Molecular genetic evaluation of pediatric renovascular hypertension due to renal artery stenosis and abdominal aortic coarctation in neurofibromatosis type 1

Abstract

The etiology of renal artery stenosis (RAS) and abdominal aortic coarctation (AAC) causing the midaortic syndrome (MAS), often resulting in renovascular hypertension (RVH), remains ill-defined. Neurofibromatosis type 1 (NF-1) is frequently observed in children with RVH. Consecutive pediatric patients (N = 102) presenting with RVH secondary to RAS with and without concurrent AAC were prospectively enrolled in a clinical data base, and blood, saliva and operative tissue, when available, were collected. Among the 102 children, 13 were having a concurrent clinical diagnosis of NF-1 (12.5%). Whole exome sequencing was performed for germline variant detection, and RNA-Seq analysis of NF1, MAPK pathway genes and MCP1 levels were undertaken in five NF-1 stenotic renal arteries, as well as control renal and mesenteric arteries from children with no known vasculopathy or NF-1. In 11 unrelated children with sequencing data, 11 NF1 genetic variants were identified, of which 10 had not been reported in gnomAD. Histologic analysis of NF-1 RAS specimens consistently revealed intimal thickening, disruption of the internal elastic lamina and medial thinning. Analysis of transcript expression in arterial lesions documented an approximately 5-fold reduction in NF1 expression, confirming heterozygosity, MAPK pathway activation and increased MCP1 expression. In summary, NF-1-related RVH in children is rare but often severe and progressive and, as such, important to recognize. It is associated with histologic and molecular features consistent with an aggressive adverse vascular remodeling process. Further research is necessary to define the mechanisms underlying these findings.

Introduction

Renovascular hypertension (RVH) secondary to renal artery stenosis (RAS) without or with abdominal aortic coarctation (AAC) often presenting as the midaortic syndrome (MAS) is an important cause of pediatric hypertension (Fig. 1). The etiology of the underlying vascular disease is poorly understood, but its clinical complications are well described and include failure to thrive, heart failure, hypertensive encephalopathy, hemorrhagic stroke and early mortality (1). Although pediatric hypertension is most common due to thoracic aortic coarctation or parenchymal renal disease, renovascular disease accounts for as many as 5–10% of these cases (2). The renal artery and aortic narrowings in these cases appear to represent a developmental anomaly. MAS is a clinical entity believed to be related to events occurring near the 25th day of fetal development; specifically, over-fusion of the two embryonic dorsal aortas or the absence of one of the paired aortas. The renal artery and aortic narrowings reduce intrarenal arterial blood pressure resulting in the production of excessive renin, subsequent activation of the angiotensin system and secondary hypertension (3,4).

Figure 1.

(A) CT-angiogram of a 9-month-old with NF-1 and bilateral ostial renal artery stenosis (RAS) associated with post-stenotic dilation of the mid-left renal artery; (B) catheter-based arteriogram of a 3-year-old with NF-1 having an abdominal aortic coarctation and bilateral renal artery ostial stenoses.

Patients with neurofibromatosis type 1 (NF-1, OMIM #162200) exhibit an unusually high frequency of arterial abnormalities, including those associated with pediatric RVH. In the past, nearly 20% of pediatric patients undergoing surgical interventions for RAS and AAC at the University of Michigan carried a diagnosis of NF-1, and others have identified similar arterial disease in up to 18% of NF-1 children (5,6). While NF-1 is inherited in an autosomal dominant manner, the etiology of vascular lesions in NF-1 remains undefined. Nonvascular genotype–phenotype correlations in NF-1 have been reported, predominantly in association with large NF1 gene deletions in children having dysmorphic facial features, developmental delays and increased frequencies of cutaneous, subcutaneous and plexiform neurofibromas (7–13). The spectrum of NF1 genetic variation in clinically affected individuals and the severity of NF-1-related arterial diseases have not been previously reported. The current investigation was undertaken to characterize the vascular disease as it may relate to genetic NF1 variants in pediatric patients with clinical NF-1 and RVH.

Results

Clinical presentation

In total, 13 unrelated patients in a biorepository of 102 patients with pediatric arterial dysplasia (12.7%) carried a clinical diagnosis of NF-1 at the time of enrollment. All NF-1 patients presented for surgical treatment of RVH (Table 1). Medically refractory HTN unresponsive to optimal drug therapy was the basis for surgical intervention in most patients. Additional indications for surgery included medically managed hypertension with need for renal preservation following failed renal artery stent (Probands 9 and 11), existence of a solitary kidney (Proband 12) or post-stenotic renal artery dilation/aneurysm (Probands 7 and 10). There was a slight male predominance (eight boys, five girls). Average age at time of RVH diagnosis was 3.6 years. Anatomic features of arteriopathy most commonly included an ostial RAS (11 of 13; Fig. 1A), MAS (Fig. 1B) and ostial celiac or superior mesenteric artery stenoses (Table 2). Notably, of the 11 patients with an ostial RAS, 9 had other vascular stenoses or an aneurysm. There were no cases of Moya-Moya syndrome, nor stroke identified. Common clinical features of NF-1 (Table 3) included cutaneous café-au-lait spots (N = 13), behavioral and cognitive challenges (N = 7), optic manifestations including optic gliomas most often (N = 4), skeletal manifestation including scoliosis most often (N = 4) and peripheral plexiform neurofibromas (N = 3). A comparison of those children diagnosed with RVH early in life (0–3 years) and those diagnosed later in life (4–18 years) is included in Supplementary Material, Table S2. Notably, there were no malignant tumors identified in this cohort, neither at the sites of vascular lesions nor elsewhere. Operations were considered successful with the exception of one child’s death from multi-system organ failure (hepatic failure and abdominal compartment syndrome) following an aorto-renal reconstruction. Postoperatively, hypertension outcomes were classified as cured if children were normotensive with blood pressures <90th percentile for sex, age and height while receiving no antihypertensive medications, or improved if their blood pressures were within normotensive ranges while on a less intensive antihypertensive regimen. While five patients required reoperation for recurrent arterial dysplastic stenoses, all reoperations were successful.

Table 1.

Clinical characteristics of children with neurofibromatosis type 1 (NF-1) and renovascular hypertension

Proband #	Sex	Age at diagnosis (years)	Presenting symptoms and signs	Additional medical history	Family history of NF-1 (relation)	Recurrence/reoperation	Hypertension outcome at last follow-up
1	M	3	Asymptomatic, medically refractory three-drug HTN (CCB, BB, Diur), LVH		No	No	Improved
2	M	2	Dyspnea on exertion, medically refractory three-drug HTN (CCB, ACE-I, Clonidine), LVH	Juvenile xanthogranulomatosis	No	Yes	Improved
3	F	6	Headache, failure to thrive, CMI, medically refractory three-drug HTN (CCB, BB, Clonidine), LVH		Yes (Mother, m-Cousin, m-Grandmother)	No	Improved
4	M	<1	Asymptomatic, medically refractory four-drug HTN (CCB, BB, ACE-I, ARB)		Unknown (Adopted)	No	Improved
5	F	1	Asymptomatic, medically refractory three-drug HTN (CCB, BB, Clonidine), LVH		Yes (Father)	Yes	Improved
6	M	2	Lower extremity fatigue with exertion, medically refractory four-drug HTN (AB, BB, Clonidine, Diur), LVH		Yes (Father)	Yes	Improved
7	F	7	Failure to thrive, medically refractory two-drug HTN (CCB, BB)	Craniofacial syndrome	Yes (Mother, m-Grandmother)	No	Cure
8	F	3	Headache, epistaxis, medically refractory four-drug HTN, LVH		Yes (Father, Brother)	No	Improved
9	M	6	Asymptomatic one-drug HTN (CCB), failed renal artery stent		No	Yes	Improved
10	M	4	Asymptomatic poorly controlled one-drug HTN (ACE-I)		No	No	Cure
11	M	10	Asymptomatic poorly controlled two-drug HTN (CCB, ACE-I), failed renal artery stent	Strabismus	Yes (Mother, Brother, Sister)	Yes	Improved
12	M	2	Epistaxis, two-drug HTN (CCB, ACE-I), single kidney		No	Yes	Cure
13	F	1	Asymptomatic hypertensive urgency (>five-drug HTN (AB, BB, CCB, ACE-I, Clonidine), LVH		No	No	Mortality

HTN = hypertension; CCB = calcium channel blocker; BB = beta-blocker; Diur = thiazide diuretic; LVH = left ventricular hypertrophy; ACE-I = angiotensin-converting enzyme inhibitor; ARB = angiotensin receptor blocker; M = male; F = female; m = maternal; AB = alpha-blocker

Table 2.

Vascular anatomy of arteriopathy in NF-1 children with renovascular hypertension

Proband #	Renal artery stenosis	Mesenteric stenosis	Aortic stenosis	Aneurysm	Intra-cranial or cervical AD	Other
1	Ostial	SMA (ostial)	Infra-renal	0	(no imaging)
2	Ostial	SMA (ostial)	Supra-renal	0	0
3	Ostial	Celiac and SMA (ostial)	Supra-renal	0	(no imaging)	Hypogastric artery hypoplasia
4	Ostial	0	0	0	(no imaging)
5	Ostial	Celiac and SMA (ostial)	Supra-renal	0	(no imaging)
6	Ostial	Celiac and SMA (ostial)	Supra-celiac	0	(no imaging)	Congenital absence of left kidney
7	Diffuse ectasia and dysplasia; ostial and distal stenoses with bi-lobed aneurysm	0	0	Renal	(no imaging)
8	Ostial	0	Supra-renal	0	(no imaging)
9	Ostial	0	0	0	(no imaging)
10	Right renal artery aneurysm with diffuse intra-renal stenoses; left renal artery ostial stenosis	Celiac (ostial)	Supra-celiac	Renal	0
11	Ostial	Celiac (ostial)	Supra-celiac	0	(no imaging)
12	Ostial	0	Supra-renal	0	(no imaging)
13	Ostial	Celiac and SMA (ostial)	Supra-celiac	0	Tortuosity of distal L ICA/vertebral arteries

AD = arterial dysplasia, SMA = superior mesenteric artery, L = left, ICA = internal carotid artery

Table 3.

Clinical NF-1 features in pediatric patients with renovascular hypertension

Proband #	Skin	CNS	Peripheral nervous system	Optic	Skeletal	Cardiac	Other/cognitive
1	CAL spots			Unspecified NF involvement of the eye requiring orbitotomy			Behavioral issues
2	CAL spots	Pontine mass	Multiple plexiform NFs	No		No	Articulation delays; fine motor impairment; ADD
3	CAL spots			Optic Glioma		No
4	CAL spots; freckling; subcut NFs		No	No	Tibial pseudoarthrosis; multiple fractures; LLD	No	Learning disability; speech delay; macrocephaly
5	CAL spots	Chiari malformation		Optic glioma		No
6	CAL spots		No	Optic glioma	Scoliosis	No	Global developmental delay; ADHD; hearing loss
7	CAL spots; subcut NFs	FASI		No	Levo-scoliosis	No
8	CAL spots
9	CAL spots
10	CAL spots, subcut NF, gluteal hirsutism	No	Large plexiform NFs; chronic ‘nerve pain’ in arms/legs	No	Scoliosis; inter-trochanteric femur NF; LLD	No	Adjustment disorder; mixed anxiety/depression; intermittent explosive behavior d/o
11	CAL spots
12	CAL spots					No	Learning disability; developmental delay
13	CAL spots; freckling	No	Plexiform NF	No	No	VSD	Gross motor delay

No malignant tumors were identified in any patient, and clinical findings documented as of the most recent visit were recorded. CNS = central nervous system; CAL = café-au-lait; NF = neurofibroma; ADD = attention deficit disorder; Subcut = subcutaneous; LLD = limb length discrepancy; adHD = attention deficit hyperactivity disorder; FASI = focal areas of signal intensity or bright areas on T2-weighted brain MRI; d/o = disorder; VSD = ventricular septal defect)

Genetic variant annotation and pathogenicity estimation

In total, 11 of the 13 pediatric patients with a clinical diagnosis of NF-1 and RVH had DNA samples available for sequencing, although only 10 passed quality control for variant calling and annotation. One additional variant was included based upon RNA-Seq-identification, subsequently validated to be germline by Sanger sequencing of renal parenchymal tissue (case) and whole blood (control) (Supplementary Material, Fig. S1). Variants were predicted to be deleterious by in silico analysis in a candidate gene list of variants associated with arteriopathy/aortopathy (Supplementary Material, Table S1), from which five non-synonymous variants derived from four patients were identified (Supplementary Material, Table S3). Each variant identified by exome sequencing in the NF1 gene was annotated according to its predicted effect on the neurofibromin protein product (Table 4). NF1 truncating mutations were most often identified (Fig. 2), occurring in 8 of the 11 patients. Of the 11 NF1 variants, 10 were annotated as likely pathogenic or pathogenic according to the ACMG/ACGS 2020 guidelines, and 1 was classified as a variant of uncertain significance (VUS). Most of the NF1 variants were not observed in the gnomAD database (Table 4 accessed 24 July 2020), and three variants were annotated in ClinVar as pathogenic for NF-1 in ClinVar (accessed 24 July 2020), with no record of vascular disease or whether phenotyping evaluations had been performed (Supplementary Material, Table S4). Notably, Proband 9 carried a clinical diagnosis of NF-1 made previously by Medical Genetics at an outside hospital prior to consultation and treatment at our institution, with incomplete records regarding NF-1 phenotype data shown in Table 3.

Table 4.

NF1 genetic variants in NF-1 children with renovascular hypertension

Proband #	Exonic change	Exon	Base change	Amino acid change	GnomAD	ClinVar	CADD/SIFT/poly-phen2/UMD	ACMG & ACGS 2020
2	frameshift deletion	35	c.4714_4715del	p.(Phe1572Leufs*28)	–	–	−/−/−/−	P(PVS1, PS2, PM2)
3a	stopgain	36	c.4923G > A	p.(Trp1641*)	–	–	41/−/−/D	P(PVS1, PM2, PP3)
4	frameshift deletion	44	c.6648del	p.(Phe2217Serfs*27)	–	–	−/−/−/−	LP(PVS1_strong, PM2)
5	stopgain	1	c.26G > A	p.(Trp9*)	–	–	34/−/−/D	P(PVS1, PM2, PP3)
6	frameshift deletion	36	c.4881del	p.(Pro1629Leufs*48)	–	–	−/−/−/−	P(PVS1, PM2)
7	frameshift deletion	30	c.4000del	p.(Glu1334Lysfs*9)	–	P	−/−/−/−	P(PVS1, PM2, PP5)
8	frameshift deletion	42	c.6519del	p.(Glu2174Argfs*5)	–	–	−/−/−/−	LP(PVS1_strong, PM2)
9	non-frameshift deletion	8	c.806_808del	p.(Leu269del)	–	–	−/−/−/−	VUS(PM2)
10	non-synonymous	14	c.1586 T > C	p.(Leu529Pro)	–	–	26/D/D/−	LP(PS2, PM2, PP3)
12	non-synonymous	17	c.1885G > A	p.(Gly629Arg)	4 × 10−6	P	16/T/D/T	LP(PM1, PM2, PP3, PP5, BP4)
13	stopgain	26	c.3376C > T	p.(Gln1126*)	–	P	38/−/−/D	P(PVS1, PM2, PP3, PP5)

D = damaging, T = tolerated, P = pathogenic, LP = likely pathogenic, VUS = variant of uncertain significance; Accession number for NF1 annotation: NM_000267.3

^aThis variant was identified from RNA-Seq of RAS tissue in the absence of available blood or saliva and was validated by Sanger sequencing of renal parenchymal tissue (non-RAS), Supplementary Material, Fig. S1.

Figure 2.

Distribution of predicted deleterious NF1 variants. The NF1 domains were defined as follows: CSRD: cysteine/serine-rich domain (residues 543–909), TBD: tubulin-binding domain (residues 1095–1197), GRD: GTPase-activating protein-related domain (residues 1198–1530), SEC: Sec14-like domain (residues 1560–1705), PH: pleckstrin homology-like domain (residues 1716–1816), HLR: HEAT-like repeat regions (residues 1825–2428), SBR: syndecan-binding region (residues 2619–2719) (41). Additional gene domains not depicted include the NLS (nuclear localization signal 2534–2550) due to small size. The CTD (2260–2817) overlaps with the HLR and SBR domains. (Accession number for NF1 annotation: NM_000267.3).

Pedigree analysis and inheritance of NF1 variants

NF-1 was presumed inherited in 6 of 13 NF-1 pediatric patients with RVH with parental diagnoses of NF-1 in 3 fathers and 3 mothers (Table 1). Upon review of the cases of inherited NF-1, two patients with heritable NF-1 shared a history of early-onset hypertension with NF-1-affected family members (Fig. 3A and B). An additional patient with heritable NF-1 had an affected maternal grandfather with end-stage renal disease requiring hemodialysis (Fig. 3C). Finally, a patient with family history of NF-1 had at least two NF-1 non-affected family members with a diagnosis of early-onset HTN and/or end-stage renal disease (Fig. 3D). Those children with a presumed de novo mutation, according to a lack of family history of NF-1 diagnosis, did not have a family history notable for early-onset hypertension (i.e.: age < 50 years), chronic kidney disease, renal artery stenosis or aneurysm with one exception (Fig. 3E).

Figure 3.

Three-generation pedigrees of heritable (A–D) and de novo (E) pediatric NF-1 associated with a family history of early-onset hypertension and/or chronic kidney disease. Additional family history diagnoses include end-stage renal disease in maternal grandfather with NF-1 (C). (HTN = hypertension, RAS = renal artery stenosis, MAS = midaortic syndrome).

Histopathology of NF-1 renal arteries

The histopathologic appearance of the NF-1 ostial renal artery stenoses consistently revealed intimal fibroplasia, disruption of the internal elastic lamina and medial discontinuity (Fig. 4). The histologic character of arteries from four probands, including three probands with known truncating NF1 variants, included intimal hyperplasia and focal medial fibrosis as predominant features (Table 5). No Wagner-Meissner bodies were identified.

Figure 4.

Histopathologic appearance of a renal artery from a 3-year-old with NF-1, an abdominal aortic coarctation and bilateral renal artery stenoses, demonstrating concentric intimal fibrodysplasia, disruption and duplication of the internal elastic lamina, and moderate fibrodysplasia of scant media (Masson stain, original magnification ×10).

Table 5.

Available histopathology of NF-1 renal arteries

Proband #	IH Prominent	Medial irregularity	Outer diameter/luminal diameter (mm)	Other complication/comments
1	Yes	Yes	1.5/0.8	Focal fibrointimal hyperplasia with disrupted internal elastic lamina, focal medial fibrosis and disorganized/lost muscle fibers
2	Yes	–	3.2/1.8a	Prominent fibrointimal hyperplasia/fibrosis with disrupted internal elastic lamina, scarring suture granulomas, previous surgery
4	Yes	Yes	2.1/0.7	Not well oriented; prominent fibrointimal hyperplasia/fibrosis with disruption of internal elastic lamina and focal mild medial fibrosis
8	Yes	Yes	1.7/0.3	Marked fibrointimal hyperplasia with disrupted internal elastic lamina focal duplication and focal medial fibrosis

IH = intimal hyperplasia

^aArchitecture disrupted from prior intervention

Transcript analysis of arterial lesions

NF1 gene expression from resected dysplastic arterial tissue was significantly decreased in NF-1 children with RVH, when compared to donor controls. Five cases with operative renal tissue available (probands 1, 3, 4, 8, and 13) had a global reduction of NF1, evaluated by the transcripts per millions reads (TPM, fold change: 5, t-test P-value: 0.002), including Proband 3 whose DNA sample failed exome sequencing (Fig. 5A). Additional analysis on each exon supported the same conclusion for all five samples (Supplementary Material, Fig. S2A) and individual level of Proband 13 with a stop-gain at exon 26 (Supplementary Material, Fig. S2B). Sanger sequencing was used to validate the SNV for Proband 3, annotated as a truncating (stop-gain) variant (Table 4). We did not identify any additional NF1 variants by RNA-Seq analysis.

Figure 5.

RNA-Seq analysis of pediatric NF-1 renal arteries with RAS compared to control renal/SMA arteries. (A) NF1 gene expression is significantly reduced in NF-1 renal artery samples compared to controls. (B) Volcano plot demonstrating differential gene expression (top 30 shown, any genes with fold change absolute value >10 and adjusted P-value ≤2 × 10⁻⁸). Fold change lower than −10 was included as fold-change −10, with MCP1 expression increased in NF-1 renal arteries as compared to control arteries (C); gene set enrichment analysis of GO gene list showed a significant upregulation of MAPK-related pathways (FDR < 0.1), with the size of gene list indicated by dot size and the normalized enrichment score (NES) indicated by color. A positive NES corresponds to increased expression in NF-1 cases as compared to controls. The top significant signals with both lowest FDR and largest size include regulation of ‘MAPK cascade’ with the subgroup ‘positive regulation of MAPK cascade’, suggesting upregulation of the MAPK cascade in NF-1 renal artery tissues. (FDR = false discovery rate).

Globally, several genes were observed to be differentially expressed in NF-1 renal artery tissue (Fig. 5B). CCL2, also known as MCP1, an inflammatory marker for tissue injury, was significantly upregulated in NF-1 renal arteries as compared to control arteries (Fig. 5B). Gene Set Enrichment Analysis (GSEA) of Hallmark, Curated pathway, and Gene Ontology (GO) gene sets identified enriched pathways with FDR < 0.1. Specifically, all MAPK-related pathways were queried as NF1 is a key component of the Ras-MAPK pathway. Three Curated pathways (Supplementary Material, Fig. S3) and four Gene Ontology gene sets (Fig. 5C) were significantly upregulated in the NF-1 renal artery tissues. Both the increased level of inflammation and upregulation of the MAPK pathway may suggest the downstream effect of observed NF1 variants in renal artery tissues, particularly of intimal hyperplasia.

Validation of RNA-Seq MAPK/Erk1 pathway activation

To evaluate activation of the Ras-MAPK pathway and previously described Erk1 activity, phosphorylated Erk1 (pErk) expression was assessed by immunohistochemistry on the renal artery sample obtained from Proband 13, as this individual had a more severe clinical phenotype and the lowest NF1 transcript expression in the RNA-Seq analysis. Histological findings in Proband 13 included moderate intimal and prominent adventitial thickening as well as intimal thickening (Fig. 6A). Cytoplasmic and nuclear immunolabeling for pERK was ~4–20-fold higher in the media and intima of this sample in comparison to control arteries (Fig. 6B, D and E).

Figure 6.

Immunohistochemistry analysis of phosphorylated Erk1. Representative images showing arterial dysplasia with markedly thickened intima and adventitia in hematoxylin and eosin-stained renal artery of Proband #13 (A) in comparison to Control 1 (C). pERK1 immunolabeled cells were digitally quantified in annotated media and intima (red outlines, B, D) and found to be ~4–20-fold higher in Proband #13 in comparison to four pediatric control arteries (E). Scale bars 100 μm.

Discussion

NF-1 is a common cause of pediatric RVH. The genetic basis of vascular involvement in NF-1 has not been previously evaluated. At the University of Michigan, nearly 20% of pediatric patients managed surgically for RVH secondary to renal artery occlusive disease and AAC resulting in the MAS have been reported to carry a diagnosis of NF-1. MAS is a clinical term representing heterogeneous clinical findings that involve narrowings of the abdominal aorta and often, but not always, narrowings of the renal as well as splanchnic arteries. While a vasculopathy (inclusive of Moya-Moya) has been identified in up to 18% of individuals with NF-1 in other series, the incidence of NF-1-associated renal artery stenosis is estimated at 2% (5,6,14). As such, the current series of 13 patients with NF-1 and concurrent RVH, with or without AAC, represents a relatively large cohort with a severe NF-1 arteriopathy phenotype. Transcript analysis of arterial tissues from NF-1 patients and control arteries validated proposed mechanisms of NF-1 arteriopathy involving reduced NF1 mRNA, MAPK pathways activation and increased pro-inflammatory MCP-1. Analysis of mRNA expression in the dysplastic arterial stenoses demonstrated reduced NF1 transcript expression, confirming heterozygosity as the mechanism within the lesion. The study validated activation of the MAPK signaling pathway and increased MCP-1 expression, which have been previously implicated in model systems of NF1 deficiency (15–20).

Exome sequencing or Sanger sequencing in the present study confirmed pathogenic variants in 11 of the 13 sequenced children with NF-1-associated RVH. These patients often presented early in life, with a more severe phenotype (i.e.: with LVH and mesenteric stenoses). There was no significant gender predilection and, in line with existing data, approximately 50% of mutations were de novo. Clinical NF-1 features of this study’s children were similarly distributed to well-described classical features of NF-1, with the exception that those with RVH demonstrated a 2-fold higher incidence of optic glioma (30%) than that reported for generalized NF-1 patients (15% by age 6) (14). Whether the rate of detection of asymptomatic optic gliomas may be elevated due to increased screening in patients with diagnosed vascular disease warrants further study in larger samples.

While there has been no proposed Mendelian form of pediatric RVH, exome sequencing of 36 patients with hypertension and MAS, from 35 families, was recently reported to reveal a monogenic cause of disease in 43% of cases (Supplementary Material, Table S5) (21). Warejko et al. reported 6 NF1 variants in this series and noted that in five of seven individuals (four of six families) with NF1 pathogenic variants, other syndromic features of NF-1 were also present at the time of diagnosis of MAS. Viering et al. similarly performed exome sequencing on 37 unrelated pediatric patients with RVH, including 6 patients with a clinical diagnosis of NF-1, and identified a monogenic cause in 5 children (13%), including 3 patients harboring pathogenic NF1 variants (22). In addition to confirming pathogenic NF1 variants across two other reports, additional proposed causal variants that were predicted to be deleterious based upon in silico analysis in genes associated with vasculopathy reported by both studies included RNF213 (associated with Moya-Moya disease), GATA6 (associated with cardiac outflow tract obstruction), ELN (associated with Williams syndrome, pulmonary artery stenosis and supra-valvular aortic stenosis), JAG1 (associated with Alagille syndrome), GLA (associated with Fabry’s disease) and SMAD6 (associated with bicuspid aortic valve, supra-valvular aortic stenosis and aortic aneurysms). Whether the non-NF1 genetic findings of this study may act as modifiers of NF-1-associated disease was not explored.

Individuals in the current study with NF-1 did not harbor additional non-NF1 pathogenic variants in these genes. Histopathology of stenotic ostial lesions from operative renal artery samples from patients with NF-1 consistently identified intimal hyperplasia, medial thinning and disruption of the elastic lamina, similar to the observations reported in stenotic renal arteries from non-NF-1 pediatric patients with RVH (23). Intra-renal dysplastic arterial tissue was not available for analysis. Wagner-Meissner bodies have been described in NF-1 arteriopathy but were not observed in the current study.

NF-1 is an autosomal dominant genetic disorder caused by pathogenic variants in the tumor suppressor gene NF1. (24) It is characterized by extreme clinical variability and the gene is large, spanning ~350 kb of genomic DNA containing 60 exons (25,26). The protein product of NF1 is neurofibromin, a guanosine triphosphatase (GTP)-activating tumor suppressor protein that negatively regulates Ras activity (15,27). Ras is an essential cellular hub for a wide variety of survival, proliferation, differentiation and senescence signaling pathways (28). NF1 pathogenic variants encode dys- or nonfunctional neurofibromin molecules and accelerate the hydrolysis of active Ras-GTP to its inactive diphosphate conformation to disinhibit/accelerate proliferative Ras signaling (29,30).

It is hypothesized that vascular smooth muscle cells (VSMCs) that have lost NF1 exhibit enhanced neointimal formation following injury, which may account for the development of stenoses (15). Medium and smaller arteries of NF-1 patients display hyperplastic VSMCs within the intimal layer (31,32). Loss-of-function experiments confirm this effect as well. Aortic VSMCs explanted from mice modeled with partial or complete loss of neurofibromin proliferate more rapidly than wild-type littermates (15,16,33,34). In vivo, intact carotid arteries from these mice respond to injury with VSMC hyperproliferation of neointimal lesions. Neurofibromin-deficient myeloid cells demonstrate preferential activation of the Ras–Erk signaling pathway in response to in vitro growth factors (17,18,33).

Altered Ras–Erk signaling and inflammation have been implicated in NF-1 arteriopathy. The analysis of transcript expression in renal arterial stenotic lesions showed approximately 5-fold reduction in NF1 mRNA. Validating the findings of murine models of Nf1 deficiency, increased MAPK pathway transcripts consistent with activation of this pathway and increased MCP-1 expression were observed, as well. Lineage-specific inactivation of a single Nf1 gene copy in monocytes/macrophages has been shown to sufficiently reproduce neointima formation in Nf1 heterozygous mice compared with wild type (35). In this same murine model of Nf1 arterial stenosis, Ras–Mek–Erk signaling has recently been shown to regulate neurofibromin-deficient and wild-type macrophage function, and Ras–Erk inhibition reduces neointimal formation in NF1 heterozygous mice (19).

Moreover, asymptomatic individuals with NF-1 have increased circulating pro-inflammatory cytokines and monocytes (CD14⁺CD16⁺) in the blood compared with controls (17). Similarly, Nf1^+/− mice have increased circulating Ly6C^hiCCR2⁺ monocytes, the murine correlate of human pro-inflammatory monocytes (35). CCR2+ receptor activation by its primary ligand monocyte chemotactic protein-1 (MCP-1) is critical for monocyte infiltration into the arterial wall and neointima formation in Nf1^+/− mice. MCP-1 induces a dose-responsive increase in Nf1^+/− macrophage migration and proliferation that corresponds with activation of multiple Ras kinases; additionally, NF1^+/− VSMCs, which express CCR2, demonstrate an enhanced proliferative response to MCP-1 when compared to WT VSMCs making MCP-1 a potent chemokine for NF1^+/− monocytes/macrophages and CCR2 as a viable therapeutic target for NF-1 arterial stenosis (20). Notably, the human arterial transcript data in the current study validated these findings.

Recent studies using Nf1 genetically engineered mouse models of specific genetic variants have highlighted the various pathways (i.e.: modifier genes) by which these factors influence disease pathogenesis and progression (36). Nf1 murine models of optic glioma have revealed the importance of the cell of origin and the tumor microenvironment in cancer development and progression; for example, mouse optic glioma is maintained by immune system-like cells (microglia) and an analogous paracrine circuit (37). Germline variation may in part contribute to NF-1 disease heterogeneity, and while germline NF1 variants may characterize the disorder, tumor formation requires somatic inactivation of the second normal NF1 allele (especially important as stromal cells and neurons do not exhibit bi-allelic NF1 loss) (36). In the setting of optic glioma as well, studies of the combination of germline and somatic genetic variants have supported pathogenicity (38,39). Whether such mechanisms may be implicated in NF-1-related vascular disease requires study and is warranted based upon the findings of early genotype–phenotype and phenotype–phenotype correlations of NF-1 vasculopathy. In vivo murine models of vasculopathy extending prior studies of global gene deletions to specific genetic variants that may be introduced by CRISPR/Cas9 as well as patient-derived primary cell lines or stem cells with germline NF1 variants may have utility to not only further define clinically relevant patient subgroups but also conduct preclinical studies of potential targeted therapies against vascular lesion formation (36).

Open surgical treatment of both renal artery stenoses and AAC at a high-volume referral center provides sustainable hypertension benefits in nearly 90% of children (6). Nevertheless, interventions in the very young (<3 years) and the presence of concurrent aortic disease increase the likelihood of a later reoperation. Moreover, those undergoing secondary procedures are less likely to be cured of hypertension. Additionally, postoperative hypertension benefits were markedly limited among patients with NF-1 and concurrent AAC requiring treatment (16% cure, 58% improved, 26% no benefit) when compared to the total cohort (44%, 46% and 10%, respectively). Established criteria for the diagnosis and management of pediatric RVH must consider specific patient phenotypes, and perhaps genotype, to define the most appropriate interventions and identify best practices, especially when balancing the risks of open operative versus endovascular interventions. NF-1 patients warrant careful patient selection for revascularization, and families intentionally counseled regarding post-surgical outcomes and the need for judicious surveillance and possible reoperation.

The current study is limited by the small sample size, limited follow-up for some patients (particularly for international referrals) and non-standardized clinical assessment of NF-1 features. Classically, complete sequencing coverage of the NF1 gene has been challenged by its large size, lack of mutation hotspots and somatic alterations present in tumors (40). However, in the present study’s exome sequencing, all coding regions were sequenced, with 99.5% of coding regions covered by at least 10×. Dedicated studies of somatic variation in the NF1 gene will be important to define whether loss of heterozygosity is a relevant mechanism and will be pursued in the future. Taken together, the findings support the need for further investigations and scientific evidence to base future interventional studies, including clinical trials.

Materials and Methods

Study subjects

Consecutive pediatric patients presenting with RVH secondary to RAS, with and without concurrent AAC, were prospectively enrolled between January 2013 and July 2020 into the University of Michigan Genetic Study of Arterial Dysplasia. This IRB-approved study (HUM00044507) was established to enroll patients with arterial dysplasia phenotypes and unaffected controls for genomic investigations. The phenotype of RAS and AAC contributing to the MAS was confirmed by reviewing the children’s medical records as well as arterial images obtained by digital subtraction angiography, computed tomography or magnetic resonance imaging. Blood, saliva and surgical vascular specimens, when available, were collected from each child. Family histories were reviewed for genetic disorders and vascular diseases, and clinical three-generation pedigrees were constructed. The histologic character of the operative renal artery specimens was documented, as was that of age-matched control arteries.

DNA preparation and whole exome sequencing

Saliva samples were collected using Oragene (DNAGenotek), and genomic DNA was isolated using prepIT L2P (DNAGenotek, Ontario, Canada). Genomic DNA was isolated from blood and tissue samples using the Nucleopsin Tissue Kit (Takara Bio, CA). DNA was assayed with Quant-iT PicoGreen dsDNA Assay kit (ThermoFisher, NY). Genomic DNA samples were used for exome sequencing library construction and exome capture was automated (Perkin-Elmer Janus II) in a 96-well plate format. Exome sequencing was performed in two batches, with the first batch sequenced at the Northwest Genomics Center (University of Washington, Seattle, WA) and the second batch sequenced by Psomagen Inc. (Rockville, MD).

After quality control, genomic DNA was subjected to a series of shotgun library construction steps, including tagmentation (in Psomagen, or fragmentation and ligation in NWGC), PCR amplification and gel purification. Libraries underwent exome capture using either the Roche/NimbleGen SeqCap EZ v2.0 (~36.5 Mb target) or Agilent SureSelect V5 (~50.4 Mb target). All samples were pair-end sequenced on Hiseq 4000 or Nova 6000, with a read length of 75 bp or 100 bp, respectively. The mean read depth was 64× for all NF-1 samples on target regions (median: 67, range: 47–82).

Exome sequencing alignment, variant calling, quality control and annotation

All data were aligned to the reference genome GRCh37 (hg19) with bwa 0.7.17 mem. The GATK (4.0.5.1) SNV calling best practices were followed (https://gatk.broadinstitute.org/hc/en-us/articles/360035535932-Germline-short-variant-discovery-SNPs-Indels-), having two parameters modified to accommodate computational resources (—max_records_in_ram and —max-reads-per-alignment-start) (14–16). Finally, all high-quality variants were preserved if 1) the quality score was high (QD > 5.0, QUAL >50.0) and 2) there were no SNV cluster within 10 bp (—cluster-size 3, —cluster-window-size 10). All variants were annotated by Annovar with the following databases: gnomAD population frequency (v2), dbNSFP (v3.5), CADD score (v1.3) and ClinVar database (accessed 3 June 2018) (17–21). All variants were predicted to be deleterious (i.e.: disease-causing) by in silico analysis if meeting any of the following criteria: (1) nonsynonymous variant with phred CADD score > 20 and gnomAD allele frequency ≤ 0.01%; (2) variants considered to have evidence of pathogenicity according to predicted loss of function (start-loss, stop-loss, stop-gain, frameshift indels or splice-site) or non-frameshift indels, with gnomAD allele frequency ≤ 0.01%; or (3) defined as likely pathogenic or pathogenic variants in ClinVar.

Annotation of sequence variants

Variants in the NF1 gene were compared against a clinical diagnosis of NF-1 and other clinical and angiographic features in 11 children having exome sequencing or RNA-Seq data. Variant location and predicted effect, including exon structure of NF1 cDNA, and domain structure of the encoded protein neurofibromin, were annotated. Non-synonymous variants underlying monogenic arteriopathies including ACTA2, COL3A1, COL5A1, COL5A2, ELN, FBN1, TGFBR1, TGFBR2, TGFB2, TGFB3, SMAD2, SMAD3, SMAD4, SMAD6, PRKG1 and LOX (Supplementary Material, Table S1) were annotated according to gnomAD allele frequency ≤ 0.01% and phred CADD score > 20, and then reviewed. Additional genes of potential relevance in children with RVH were similarly reviewed, including ELN, JAG1, RNF213, TSC1 and TSC2 (22,23).

RNA-Seq library preparation and sequencing

To assess global and specific transcriptional changes associated with NF-1 arteriopathy, RNA-Seq was performed on renal arteries from five pediatric NF-1 cases of RVH, renal or superior mesenteric arteries from four control individuals having no vascular disease, along with other biorepository arterial specimens. All libraries were prepared with 25 ng of RNA using KapaHyperErase (HMR) kit with the rRNA depleted. After quality control, all samples were pair-end sequenced by NovaSeq 6000 in MedGenome (Foster City, CA), with a read length of 100 bp. Each sample acquired ~13 Gb data on average (median: 10 Gb, range: 8–26 Gb).

RNA-Seq analysis of differential gene expression

All data were aligned to reference genome GRCh37 (hg19) by STAR2, with the isoform annotation from genecode (V19) (24,25). Raw counts were extracted and then normalized with DESeq2 (26). After quality assessment, analyses were performed in DESeq2 to determine differential transcript expression between case and control groups. Gene Set Enrichment Analysis (GSEA) was performed using Hallmark, Gene Ontology, and Curated Pathways set files from the Molecular Signatures (27). The significant threshold is defined as FDR (false discovery rate) < 0.1.

A normalized read count ratio was used to evaluate the exon-wise difference between NF-1 cases and controls. The number of uniquely mapped reads were first extracted through samtools (28). Then they were normalized by the total number of reads per sample and divided by the median normalized read count of controls.

Variant calling from RNA-Seq data

SNVs were called from the RNA-Seq data followed the GATK (4.0.5.1) best practice for RNA-Seq variant calling (https://gatk.broadinstitute.org/hc/en-us/articles/360035531192-RNAseq-short-variant-discovery-SNPs-Indels-). Low-quality variants were removed if matching any of the following criteria: (1) the quality score was low (QD < 5.0, QUAL <50.0); (2) the indel size was longer than 3 bp; or (3) from non-protein coding region, due to lower coverage.

Sanger sequencing validation

Sanger sequencing validation was performed of RNA-Seq-detected SNVs. While Proband 3 failed quality control for exome sequencing of the small volume of renal artery tissue available, operative tissue had a global reduction of NF1 by RNA-Seq. An NF1 variant was validated by Sanger sequencing of operative renal parenchymal tissue in the absence of available blood or saliva. About 10 ng of genomic DNA was used to amplify a 440 bp amplicon in exon 36 of the NF1 gene. The following primers were used: Forward 5′-CTTCTCTTAGCCTTATTTCTCAGTGTCC, Reverse 5′-GCAGCCGCTCATGATACTTGG. Amplicons were isolated and sent for sequencing (Genewiz, South Plainfield, NJ).

Immunohistochemistry

Histology and immunohistochemistry slides were prepared by routine methods. The primary antibody (rabbit anti-human phosphorylated p44/42 MAPK[Erk1/2], Cell Signaling Technology #4370, Danvers, MA) was applied at a titration of 1:300 for 30 min on an automated immunohistochemistry stainer after low pH heat-induced antigen retrieval and labeling was detected with Agilent Dako EnVision FLEX system. Slides were digitized on a Leica Aperio AT2 digital slide scanner (Leica Biosystems) at magnification up to 20× (0.50 μm/pixel). Immunohistochemistry quantitation was performed using the open-source program QuPath v0.2.1 (https://github.com/qupath/qupath) (29). Digitized slides were positively annotated to select intima and media for analysis and visible artifacts or confounding areas (folds, debris) were excluded by negative annotation. Slides with inadequate material or orientation for evaluation were excluded from analysis. RGB values for DAB and hematoxylin staining were set directly from singly-stained areas in the slides and a pixel classifier was trained to detect pERK-labeled pixels and quantify them as objects (cells). Digital image overlays were visually checked for accuracy. The number of labeled cells for each marker was reported out as cells * 10⁴/area tissue (μm²) where tissue represented the total intima and media.

Web Resources
GSEA database	Gene Set Enrichment Analysis	https://www.gsea-msigdb.org/gsea/msigdb
ClinVar database	National Center for Biotechnology Information (NCBI)	https://www.ncbi.nlm.nih.gov/clinvar
gnomAD database	The Genome Aggregation Database (gnomAD)	http://gnomad.broadinstitute.org
LOVD database	Leiden Open Variation Database	https://www.lovd.nl/
CADD	Combined Annotation Dependent Depletion	https://cadd.gs.washington.edu
UMD	UMD-predictor	http://umd-predictor.eu/
SIFT	Sorting Intolerant From Tolerant	http://sift.bii.a-star.edu.sg
Polyphen2	Polymorphism Phenotyping v2	http://genetics.bwh.harvard.edu/pph2/
ACMG/ACGS 2020	ACGS Best Practice Guidelines for variant classification in rare disease 2020	https://www.acgs.uk.com/media/11631/uk-practice-guidelines-for-variant-classification-v4-01-2020.pdf

Data Availability

RNA-Seq data will be deposited into GEO.

Funding

National Heart, Lung, and Blood Institute (NHLBI), Department of Defense, Frankel Cardiovascular Center and The University of Michigan A. Alfred Taubman Institute. S.K.G., D.C., Y.W., and K.H. were supported by NHLBI R01HL139672, University of Michigan A. Alfred Taubman Institute, University of Michigan Frankel Cardiovascular Center, and Department of Defense #NF190071. S.K.G. is also supported by NHLBI R01HL122684, R01HL086694. Y.W. was supported by NHLBI T32-HL007853.

Author information

Conceptualization: D.C., S.G.; Data curation: Y.W., D.C., J.S., S.G.; Formal Analysis: Y.W., M.Y., I.B., S.G.; Funding acquisition: S.G.; Investigation: K.H., I.B., S.G.; Methodology: Y.W., M.Y., S.G.; Project administration: D.C., S.G.; Resources: I.B., S.G.; Visualization: D.C., I.B.; Y.W., S.G.; Writing—original draft: D.C., Y.W., S.G.; Writing—review & editing: D.C., Y.W., J.S., S.G.

Ethics Declaration

This University of Michigan Genetic Study of Arterial Dysplasia IRB-approved study (HUM00044507) was established to enroll patients with arterial dysplasia phenotypes and unaffected controls for genomic investigations. Informed consent was obtained from all participants as required by the IRB, with assent as appropriate. Clinical data were de-identified, and written consent has been obtained and archived for participation/publication from every individual whose data are included in this manuscript. This study adheres to the principles set out in the Declaration of Helsinki.