Genotype-phenotype analysis in congenital adrenal hyperplasia due to P450 oxidoreductase deficiency.

CONTEXT: P450 oxidoreductase deficiency (PORD) is a unique congenital adrenal hyperplasia variant that manifests with glucocorticoid deficiency, disordered sex development (DSD), and skeletal malformations. No comprehensive data on genotype-phenotype correlations in Caucasian patients are available.
OBJECTIVE: The objective of the study was to establish genotype-phenotype correlations in a large PORD cohort.
DESIGN: The design of the study was the clinical, biochemical, and genetic assessment including multiplex ligation-dependent probe amplification (MLPA) in 30 PORD patients from 11 countries.
RESULTS: We identified 23 P450 oxidoreductase (POR) mutations (14 novel) including an exonic deletion and a partial duplication detected by MLPA. Only 22% of unrelated patients carried homozygous POR mutations. p.A287P was the most common mutation (43% of unrelated alleles); no other hot spot was identified. Urinary steroid profiling showed characteristic PORD metabolomes with variable impairment of 17α-hydroxylase and 21-hydroxylase. Short cosyntropin testing revealed adrenal insufficiency in 89%. DSD was present in 15 of 18 46,XX and seven of 12 46,XY individuals. Homozygosity for p.A287P was invariably associated with 46,XX DSD but normal genitalia in 46,XY individuals. The majority of patients with mild to moderate skeletal malformations, assessed by a novel scoring system, were compound heterozygous for missense mutations, whereas nearly all patients with severe malformations carried a major loss-of-function defect on one of the affected alleles.
CONCLUSIONS: We report clinical, biochemical, and genetic findings in a large PORD cohort and show that MLPA is a useful addition to POR mutation analysis. Homozygosity for the most frequent mutation in Caucasians, p.A287P, allows for prediction of genital phenotype and moderate malformations. Adrenal insufficiency is frequent, easily overlooked, but readily detected by cosyntropin testing.

Congenital adrenal hyperplasia (CAH) is commonly caused by mutations in genes encoding steroidogenic enzymes (1, 2). By contrast, the CAH variant P450 oxidoreductase (POR) deficiency (PORD) is due to mutations affecting POR that serves as mandatory electron donor enzyme to all microsomal cytochrome P450 (CYP) enzymes (3–7). Thus, in PORD, deficient steroidogenesis is caused by indirect impairment of key enzymes involved in glucocorticoid and sex steroid synthesis including 17α-hydroxylase (CYP17A1), 21-hydroxylase (CYP21A2), and also P450 aromatase (CYP19A1). Clinical manifestations of PORD include adrenal insufficiency and neonatal presentation with disordered sex development (DSD). Of note, DSD can occur in both sexes. Male undervirilization can be explained by impaired sex steroid synthesis, whereas 46,XX DSD despite low circulating androgens has been suggested to be explained by an alternative backdoor pathway to androgens active in fetal life, thereby leading to prenatal virilization of affected female PORD neonates (4).

PORD patients may present with a complex malformation phenotype resembling that described for Antley-Bixler syndrome (ABS; online inheritance in man no. 207410). However, an ABS phenotype can also be caused by FGFR2 mutations (7) and furthermore has been observed in patients without evidence for FGFR2 or POR mutations, suggesting hitherto unidentified genetic defects as underlying cause for the ABS presentation. The skeletal malformations observed in many but not all patients with PORD are thought to be due to disruption of enzymes involved in sterol synthesis, namely 14α-lanosterol demethylase (CYP51A1) and squalene epoxidase, and disruption of retinoic acid metabolism catalyzed by CYP26 isozymes (8) that also depend on electron transfer from POR (9, 10). Furthermore, in vitro activity of microsomal CYP enzymes involved in drug metabolism can be impaired by mutant POR (11). In vivo impairment, particularly of CYP3A4, has recently been demonstrated in a patient with PORD (12), which has important implications for steroid replacement and drug therapy in affected patients.

The clinical characteristics and corresponding genotypes have been described in a number of PORD patients (3, 4, 6, 7, 13–18). The mutation p.A287P is the most frequently reported mutation in patients of Caucasian origin (3, 4, 7), whereas p.R457H is most commonly found in the Japanese population (7, 14, 16). No PORD patient carrying null mutations on both alleles has been described so far, suggesting incompatibility of such genotype with postnatal life, similar to the observation of early fetal death in the murine por deletion model (19, 20).

Of note, the clinical presentation of PORD shows broad phenotypic variability, with adrenal insufficiency, DSD, and skeletal malformations present in some patients but not in others (3, 7, 15, 17). This calls for genotype-phenotype studies to aid in the prediction of disease manifestation and severity. However, apart from a large Japanese series of patients mainly carrying the p.R457H mutation in homozygous or compound heterozygous state (14), no report on a clinically well-characterized PORD patient cohort has addressed this important clinical issue systematically. Here we have analyzed a large mixed Caucasian PORD patient cohort, aiming to provide a comprehensive overview of the association of genotype with clinical and biochemical patient characteristics.

Patients, Materials, and Methods

Patients

The cohort consisted of 30 patients (27 unrelated) from 11 countries. Most patients were Caucasians. However, two patients were of Pakistani origin, one was Hispanic, and two had mixed Chinese/Hispanic and African/Caucasian backgrounds, respectively. Clinical assessment and serum biochemical analyses were performed by the local physicians.

Molecular genetic analysis

Genetic analysis of the POR gene was carried out after obtaining informed consent according to local institutional review board guidelines. DNA was extracted from peripheral blood leukocytes following a standard procedure. The coding sequence of the POR gene including exon-intron boundaries was amplified in 13 PCR fragments as previously described (4). Direct sequencing was carried out using an automated ABI3730 sequencer (Applied Biosystems Inc., Foster City, CA). Sequences were analyzed using the DNAStar Lasergene software package (DNASTAR Inc., Madison, WI). Sequence variants were designated according to Human Genome Variation Society recommendations (www.hgvs.org/rec.html) using the reference sequences GenBank NC_000007 (g.DNA), GenBank NM_000941.2 (c.DNA), and GenBank NP_000932.3 (protein). For genomic DNA and c.DNA numbering, the nucleotide designated +1 was the A of the ATG start codon.

We used multiplex ligation-dependent probe amplification (MLPA) for the identification of partial or complete POR gene deletions or duplications. The MLPA assay was designed by MRC Holland (Amsterdam, The Netherlands) and comprises 16 probes each specifically detecting deletions of the 5′ untranslated exon (U1) and 14 of the 15 coding exons of the POR gene. The assay does not contain a probe for exon 11 because no stable signal was achieved. MLPA was performed using standard reaction conditions according to the manufacturer's protocol. Fragment analysis was conducted on an ABI 3130 Sequencer (Applied Biosystems) using the GeneMarker 1.7 software (SoftGenetics LLC, State College, PA).

In vitro analysis of novel POR mutations

Analysis of novel splice site mutations was performed in silico, using the NetGene2 online prediction tool (www.cbs.dtu.dk/services/NetGene2/). Functional analysis of novel missense mutations was performed using a microsomal yeast coexpression assay as previously described (21). In brief, POR mutations were generated by PCR-based site-directed mutagenesis, cloned into the yeast expression vector pYcDE2, and transformed into hCYP17A1-expressing yeast. Microsomes were prepared as previously described (21) and protein content measured by the Bradford method (Bio-Rad, Hemel Hempstead, UK). To confirm expression of equivalent amounts of proteins, Western blot analysis of the microsomes was performed using polyclonal antibodies to human POR (Abcam, Cambridge, UK) and human CYP17A1 (Santa Cruz Biotechnology, Middlesex, UK). CYP17A1 17α-hydroxylase and 17,20 lyase activities with either wild-type or mutant POR were assessed as previously described (21), and calculation of enzyme kinetic parameters and subsequent statistical analysis was performed using curve-fitting software (Enzfitter 2.0.9.1; Biosoft, Cambridge, UK).

Biochemical analysis of urine steroid hormone metabolites

Analysis of urinary steroid metabolite excretion was performed by a quantitative gas chromatography/mass spectrometry (GC/MS) selected ion-monitoring method as described previously (22–24). In brief, steroids were enzymatically released from conjugation and, after extraction, chemically derivatized before GC/MS selected ion-monitoring analysis. Steroids quantified included metabolites of sex steroids, mineralocorticoids, glucocorticoids (cortisol and corticosterone), and their respective precursor steroids (Supplemental Table 1, published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org). Although numerous steroids (>30) form part of a routinely measured panel, the data have been simplified for reporting by calculating substrate metabolite to product metabolite ratios. These ratios were developed to allow diagnosis of many steroidogenic disorders and in this instance calculated to define the impact of the POR mutations on distinct steroidogenic activities, as described previously (25). Steroid 21-hydroxylase activity was defined as the ratios of pregnanetriolone to cortisol metabolites [100*PT′ONE (pregnanetriolone)/(tetrahydrocortisone [THE]+tetrahydrocortisol [THF]+5αTHF)] and 17-hydroxyprogesterone (17OHP) metabolites/cortisol metabolites [(17-hydroxypregnanolone [17HP] + pregnanetriol [PT])/(THE+THF+5αTHF)]. 17α-Hydroxylase activity was defined as the ratio of corticosterone metabolites over cortisol metabolites [(tetra-11-dehydrocorticosterone [THA]+5αTHA+tetrahydrocorticosterone [THB]+5αTHB)/(THE+THF+5αTHF)], and 17,20-lyase activity was assessed as the ratio of 17HP metabolites over androgen metabolites [(17HP+PT)/(androsterone + etiocholanolone)]. The ratio of the progesterone metabolite pregnanediol over cortisol metabolites [PD/(THE+THF+ 5αTHF)] was used as a PORD-specific diagnostic ratio. A second PORD-specific ratio has not been included here in aid of simplicity, although it is reported in other publications from our group. This is the ratio of the pregnenolone metabolite pregnenediol (5-pregnene-3β,20α-diol) to cortisol metabolites. This ratio is particularly valuable in diagnosing neonates and young infants in which 3β-hydroxy-Δ (5) steroids dominate.

PORD malformation score

Malformations observed in PORD include midface hypoplasia, craniosynostosis, phalangeal malformations (such as arachno-/campto-/clinodactyly, brachytelephalangia, rocker bottom feet), synostosis of large joints, and additional, nonskeletal malformations. We developed a malformation scoring system to assess the interindividual differences in the severity of skeletal malformations based on clinical and radiological assessments. This scoring system consists of six domains: midface hypoplasia, craniosynostosis, hand and feet malformations, large joint synostosis, femoral bowing, and the presence of additional malformations (Table 1). The minimal score is 0 (no malformations) and the maximum score of 16 indicates the highest possible severity. Information on malformations was requested from the referring clinician in a standardized fashion; subsequently, three clinicians independently scored each patient based on the information provided and agreed on a final score for each domain.

Table 1.

The six domains of the proposed PORD malformation score

Score	Midface hypoplasia	Craniosynostosis	Hand and feet malformationsa	Large joint synostosisb	Femoral bowing	Additional malformations
0	None	None	None	None	None	None
1	Moderate (low set ears, pear shaped nose)	Moderate (one suture affected, e.g. brachycephaly)	One hand/feet malformation	Extension deficit	Present	One additional malformation
2	Severe (compressed skeletal midface structure, proptosis)	Severe (two or more sutures affected, e.g. turricephaly)	Two hand/feet malformations	Fixed synostosis, joint contractures of one large joint	Complicated by neonatal fractures	Two or more additional malformations
3	Complicated by choanal stenosis/atresia that might require tracheostomy	Complicated by hydrocephalus, with ventriculoperitoneal shunt requirement	Three or more hand/feet malformations	Fixed synostosis, joint contractures of several large joints
Maximum domain score	3	3	3	3	2	2

The TMS ranged from a minimum of 0 to a maximum of 16 (TMS 0, no malformations; TMS 1–4, mild malformations; TMS 5–8, moderate malformations; TMS 9–16, severe malformations).

Phenotypic spectrum: arachnodactyly, clinodactyly, camptodactyly, metacarpal and metatarsal synostoses, wrist deviation, rocker bottom feet, talipes.

Predominantly radiohumeral and radioulnar malformations, but in more severe cases, also other joints including knees and ankles.

Results

Spectrum of POR gene mutations

Direct sequencing of the POR gene showed a variety of different molecular genetic defects including 14 novel mutations. These include three nonsense, three splice site, four frame shift, and one missense mutation as well as a 12-bp in-frame duplication (Fig. 1, A and B, and Table 2). In addition, the patients harbored one previously described frame shift, one known nonsense, and seven known missense mutations. In four unrelated patients presenting with hallmark clinical and biochemical features of PORD, only one mutant allele was identified by direct sequencing; in two of these, MLPA detected a deletion of exons U1 and 1 and a partial duplication affecting exons 2–5, respectively (Supplemental Fig. 1). In two unrelated patients with classic clinical and biochemical PORD presentation, no POR mutation was detected either by direct sequencing or by MLPA analysis (Table 3).

Fig. 1.

Localization of POR mutations at the DNA and protein level. A, Schematic representation of the POR gene. The untranslated exon U1 is given as a black box and coding exons are numbered. Novel mutations are given in red font, and the most common mutation p.A287P in exon 8 is marked in blue. Mutations cluster in the 3′ region without any additional common mutation apart from the p.A287P mutation. The partial deletion del ex U1–1 and the partial duplication dup ex2–5 that were identified by MLPA are represented by labeled red boxes. B, Schematic representation of the POR protein. Mutant residues are located in various locations all over the three functional domains of the POR protein. C, Three-dimensional POR protein model illustrating the localization of mutant residues in relation to the three functional moieties: nicotinamide adenine dinucleotide phosphate (NADPH; blue), flavin adenine dinucleotide (FAD; yellow), and flavin mononucleotide (FMN; green). Residues affected by missense mutations are given in dark yellow, frame shift mutations in orange, and residues affected by nonsense mutations as well as an in-frame duplication are marked in red.

Table 2.

Identified POR gene mutations on 54 unrelated alleles

g.DNA position	Protein position	Location	Coding effect	Alleles	First published by
Del ex U1–1		Ex U1, 1	Deletion	1	Novel
Dup ex 2–5		Ex 2, 3, 4, 5	Duplication	1	Novel
25,481dupA	p.Y87X	Ex 3	Nonsense	1	Novel
26,404 A>G	p.T142A	Ex 4	Missense	1	(7)
27,080 T>G	p.Y181D	Ex 5	Missense	1	(4)
27,102_27,113dup	p.G188_V191dup	Ex 5	Duplication	1	Novel
28,230 A>T	IVS6 − 2A>T	IVS 6	Splice site	1	Novel
27,550 C>T	p.R223X	Ex 6	Nonsense	1	Novel
28,332 dupT	IVS7 + 2dupT	IVS 7	Splice site	2	Novel
29,556 G>C	p.A287P	Ex 8	Missense	23	(3, 4)
29,645 G>A	IVS8 + 1G>A	IVS 8	Splice site	1	Novel
30,843dupC	p.Y376LfsX74	Ex 10	Frameshift	1	Novel
31,146dupC	p.I444HfsX6	Ex 11	Frameshift	1	(6)
31,180 delC	p.Q455RfsX90	Ex 11	Frameshift	1	Novel
31,187 G>A	p.R457H	Ex 11	Missense	2	(3, 4, 6)
31,603_31,604del	p.V472AfsX102	Ex 12	Frameshift	1	Novel
s31,681G>C	p.R498P	Ex 12	Missense	2	Novel
31,967G>A	p.C569Y	Ex 13	Missense	1	(3, 4)
31,988dupA	p.Y576X	Ex 13	Nonsense	1	Novel
32,062delG	p.E601SfsX12	Ex 13	Frameshift	1	Novel
32,171 G>A	p.Y607C	Ex 14	Missense	1	(31, 32)
32197C>T	p.R616X	Ex 14	Nonsense	1	(7)
32,234 A>C	p.H628P	Ex 14	Missense	1	(21)
Not detected				6

Ex, Exonic location; IVS, intervening sequence (designates intronic location).

Table 3.

Summary of clinical, genetic, and serum biochemical characteristics of PORD patients (n = 30)

Patient no.	Country of origin	Mutation on maternal allele	Mutation on paternal allele	Age at presentation	Karyotype	DSD (virilization/undermasulinization)	CYP17/CYP21 deficiency (GC/MS)	Increased baseline 17OHP	Baseline cortisol (nmol/liter)	Maximum cortisol after ACTH (nmol/liter)	Glucocortioid replacement
P01[a,b]	PL	R457H	A287P	16 yr	XX	Y	Y	26	290	436	+
P02	PL	—	—	NN	XX	Y	Y	43	745	866	−
P03a	D	C569Y	Y181D	NN	XX	Y	Y	5	291	280	+
P04a	D	C569Y	Y181D	NN	XY	N	Y	49	203	316	+
P05c	D	A287P	—	NN	XX	Y	Y	20	n/a	127	++
P06c	USA	A287P	IVS6 − 2A>T	NN	XX	N	Y	n/a	462	512	+
P07c	USA	A287P	IVS6 − 2A>T	NN	XY	Y	ND	16	253	536	+
P08a	D	A287P	A287P	NN	XY	N	Y	103	165	248	++
P09c	USA	A287P	V472AfsX102	NN	XY	Y	Y	n/a	209	481	+
P10	UK	Q455RfsX544	IVS7 + 2dupT	NN	XY	Y	Y	n/a	142	866	−
P11a	USA	A287P	H628P	NN	XX	Y	Y	31	170	209	++
P12b	PL	A287P	A287P	12 yr	XX	Y	Y	72	607	717	−
P13c	NL	A287P	Del ex U1–1	NN	XX	N	Y	128	140	210	++
P14	UK	A287P	IVS8 + 1G>A	NN	XX	Y	Y	39	n/a	n/a	++
P15	USA	A287P	A287P	NN	XX	Y	Y	n/a	n/a	n/a	++
P16	USA	A287P	A287P	NN	XY	N	Y	747	n/a	n/a	++
P17	CAN	A287P	G188_V191dup	NN	XY	Y	Y	n/a	20	209	++
P18	USA	A287P	A287P	NN	XX	Y	ND	n/a	121	303	+
P19	PAK	A287P	I444HfsX6	NN	XY	Y	ND	ND	ND	ND	ND
P20b	ITA	A287P	A287P	24 yr	XX	Y	ND	68	149	158	++
P21	AUT	Y87X	—	NN	XX	Y	Y	95	n/a	n/a	++
P22	MEX	—	—	Stillborn	XX	Y	Y	ND	ND	ND	ND
P23b	AUS	R457H	Y576X	NN	XY	Y	Y	112	198	229	++
P24a	PL	Y607C	E601SfsX12	NN	XY	Y	Y	10	391	494	+
P25	PAK	R498P	R498P	NN	XX	Y	Y	n/a	495	548	+
P26b	NL	Y376LfsX74	T142A	18 yr	XX	N	Y	25	375	395	+
P27	D	A287P	R616X	NN	XY	N	Y	6	115	203	++
P28b	UK	A287P	IVS7 + 2dupT	12 yr	XY	N	Y	14	353	411	+
P29b	D	A287P	R223X	16 yr	XX	Y	Y	15	190	314	++
P30	NL	A287P	Dup ex 2–5	31 yr	XX	Y	Y	20	360	350	+

Glucocorticoid replacement is as follows: −, none; +, hydrocortisone cover in stress/intercurrent illness; ++, permanent hydrocortisone replacement; ND, no glucocorticoid replacement implemented as no live birth. NN, Neonatal; Y, yes; N, no; n/a, not available; ND, not done; PL, Poland; D, Germany; USA, United States of America; UK, United Kingdom; NL, The Netherlands; CAN, Canada; PAK, Pakistan; ITA, Italy; AUT, Austria; MEX, Mexico; AUS, Australia. Mutation: —, Alleles without identified disease-causing mutation.

Patients whose geno- and phenotype was reported in a previous publication: P01 [case 1 (4)], P03 [case 2 (4)], P04 [case 3 (4)]; P08 (33); P11 (34); and P24 (32).

Patients with a description of pubertal development in a previous publication (18): P01 (case 1), P12 (case 2), P20 (case 3), P26 (case 4), P29 (case 5), P23 (case 6), and P28 (case 7).

Patients with a description of clinical phenotype only in a previous publication: P13 [case 2 (22)], P01 [case 3 (22)], P06 [case 4 (22)], P09 [case 5 (22)], P3 [case 6 (22)], P4 [case 7 (22)]; P07 [case 1 (35)], P06 [case 2 (35)]; and P05 (36).

The missense mutation p.A287P (g.29,556G>C) was the most common point mutation, detected in 23 of 54 unrelated alleles (43%). Six patients (five unrelated) carried p.A287P on both alleles and one unrelated patient was homozygous for the novel point mutation p.R498P; the remaining 23 patients (21 unrelated) were compound heterozygous. The missense mutation p.R457H (g.31,187G>A) was found on two unrelated alleles in patients of Polish and mixed Chinese/Hispanic origin, respectively. A splice site mutation in intron 7 (IVS7 + 2dupT) was detected in two unrelated patients. All other mutations were identified only in single families (Table 3). Apart from p.A287P, no other frequently occurring mutation was identified. However, mutations clustered in the 3′ portion of the POR gene, with their translation impacting on the flavin adenine dinucleotide and nicotinamide adenine dinucleotide phosphate binding domains of the POR protein (Fig. 1, B and C).

Functional analysis of novel mutants

Results of the in silico analysis of the three novel splice site mutations (Supplemental Table 2) confirmed complete loss of predicted POR wild-type splice donor and acceptor sites, respectively, and thus loss of POR function. Confidence values for the wild-type sequence predicting splicing ranged between 0.95 and 0.99, indicative of a high accuracy of prediction.

In vitro functional analysis of p.R498P and p.G188_V191dup confirmed reduction of CYP17A1 17α-hydroxylase activity to 66 and 37% of wild-type catalytic efficiency [maximum enzyme velocity (Vmax)/Michaelis-Menten constant (Km)], respectively. Coexpression of both mutants with human CYP17A1 reduced the maximal velocity of the enzymatic reaction, whereas the Michaelis constant remained essentially unaltered. Impairment of 17,20 lyase by displayed a similar pattern, consequently decreasing catalytic efficiency to 58 and 70% of wild type (Supplemental Table 3).

Urinary steroid profiling

In 25 patients the diagnosis of PORD had been established employing urinary steroid profiling by GC/MS; detailed results were available in 23 of them. Five established diagnostic substrate to product steroid metabolite ratios were applied to measure the impact on specific steroidogenic enzymes, namely 17α-hydroxylase and 21-hydroxylase (Fig. 2). All five diagnostic ratios were increased above the normal reference range in 16 of 23 patients; an additional four had four of five ratios pathologically increased (Fig. 2).

Fig. 2.

In vivo steroidogenic enzyme activities in PORD patients (n = 23) as determined by urinary steroid profiling. Diagnostic steroid substrate over product ratios reflective of distinct steroidogenic enzyme activities and measured by GC/MS are shown in comparison with an age-matched reference cohort. Box plots represent the interquartile ranges (25th to 75th percentile), whiskers the fifth and 95th percentiles, respectively, of the reference cohort; each PORD case is represented by specific symbols.

Adrenal insufficiency

The outcome of adrenal function assessment by a short cosyntropin test was documented in 27 patients, and 24 were found to have adrenal insufficiency (Table 3). Of those, 13 received permanent hydrocortisone replacement, whereas the remaining 11 were treated with hydrocortisone cover during intercurrent illness and surgical stress (Table 3). Only three patients showed a sufficient cortisol response to ACTH stimulation; one of these patients, P10, had shown normal values for both urinary steroid metabolite ratios reflective of 21-hydroxylase activity (Fig. 2). Of note, there was no correlation between POR genotype and the presence and severity of glucocorticoid deficiency.

Baseline 17OHP concentrations were available in 21 patients and increased in all but one patient, with a wide range of 17OHP concentrations [median 31 nmol/liter (942 ng/dl), range 5–747 nmol/liter (162–24,651 ng/dl)] (Table 3). However, in the majority of patients, apart from one patient (P16), serum 17OHP levels were only mildly elevated and considerably lower than typically observed in classic 21-hydroxylase deficiency due to mutant CYP21A2.

Genital phenotype

The majority of patients (70%) presented with either 46,XX DSD (15 of 18; 83%) or 46,XY DSD (seven of 12; 58%). However, patients of both sexes also presented with normal genital appearance consistent with their chromosomal sex (three females and six males), all of whom carried one major loss-of-function mutation. By contrast, all 12 patients with homozygous or compound heterozygous missense mutations presented with 46,XX DSD and normal male genital appearance in 46,XY individuals. Six patients were homozygous for p.A287P; all four 46,XX individuals presented with 46,XX DSD, whereas the two 46,XY individuals had normal male external genitalia. Intronic splice site mutations were of intermediate effect, largely defined by the mutation severity on the other allele (Table 3 and Supplemental Table 4).

Malformation phenotype

The majority of patients with no, mild [total malformation score (TMS) 1–4)], to intermediate malformations (TMS 5–8) were compound heterozygous or homozygous for two missense mutations (Table 4). By contrast, all patients with more severe malformations (TMS 9–16) were compound heterozygous, carrying a null mutation on one allele (Table 4). None of the patients with mild to intermediate malformations reached a maximal domain score for midface hypoplasia, craniosynostosis, or synostosis of the large joints. The most variable presentations were observed for hand and feet malformations, which were severe in some patients with otherwise mild malformation phenotype and vice versa (Table 4). All patients homozygous for p.A287P had mild to moderate malformations, with total malformation scores varying between 4 and 8, and none of these patients had a maximum score in any of the malformation domains, indicating the absence of severe malformations and their complications.

Table 4.

POR genotype and malformation type and severity in the PORD cohort (n = 29)

TMS	Mutant allele 1	Mutant allele 2	Patient no.	Midface hypoplasia	Craniosynostosis	Phalangeal malformations	Large joint synostosis	Femoral bowing	Additional malformations	Type of additional malformation
0	Y181D	C569Y	P4	0	0	0	0	0	0
0	Y87X	—	P21	0	0	0	0	0	0
1	Y607C	E601SfsX12	P24	1	0	0	0	0	0
4	A287P	R457H	P01	1	0	2	0	0	1	Scoliosis
4	A287P	Del ex U1–1	P13	2	0	0	2	0	0
4	Y181D	C569Y	P03	1	0	3	0	0	0
4	A287P	A287P	P08	1	2	1	0	0	0
5	A287P	A287P	P12	1	1	2	1	0	0
5	T142A	Y376LfsX74	P26	2	0	3	0	0	0
5	A287P	A287P	P20	1	0	3	0	0	1	Scoliosis
5	A287P	IVS7 + 2dupT	P28	3	1	0	1	0	0
5	A287P	Dup ex 2–5	P30	2	0	2	1	0	0
6	—	—	P02	2	1	1	0	1	1	Dysplastic ears
6	A287P	R223X	P29	3	1	0	2	0	0
7	A287P	H628P	P11	2	1	1	2	0	1	Scoliosis
7	R498P	R498P	P25	2	0	2	2	1	0
7	A287P	A287P	P15	2	2	1	2	0	0
8	A287P	A287P	P16	2	2	2	1	0	1	Pectus excavatum
8	A287P	R616X	P27	2	2	1	3	0	0
9	A287P	—	P05	2	2	2	3	0	0
9	A287P	IVS8 + 1G>A	P14	2	2	2	3	0	0
9	A287P	G188_V191dup	P17	3	0	2	0	2	2	Absent left kidney, frontal capillary hemangioma
10	A287P	I444HfsX6	P19	3	3	0	2	1	1	Scoliosis
10	A287P	V472AfsX102	P09	2	2	2	3	0	1	Two-vessel umbilical cord
13	R457H	Y576X	P23	3	2	3	3	0	2	Scoliosis, preauricular pits
14	—	—	P22	2	2	3	3	2	2
15	A287P	IVS6 − 2A>T	P06	3	3	3	2	1	3	Anteriorly placed anus, single umbilical artery, dysplastic ears
15	A287P	IVS6 − 2A>T	P07	3	3	3	3	2	1	Dysplastic ears
15	Q455RfsX544	IVS7 + 2dupT	P10	3	3	3	3	2	1	Arnold-Chiari malformation

For definition of PORD malformation score, please see Table 1. Mutation: —, Alleles without identified disease-causing mutation.

Discussion

PORD manifests with a wide range of clinical signs and symptoms, and our study shows a broader phenotypic expression in Caucasian patients than previously observed in a large Japanese cohort (3, 7, 14, 15). This is mainly attributable to a greater variety of detected POR mutations. In this cohort containing the largest number of well-described Caucasian PORD patients to date, we validated the previous notion that p.A287P is the most prevalent mutation in Caucasians. Using MLPA, we were able to identify a partial deletion (exons U1–1), also previously described elsewhere (14), and for the first time a partial duplication (exons 2–5) of the POR gene on alleles with no mutation detected by direct sequencing. This illustrates that MLPA is a highly useful addition to the genetic work-up in PORD. However, despite using MLPA, we did not detect mutations on 11% of unrelated alleles. In a Japanese PORD cohort (14), no mutations were identified in 5% unrelated alleles (three of 58) in a largely Caucasian cohort (7) in 26% of alleles (12 of 46).

Our data illustrate that urinary steroid metabolite profiling by GC/MS can be considered a valuable noninvasive tool for the comprehensive diagnosis of PORD with clear diagnostic criteria. Applying diagnostic steroid substrate to product ratios can establish an unambiguous diagnosis of PORD and also provides insight into the differential impairment of steroidogenic enzyme function by mutant POR. Although the data in our study were based on 24-h urine collections, similar information can be obtained from a GC/MS analysis of spot urines (25). Obviously the diagnosis of PORD can also be established by the combined measurement of serum pregnenolone, progesterone, and 17-hydroxypregnenolone before and after ACTH stimulation (15).

GC/MS analysis can also be instrumental in analyzing the differential effects of POR mutations on different steroidogenic CYP enzymes. We have demonstrated this in a previous study by describing significantly increased corticosterone over cortisol metabolite ratios in patients with homozygous p.A287P (21). These in vivo findings were entirely consistent with the in vitro findings that indicated a significantly higher impairment of 17α-hydroxylase over 21-hydroxylase for p.A287P (21).

Based on clinical presentation and steroid measurements, some cases of PORD have previously been mistakenly diagnosed as 21-hydroxylase deficiency, aromatase deficiency (16), or even isolated 17,20 lyase deficiency (17). It is particularly important to consider PORD in the differential diagnosis of 21-hydroxylase deficiency because some of the previously described cases have been detected in 21-hydroxylase newborn screening programs (3, 14). In our cohort, serum 17OHP was increased in almost all patients but to considerably lower levels than usually observed in classic 21-hydroxylase deficiency. Establishing the correct diagnosis has significant consequences for the choice of glucocorticoid dose because PORD patients require only glucocorticoid replacement, whereas 21-hydroxylase-deficiency patients usually require higher glucocorticoid doses to control the concomitant adrenal androgen excess. The degree of 17-hydroxylase impairment observed in our PORD patients determined by the corticosterone metabolite to cortisol metabolite ratio [(THA+5αTHA+THB+5αTHB)/(THE+THF+5αTHF; median 0.95, range 0.12–1.8)] is significantly lower than previously observed in patients with 17-hydroxylase deficiency due to inactivating CYP17A1 mutations [(THA+5αTHA+THB+5αTHB)/(THE+THF+5αTHF); median 95, range 64–11] (26). Similarly, the diagnostic steroid ratio indicative of 21-hydroxylase deficiency [17OHP over cortisol metabolites, (17HP+PT)/(THE+THF+5αTHF] has been shown to be increased above 10 in patients with classic CAH due to mutant CYP21A2 and above 0.34 in nonclassic 21-hydroxylase deficiency patients (23). In our PORD cohort, this ratio has a median of 0.52 (range 0.04–1.20), an excellent fit with the serum 17OHP concentrations in our PORD patients, which largely varied within the range observed for nonclassic 21-hydroxylase deficiency patients.

The overwhelming majority of our PORD patients had adrenal insufficiency manifesting with isolated glucocorticoid deficiency, as documented by cosyntropin testing, and required either permanent hydrocortisone replacement or at least stress dose cover. In contrast to other reports (3, 6, 7, 13), none of our patients presented with clinical signs and symptoms of adrenal insufficiency, and several of our patients (18) underwent only formal testing many years after presenting with DSD and/or skeletal malformations. Random baseline cortisol is not a sensitive tool for detection of adrenal insufficiency, and therefore, it is mandatory to assess adrenal function by measuring serum cortisol before and after ACTH, which is the accepted standard for the diagnosis of adrenal insufficiency (27, 28). This is particularly true in PORD patients who often have normal basal cortisol secretion but fail to increase cortisol production in response to ACTH and thus are prone to deterioration during intercurrent illness.

Mineralocorticoid deficiency is not part of the biochemical presentation in PORD. By contrast, as demonstrated by the steroid metabolome analysis in our cohort, elevated corticosterone metabolites indicative of relative mineralocorticoid precursor excess are a frequent finding in PORD. We previously demonstrated predominant 17α-hydroxylase over 21-hydroxylase impairment by p.A287P (21), and indeed, arterial hypertension has been documented in homozygous carriers (3, 12), probably caused by excess deoxycorticosterone production as in classical 17α-hydroxylase deficiency. Systematic analysis of blood pressure should therefore become an integral part of clinical monitoring in PORD.

Genotype-phenotype correlations with regard to DSD presentation in PORD are most puzzling, and in our study we found DSD in both 46,XX and 46,XY individuals but also normal genital development in individuals of both sexes. A study performed in Japan showed a predominance of 46,XX DSD over 46,XY DSD, with 100% of 46,XX DSD but only 44% of 46,XY DSD patients classified as any genital feature at birth (14). Our findings are similar, despite a completely different and diverse ethnic/racial background in our population: more than 80% of our 46,XX patients presented with DSD, whereas DSD was present in only half of affected 46,XY individuals (Supplemental Table 3). The Japanese data suggest that p.R457H might be associated with virilization in 46,XX patients but most likely results in normal male genital development (14). Before our study only two p.A287P homozygous patients have been described; one with an unclear karyotype and genital appearance (7) and the other with ambiguous genitalia and 46,XY DSD (3). Our data based on six carefully characterized patients carrying homozygous p.A287P reveal p.A287P as associated with 46,XX DSD in four of four 46,XX individuals and with normal male genitalia in two of two 46,XY individuals.

It has been previously suggested that impairment of CYP19A1 aromatase activity by mutant POR may be responsible for the 46,XX DSD presentation in PORD because inactivating CYP19A1 mutations result in female virilization. However, if the classic androgen pathway via dehydroepiandrosterone is down-regulated due to significantly impaired 17,20 lyase activity in PORD patients, then there would not be a relevant amount of substrate for CYP19A1 to make this matter. Moreover, a previous in vitro study found that p.A287P does not impair CYP19A1 activity, but p.R457H abolishes aromatase activity (29). However, both mutations result in 46,XX DSD when present in the homozygous state, thus precluding a significant influence of aromatase on the 46,XX DSD phenotype. We have previously suggested that 46,XX DSD in PORD occurs through androgen generation via the alternative androgen synthesis pathway (4). This pathway also involved CYP17A1 17,20 lyase activity, and it has been previously shown that the CYP17A1 enzyme has a higher substrate preference for the alternative pathway over the classic pathway steroid substrates for its 17,20 lyase reaction (30). Thus, one could speculate that p.A287P and p.R457H associated with 46,XX DSD might have a high residual activity in the alternative pathway, whereas p.H628P, shown to result in 46,XY DSD in our cohort, may have low residual activity, a hypothesis to be explored by future studies.

PORD-associated malformations have previously been reported only in a qualitative fashion. To more objectively assess the spectrum and quantify the severity of malformations in PORD patients, we developed a novel scoring system to standardize malformation assessment reported in this study. Our data provide the basis for an improved detection and more accurate assessment of PORD-associated malformations based on clinical and radiological assessment, thereby facilitating early diagnosis and genetic counseling. Although PORD patients without obvious malformations have been described (3, 7, 17), it is possible that mild malformations may be overlooked. In several of our patients, neonatal assessment had not revealed obvious abnormalities, but careful reassessment after the diagnosis of PORD established the existence of several malformations, mostly affecting the midface, fingers, and toes. We found that patients with an absent or mild to moderate skeletal malformation phenotype were either homozygous or compound heterozygous for missense mutations, whereas severe malformations were invariably associated with a major loss-of-function mutation on one of the affected alleles (Table 4).

The 46,XY individuals with DSD showed a trend toward higher malformation scores (Supplemental Fig. 2), possibly suggesting that more severe mutations equally disrupt sex steroid synthesis and bone development. Interestingly, the patient with the most severe malformation score, P10, presented with 46,XY DSD but had a normal cortisol response to ACTH, nicely illustrating that a distinct POR mutation can have very different effects on different electron-accepting CYP enzymes.

In conclusion, we have described detailed clinical, biochemical and genetic findings in a large Caucasian PORD cohort. We have established the usefulness of MLPA as an addition to molecular genetic analysis of the POR gene. We provide a novel scoring system for standardized assessment of PORD-associated malformations. Major loss-of-function mutations on one of the affected alleles are associated with severe malformations, whereas homozygosity or compound heterozygosity for missense mutations predicts a mild to moderate malformation phenotype. Importantly, homozygosity for the most common mutation in Caucasians, p.A287P, allows for prediction of the genital phenotype and is associated with mild to moderate malformations. Adrenal insufficiency is present in the large majority of patients and cannot be predicted by genotype. Long-term follow-up of well-characterized patients will reveal the full extent of the impact of mutant POR on the health status of affected patients.