of 215 mg/day. This patient began nutritional treatment at 1.5 months old, with consistent

Novel PAH Variants and Description of Resulting Phenotypes
The frameshift c. 23del or p. (Asn8Thrfs * 30) variant was found in a patient categorized as cPKU and bearing the p. (Val388Met) in the second allele (Patient 46, Table 2). Clinically, this patient showed a maximum blood Phe concentration of 2798 µmol/L, a Tyr blood concentration of 133.29 µmol/L, a Phe/Tyr ratio of 20.9 and a dietary Phe tolerance of 215 mg/day. This patient began nutritional treatment at 1.5 months old, with consistent adherence to treatment during 4 years of follow up, with most of her Phe values within the therapeutical range (<360 µmol/L) and neurodevelopmental skills according to her age.
The c. 1315 + 5_1315 + 6insGTGTAACAG variant was identified in compound heterozygosity involving the p. The homozygous c.
[625_626insC]; [625_626insC] genotype for a novel frameshift insertion was identified in a cPKU patient (Patient 72, Table 2), showing a Phe blood concentration of 1217 µmol/L, Tyr of 58 µmol/L, a Phe/Tyr ratio of 20.9 and a dietary Phe tolerance of 319 mg/day. As this patient was adopted at an early age, we did not have his family history (i.e., parental consanguinity). Early instauration of dietary management at the age of 25 days allowed him to maintain most of his blood Phe values within the therapeutic range and normal neurodevelopmental skills according to his age, in a 6.5-year follow-up period.

Protein in Silico Modeling of the p. (His264Arg) Variant
The c. 791A > G or p. (His264Arg) has been recently reported in the BIOPKUdb, although its allele frequency is still not known, results of functional assays are not available, and its resulting biochemical phenotype is still unknown. In fact, in the ClinVar database, it is considered as a variant of unknown significance (RCV001224584.2). Herein, it was identified in two unrelated patients (ID 64 and 88, Table 2). The PolyPhen, PROVEAN, and Pmut programs unanimously predicted the p. (His264Arg) variant as deleterious with high confidence scores. Due to the rarity of this variant, we evaluated in silico the potential disturbance at the tridimensional structural arrangement of the resulting protein. We found that the substitution of histidine by the positively charged amino acid arginine predicts a serious disturbing effect within the vicinity of the catalytic site in all the possible rotamers of arginine ( Figure 5).

Theoretical BH 4 Responsiveness
We found that there were 52% (65/124) of patients whose genotypes included at least one potential responder allele, or one that has still not been reported in BIOPKUdb ( Table 4). The MHP was the predominant biochemical phenotype of these patients (52.3%, 34/65), but cPKU accounted for 27.7% (18/65), and the remainder accounted for mPKU (20%, 13/65). The most frequent responder allele was c. 1162G > A or p. (Val388Met) (14 alleles), followed by c. 782G > A or p. (Arg261Gln) (seven alleles) and c. 722G > A or p. (Arg241His) and c. 1169A > G or p. (Gln390Gly) (six alleles each). identified in two unrelated patients (ID 64 and 88, Table 2). The PolyPhen, PROVEAN, and Pmut programs unanimously predicted the p. (His264Arg) variant as deleterious with high confidence scores. Due to the rarity of this variant, we evaluated in silico the potential disturbance at the tridimensional structural arrangement of the resulting protein. We found that the substitution of histidine by the positively charged amino acid arginine predicts a serious disturbing effect within the vicinity of the catalytic site in all the possible rotamers of arginine ( Figure 5).

Discussion
The epidemiology of PKU has been the objective of a growing number of publications [2,3,18]; however, the main findings of those investigations are based on European and Asiatic populations. The Latin American population is poorly represented, mainly due to the scarce reliable published data from this region. In a recent PKU work, genetic and epidemiological data from 64 countries were presented [2]; however, only three Latin American countries were considered (Argentina, Brazil and Mexico [6,19,20]), so there would be a possible underrepresentation of the mutational PAH spectrum in Latin American HPA/PKU patients. The present study aimed to update the PAH mutational spectrum of Mexican HPA/PKU patients attending a single reference center. Moreover, to the best of our knowledge, this is the largest published cohort of PKU patients from Mexico, including individuals of almost all the geographic regions of the country (patients came from 30/32 of the states that constitute the country).
In this new study, patients' phenotypes were classified according to the criteria established by Hillert et al. and van Spronsen et al. [2,3], which will facilitate international comparisons both with BIOPKUdb and other publications. By comparing the frequencies of the different PAH-related biochemical phenotypes identified herein with those previously reported [6], no statistical differences were found.
The number of patients detected by NBS in our center increased from 54.1% in 2015 [6] to 78.2%. This significant increment (p = 0.001) could be related to the optimization of the NBS system in Mexico, although it still does not reach the optimal goal of having a total NBS coverage for this disease [21]. This finding agrees with the fact that improvements in the NBS programs increase the detection of MHP forms [22,23]. Even though it is widely accepted that MHP patients are only treated when Phe blood levels are >360 µmol/L, the biochemical follow up of these patients should be periodical, in order to assure neuro-logically safe Phe values [24]. Moreover, in female patients at a reproductive stage, a close follow up is especially relevant considering the potential risk of their offspring suffering from maternal PKU syndrome [2].
As reported has been worldwide [2,3], the molecular PKU spectrum found in the present study is heterogeneous, and variants are distributed along the entire gene with clustering in exons that codify the catalytic PAH region (35%), especially exons 7 and 6 ( Figure 3). Notably, we did not find any patient with the "Celtic" c. 1222C > T or p. (Arg408Trp) pathogenic variant, which is considered the most frequent one worldwide, especially in central and eastern Europe (44.4-53.7%); however, its allele frequency in the Spanish population decreased significatively (4%), which could correlate with its absence in the Mexican population, which is characterized by bearing a Spanish-derived ancestry [3]. On the other hand, the absence of variants in exon 13 ( Figure 3) was similar to that observed in Spain and other Latin American populations [19,[25][26][27][28]. This fact is in agreement with the low number of pathogenic variants registered in exon 13 in BIOPKUdb.
The Sanger methodology allowed us to obtain a detection rate of 87.3% (124/142) of the biallelic PAH genotypes in the patients initially classified as HPA/PKU; however, this methodology identified that there were 2.81% (4/142) of patients carrying only one pathogenic allele (Figure 1). To date, there are no available data to estimate the proportion of large rearrangements unnoticed for Sanger sequencing methodology, which are expected in less than 1% of PAH-related HPA patients worldwide [3]. At least in Russian patients, gross exonic deletions (mainly involving exon 3 and 5) identified by Multiplex Ligationdependent Probe Amplification (MLPA) analysis account for 0.39% of pathogenic PAH alleles [29]. Thus, in these four patients with monoallelic genotypes-who presented persistent HPA, three of which required nutritional treatment-further studies, such as MLPA analysis, could be an option in order to discard gross PAH rearrangements.
The remaining 14 cases (9.8%) with normal PAH genotypes lacked the recommended former dihydropteridine reductase (DHPR) and pterin evaluations [2] to discard a defect in the metabolism of BH 4 .
Our study corroborates the high frequency of the c. 60 + 5G > T variant (14.5%, Table 1) and its geographical predominance in the western and central regions of Mexico (Table 2), probably due to the founder effect [6]; this has been corroborated by other authors [9]. This intronic null variant, formerly known as "IVS1 + 5g > t", was first described by Guldberg et al. in 1993 [30]. It is a rare variant reported in only 0.7% and 1.36% of Danish [30] and Spanish [31] populations, respectively. In BIOPKUdb, it is reported in 0.32% of the subjects. Due to its rarity, it has been scarcely studied. Interestingly, this variant has not been found in PKU patients from Argentina, Chile or Cuba [19,27,32], but it has been reported in Costa Rica [33], and Brazil [20,25,34,35]. Moreover, the c. 60 + 5G > T variant has not been reported in Russia [36], Japan [23], or China [37]. Therefore, to the best of our knowledge, the studied Mexican population has the highest frequency of this variant.
In the present study, all the patients harboring a homozygous c. 60 + 5G > T genotype had cPKU, which can be explained by the severity of this null variant (APV value = 0), which can theoretically lead to no enzymatic activity, in addition to an absence of BH 4 response [15]. As expected, among patients harboring c. 60 + 5G > T in compound heterozygosity involving a less severe allele with APV > 5 [14], the disease was less severe, corresponding to an MHP phenotype (patients 23 and 24, Table 3), confirming that functionally mild variants with a substantial residual PAH enzymatic activity dominate over null alleles [3]. The clinical picture of patients with homozygous genotypes c. [60 + 5G > T]; [60 + 5G > T] coincides with that described for cPKU patients, consisting of early, severe and progressive neurological manifestations, and symptomatology worsening if the start of dietary treatment is delayed. Therefore, these findings emphasize the importance of starting treatment in the first days of life, preferably in the first week [8], to prevent brain damage that leads to neurodevelopmental delay and permanent intellectual disability [38]. At least in two c. 60 + 5G > T-homozygous and lately diagnosed patients (CD group), we documented central nervous system sequelae (Figure 4), consisting of basal ganglia and white matter brain damage. Remarkably, although the male patient was diagnosed at a younger age, he showed the involvement of basal ganglia, but it is known that imaging studies of PKU patients do not always correlate with the severity of the observed phenotype [39].
The second most frequent variant of this work was c. 1162G > A p. (Val388Met) (11.2%). This variant is common in the Spanish population (6.8%) [26], as well as in other Latin American countries such as Chile (17.2%) [27], Argentina (3.9%) [19] and Brazil (12.7-21.2%) [25,28]. Interestingly, our five homozygous p. (Val388Met) patients had different biochemical phenotypes (mPKU: N = 4 and cPKU: N = 1, Table 2). The predominance of mPKU phenotypes observed herein associated with homozygous p. (Val388Met) contrasts with that reported in BIOPKUdb, where half of the patients show cPKU. Furthermore, this variant was the most frequently found in the discordant heterozygous genotypes (4/12) and in the potential BH 4 -responder genotypes (13/60, Table 4). Its predominance among patients with an inconsistent phenotype could be related to diverse proposed phenomena, such as its destabilizing PAH nature; differences in protein folding; or other, less known factors, such as the influence of modifier genes or epigenetic modifications [26].
The c. 441 + 5G > T or p. (?) intronic and severe variant (APV = 0), was found in 5.64% of the studied alleles, which contrasts with its low frequency worldwide (0.98%). However, higher allelic frequencies have been reported for other Latin American countries, such as Argentina (2.4%) [19] and Chile (3.0%) [27]. Severe cPKU is expected for those homozygous individuals bearing this null allele. Remarkably, in the present study, 50% of the patients with this variant, including two homozygous patients, are originally from the north of Mexico (Baja California, Chihuahua, Coahuila and Tamaulipas). This likely suggests a possible founder effect similar to that described for c. 60 + 5G > T [6]; therefore, further research is needed to support this hypothesis.
The c. 1066-11G > A or p. (Gln355_Tyr356insGlyLeuGln) variant has an APV = 0 leading to the cPKU phenotype. It was identified in 5.6% of the alleles (Table 1), which is similar to that reported in BIOPKUdb (6.8%), Argentina (9.3%), Brazil (9.4%) [3] and Chile (12.7%) [27]. This variant was previously proposed as being responsible for the high PKU cases from Jalisco State [40]; however, our work does not support this previous asseveration, since we did not observe any geographical predominance for patients carrying this variant.
Regarding the novel c. 1315 + 5_1315 + 6insGTGTAACAG variant, it is currently classified as likely pathogenic, as it meets the PM2, PM3, PP3 and PP4 ACMG/AMP criteria [11]. The paternal origin of this variant was confirmed in cPKU Patient #88, who has an extremely low dietary Phe tolerance (255 mg/day) associated with his compound heterozygous genotype involving c. 791A > G or p. (His264Arg) of maternal origin (Table 2). This last missense variant was recently posted in BIOPKUdb, and we predicted it to be a pathogenic change (Figure 5), supported by PM1, PM2, PM5, PP3 and PP4 ACMG/AMP criteria [11].
The novel c. 625_626insC variant was found in homozygous state in patient #77 affected by cPKU (Table 2), in addition to low dietary Phe tolerance (319 mg/day), but no familial antecedents were available as the patient was adopted. This frameshift variant was classified as pathogenic according to the ACMG/AMP (American College of Medical Genetics / Association for Molecular Pathology) criteria (very strong pathogenic variant (PVS1), pathogenic moderate (PM1, PM2) and PP4 ACMG/AMP) [11].
The third novel c. 23del or p. (Asn8Thrfs * 30) frameshift variant was also classified as pathogenic, as it meets the PVS1, PM1, PM2 and PP4 ACMG/AMP criteria [11]. This variant was in heterozygous state with c. 1162G > A or p. (Val388Met) for patient #46 (Table 2), who was classified as a cPKU with low Phe dietary tolerance (215 mg/day). To our knowledge, this variant has never been documented; however, Wang et al., 2018, reported a similar c. 23_24delinsT or p. (Asn8Ilefs * 30) variant (PAH1043 accession number BIOPKUdb) [41]. Although it is a different variant, the resulting frameshift effect at the regulatory domain is the same. The occurrence of these two indels at the same position could suggest a "hot spot" for small rearrangements in c. 23 and c. 24 positions.
In the present work, the observed general genotype/phenotype concordance was 83.8% when compared with identical genotypes reported in the BIOPKUdb. These data are slightly higher than previously reported by our group (70.8%) [6]. This higher concordance is likely due to the new classification criteria used, although other unknown epigenetic or environmental regulatory factors could be involved. One limitation for the frequency of variants and concordance comparison is that not all patients from different populations are listed in the BIOPKUdb there are few identical genotypes, especially those with rare variants, hence the importance of contributing to this database with the information of the PAH mutational spectrum from diverse world regions. Interestingly, in this study, discordance consisted mainly of the underestimation of the disease severity (8/12, 66.7%), as we classified patients with a less severe disease type. This could highlight the difficulties of establishing the correct phenotype based only on blood Phe levels and the importance of establishing the genotype in all HPA/PKU patients [42].
Finally, we found that 52% of the patients were considered potential BH 4 responders ( Table 4). This is similar to the worldwide responsiveness reported by Hillert et al., who found 43% responsiveness from a cohort of 5597 subjects [3]. The most frequent potential BH 4 responder variants are of the missense type, with residual enzymatic activity that could be enhanced with pharmacological doses of BH 4 [43]. On the other hand, we emphasize that patients carrying two non-responder alleles in trans should not be submitted to the BH 4 response test, since those variants (27/60, 45%) do not allow the synthesis of a functional protein; thus, the recommendation for those patients is to maintain a strictly nutritional treatment [15,16].

Conclusions
In summary, the high number of variants found, including the three new variants, confirms the heterogeneous and complex mutational landscape of PKU in Mexico. Half of the studied Mexican patients had cPKU, with a predominance of the c. 60 + 5G > T variant (14.5% of the alleles), which could indicate the highest allelic frequency reported for this variant worldwide. Furthermore, we highlight the severe central nervous system sequelae in two late-diagnosed symptomatic c. 60 + 5G > T-homozygous patients which points out the importance of an early instauration of dietary treatment. A high genotype-phenotype concordance was observed (>80%). A possible founder effect of the c. 441 + 5G > T variant in the north of Mexico warrants further research. Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/genes12111676/s1, Table S1: PCR oligonucleotides and sequencing conditions for PAH gene.