Clin Genet 2015: 87: 401–410Printed in Singapore. All rights reserved

© 2014 John Wiley & Sons A/S.Published by John Wiley & Sons Ltd

CLINICAL GENETICSdoi: 10.1111/cge.12498

Review

The natural history of spinal neurofibromatosis:a critical review of clinical and genetic features

Ruggieri M, Polizzi A, Spalice A, Salpietro V, Caltabiano R, D’Orazi V,Pavone P, Pirrone C, Magro G, Platania N, Cavallaro S, Muglia M, Nicita F.The natural history of spinal neurofibromatosis: a critical review of clinicaland genetic features.Clin Genet 2015: 87: 401–410. © John Wiley & Sons A/S. Published byJohn Wiley & Sons Ltd, 2014

Spinal neurofibromatosis (SNF) is a related form of neurofibromatosis 1(NF1), characterized by bilateral neurofibromas (histologically proven) of allspinal roots (and, eventually, of all the major peripheral nerve branches) withor without other manifestations of classical NF1. By rigorous application ofthese criteria to the 98 SNF cases published, we developed: (i) a cohort of 49SNF patients (21 males and 28 females; aged 4–74 years]: 9 SNF families(21/49), 1 mixed SNF/NF1 family (1/49) and 27 of 49 sporadic SNF patients(including 5 unpublished patients in this report); and (ii) a group of 49non-SNF patients including: (a) 32 patients with neurofibromas of multiplebut not all spinal roots (MNFSR): 4 mixed SNF/MNFSR families (6/32); (b)14 patients with NF1 manifestations without spinal neurofibromas,belonging to SNF (8/49) or MNFSR families (6/32); (c) 3 patients withneurofibromas in one spinal root. In addition to reduced incidence ofcafé-au-lait spots (67% in SNF vs 56% in MNFSR), other NF1manifestations were less frequent in either cohort. Molecular testing showedcommon NF1 gene abnormalities in both groups. The risk of developingSNF vs NF1 was increased for missense mutations [p= 0.0001; odds ratio(OR)= 6.16; confidence interval (CI)= 3.14–13.11], which were morefrequent in SNF vs MNFSR (p= 0.0271).

Conflict of interest

The authors declare no conflict of interest

M. Ruggieria,b, A. Polizzic,d,A. Spalicee, V. Salpietrof,R. Caltabianog, V. D’Orazih,P. Pavonei, C. Pirronej,G. Magrog, N. Plataniak,S. Cavallaroc, M. Muglial

and F. Nicitae

aDepartment of Educational Sciences,Chair of Paediatrics , bCentre forNeurocutaneous Disorders, University ofCatania, Catania, Italy, cInstitute ofNeurological Sciences, National ResearchCouncil, Catania, Italy, dNational Centrefor Rare Diseases, Istituto Superiore diSanità, Rome, Italy, eChild NeurologyDivision, Department of Paediatrics,Sapienza University of Rome, Rome, Italy,fUnit of Genetics and PediatricImmunology, Department of Pediatrics,University of Messina, Messina, Italy,gDepartment ‘G. Ingrassia’, Section ofAnatomic Pathology, University ofCatania, Catania, Italy, hDepartment ofSurgical Sciences, “Sapienza” Universityof Rome, Rome, Italy and Unit ofMicrosurgery and Hand Surgery, FabiaMater Hospital, Rome, Italy, iUnit ofPaediatrics and Paediatric Emergency,University Hospital ‘Policlinico-VittorioEmanuele’, Catania, Italy, jBranches ofPsychology, Department of EducationalSciences, University of Catania, Catania,Italy , kInstitute of Neurosurgery,Department of Neurosciences ‘G.Ingrassia’, University of Catania, Catania,Italy, and lInstitute of NeurologicalSciences, National Research Council,Cosenza, ItalyKey words: familial spinalneurofibromatosis – multipleneurofibromas – NF1 – paraspinalneurofibromatosis – spinalneurofibromatosis

Corresponding author: MartinoRuggieri, BA, MD, PhD, Department ofEducational Sciences, Chair ofPaediatrics, University of Catania, ViaCasa Nutrizione, 1, 95124 Catania, Italy.Tel.: +39 095 7466377;fax: +39 095 7466385;e-mail: m.ruggieri@unict.it

401

Ruggieri et al.

Received 10 August 2013, revised andaccepted for publication 8 September2014

Definition and historical background

Spinal neurofibromatosis (SNF) is a distinct clini-cal entity – a related form of neurofibromatosis 1(NF1) – characterized by bilateral, histologicallyproven, neurofibromas of all spinal roots leaving nointact segments, with or without other manifestations ofthe NF1 diagnostic criteria (1, 2).

This form was first recognized in the literature by Pulstet al. in 1991, in eight members of two multi-generationfamilies and named ‘familial spinal neurofibromatosis’(1). A few years later, in 1997, Poyhonen et al. (2)reported a family in which seven members over threegenerations were affected with a rare SNF (‘… a dis-tinct variant of NF1… ’), named ‘hereditary spinal neu-rofibromatosis’. Interestingly, while affected members ofthe two families of Pulst et al. (1) had ‘ . . . . multipleparaspinal neurofibromas on both sides at all levels ofmultiple (but not all) segments of the spine’, all the (four)affected adults of the family of Pohyonen et al. (2) had‘… multiple spinal neurofibromas symmetrically affect-ing all spinal roots’. Both studies noted, for the first time[irrespective of full (2) or partial (1) spinal root involve-ment], that these phenotypes were (1, 2): (i) differentfrom the well-established NF1 and NF2 phenotypes, (ii)distinct, in that the predominant feature was multiple(extensive) spinal neurofibromas, and (iii) much morerestricted in their manifestations with absence of Lischnodules and of non-tumour features of NF1 (e.g. learn-ing difficulties and skeletal dysplasia). At the time of thestudies (1, 2), linkage with the NF1 gene was recordedin family 1 of Pulst et al. (odds of 97:1) (1) and in thefamily of Pohyonen et al. (LOD score of 1.763) (2).

After these two reports (1, 2), 17 further studiesrecorded 83 additional (both familiar sporadic) patientsfor a total of 98 patients labelled under the umbrellaterm of ‘familial/hereditary spinal neurofibromatosis’ orjust ‘spinal neurofibromatosis’ (3–20). The conditionhas been regarded as a true NF1 variant or subtype, andeither linkage to the NF1 locus (1, 2) or mutations in theNF1 gene have been recorded in most reported cases (3,5–11, 13, 16–20).

In addition to its hallmark of multiple neurofibromas ofthe spinal roots and, major nerve trunks, the phenotypevariably included classical NF1 stigmata (e.g. typicalpigmentary manifestations, Lisch nodules and cutaneousand nodular neurofibromas) and other NF1-associatedmanifestations (e.g. scoliosis, long bone dysplasia, opticand non-optic pathway gliomas, plexiform neurofibro-mas (PNFs), pheochromocytoma and the typical T2hyperintensities in the brain) (3–20). Besides the classi-cal families with multiple spinal neurofibromas, severalsporadic cases have been recorded and, more recently,the phenotypic spectrum has included a set of concordanttwins with SNF (16) and individuals with SNF belongingto families with classical NF1 (20). Thus, it was proposed

to drop the term ‘familial (hereditary) spinal neurofibro-matosis’ and to replace it with ‘Spinal neurofibromato-sis’ (16, 19, 20) as the condition could simply reflect atypical autosomal dominant pattern of inheritance, there-fore one would expect a balance of familial (includingtwins and cases within classical NF1 families) vs spo-radic cases.

Rationale of review

By critically reviewing the published reports labelled as(familial/hereditary) spinal neurofibromatosis (1–20),we realized that not all the reported cases presentedwith involvement of all the 38 spinal roots. In manyindividuals, the affected status of SNF was assignedto (familial or sporadic) patients with few or multipleneurofibromas involving multiple (but not all) segmentsof the spine (either unilaterally or bilaterally). In addi-tion, some families included members with classicalNF1 and other members with multiple neurofibromasinvolving all the spinal roots. Other families includedsome members with involvement of all spinal roots andother members with involvement of multiple but not allspinal roots, irrespective of age. This implies that forsome individuals gleaned from these reports, spinal rootneurofibromas could reflect more severe of a manifes-tation of classical NF1 rather than a related or alternateform of NF1. In some other families involvement of allvs multiple but not all spinal roots could merely reflectthe phenotypic variability within the spectrum of SNF.

Aim and methods of review

Aiming to better define the phenotype of SNF wesearched existing databases (i.e. Medline, Scopus andWeb of Knowledge) for reports on individuals labelledunder the terms ‘familial spinal neurofibromatosis’,‘hereditary spinal neurofibromatosis’ or ‘spinal neurofi-bromatosis’.

We identified 20 reports (1–20), which included a totalof 98 patients with SNF.

We rigorously applied the criteria, which define SNF[i.e. the presence of multiple, histologically proven, neu-rofibromas of all spinal roots, bilaterally (and, eventu-ally, of all the major peripheral nerve branches)], to eachof these 98 individuals. Assignment of affected statusof SNF, in each of the 98 reported patients, was basedon complete phenotype assessment (including skin, oph-thalmologic and orthopaedic examination and cognitiveprofile) and results of brain and spinal magnetic reso-nance imaging (MRI) studies and biopsies of at least onelesion of the peripheral nervous system.

We added to this cohort complete phenotypes andgenotype in five unpublished cases followed at ourinstitutions (Table 1).

402

Spinal neurofibromatosis

Table 1. Clinical, laboratory and imaging features of five (unreported) patients affected by spinal neurofibromatosis seen andfollowed-up at our institutions

Patient 1 2 3 4 5

Gender F F M M MCurrent age (year) 74 66 55 44 33Age of onset (year) 33 29 22 25 22Familial (NF1 or SNF) − − − − −Sporadic + + + + +Growth parameters

(weight, length, HC)50th 50th 50th 50th 50th

Café-au-lait spotsNumber

>6 10 8 10<6 4 3NS NA NA NA

Colour Light-brown Light-brown Light-brown Light-brown Light-brownSize (cm) >2 >3 cm >2 >5 >4Margins Irregular Irregular Irregular Irregular Irregular

Freckling − − − − −Lisch nodules − − − − +Cutaneous NFs − − − − −Nodular NFs Limbs, trunk Limbs, trunk Limbs, trunk Limbs Trunk, limbMPNST − − − − +Scoliosis + − − − +Other NF1 stigmata − − − − −Brain MRI

NBO − − − − −OPG − − − − −

Spinal MRIParaspinal NFs + B (all roots) + B (all roots) + B (all roots) + B (all roots) + B (all roots)Thoracic NFs + + + + +

Body MRINodular NFs Limbs, trunk Limbs, trunk Limbs, trunk Limbs Trunk, limb

Neurology NeuropathyLL

NeuropathyLL, abdomen

NeuropathyUL

Pain, neuropathyTrunk, limbs

Pain, paresisLL

Cognitive profile IQ= 95 IQ=88 IQ= 89 IQ=90 IQ= 95Molecular testingNF1 gene Exon 21

p.(Gly848Glu)Intron 42c.6427+2T>G

Exon 12a1733T>G

Exon 376757delG

Pending

B, bilateral; F, female; HC, head circumference; IQ, intelligence quotient; LL, lower limbs; M, male; MPNST, malignant peripheral nervesheet tumour; MRI, magnetic resonance imaging; NBO, neurofibromatosis bright objects (high signal lesions on T2-weighted MRI);NF1, neurofibromatosis type 1; NFs, neurofibromas; NA, not applicable; NS, not specified; OPG, optic pathway glioma; UL, upperlimbs.

Results of review

Clinical findings and natural history

We assigned affected status of SNF to 49 patients [21males and 28 females (male:female ratio= 0.72)] aged,at their last follow-up, 4–74 years (Tables S1 and S2,Supporting information). The main clinical and imagingaspects of SNF are illustrated in Figs 1–4.

We identified a second cohort, which included 32patients (22 males and 8 females; aged 7–74 years) withmultiple neurofibromas in multiple ‘but not all’ spinalroots (MNFSR) (Table S3).

An additional group of 14 patients belonging either toSNF families (8/49; Table S1) or to MNFSR families(6/32; Table S3) were identified who had NF1 stigmatawithout neurofibromas in spinal root.

Three out of the original 98 patients, labelled as havingSNF (1–20), were excluded from present analysis as theyhad involvement of only one spinal root (patients 406,894 and 548) (8).

We generated an additional table (Table S4) thatincluded data from the five unpublished patients herebyreported (as shown in Table 1) and follow-ups of the fourpublished patients (16, 19) seen at our institutions.

Molecular testing

Results of molecular NF1 gene testing, which wasperformed in 40 of 49 SNF patients and in 25 of 32MNFSR patients are summarized in Tables S1–S3 anddetailed in Tables S5 and S6. We generated an additionaltable (Table S7) detailing the phenotypic features in SNFpatients harbouring NF1 gene deletions.

403

Ruggieri et al.

Fig. 1. Close view-up of the pigmentary macules recorded in children with spinal neurofibromatosis; note the paler brownish colour and the irregularmargins of the macules [the figures show the pigmentary lesions seen in patient 1 (a), patient 2 (b), and patient 3 (c) as reported in Ruggieri et al. (16)].

Overall, molecular NF1 gene testing identified muta-tions in 37 of 40 (92.5%) SNF patients (including adouble mutation) and 24 of 25 (96%) MNFSR. Therewere whole gene deletions [4/37 (1 familial) and 2/24(1 familial)], missense [16/37 (5 familial) and 4/24 (2familial)], splicing [8/37 (3 familial) and 6/24 (3 famil-ial)], nonsense [4/37 (1 familial) and 3/24 (2 familial)]and frameshift mutations [5/37 (1 familial) and 9/24 (3familial)].

The risk of having SNF vs classical NF1 wasincreased in individuals harbouring missense mutations[p= 0.0001; odds ratio (OR)= 6.16; confidence inter-val (CI)= 3.14–13.11], and missense mutations weremore frequent in SNF vs MNFSR patients (p= 0.0271)(Appendix S1). Other types of mutations did not have asignificant association with any one phenotype.

Phenotypic spectrum and natural history

Hallmark of disease

The hallmark of SNF is bilateral neurofibromas of allspinal roots (Tables S1 and S2; Fig. 2a,b). This featurecharacterizes SNF as a related or alternate form of NF1,and not merely as a classical NF1 phenotype manifestingwith increased occurrence of spinal neurofibromas orwith a concentration of multiple neurofibromas in somesegments of the spine.

By critical literature review and by personal experiencewe could easily identify the first cohort of 49 patientswith this feature (Table S1; see also Tables 1 and S5) (16,19). Bilateral involvement of all spinal nerve roots most

clearly defined the SNF phenotype. Lack of rigorousapplication of this criterion has perpetuated phenotypedesignations that are not SNF (Table S3).

Relaxing the application of strict criteria requiringneurofibromas of all spinal nerves generated the secondcohort, and we identified 32 patients with multipleneurofibromas in multiple (but not all) spinal roots,which we named MNFSR (Table S3). Three individuals,in this cohort had massive involvement of almost allspinal roots (Table S3) (12) and six other individuals,with MNFSR, belonged to four single-generation ormulti-generation families with SNF (Tables S1 and S2)(3, 8, 10, 13): such phenotypic variability must be takeninto consideration and is likely part of the phenotypicspectrum of SNF. Individuals with MNFSR, unless partof a family with affected members having SNF, shouldnot be regarded as having SNF.

Neurofibromas, in SNF, can be localized in theparaspinal segments or may have either intradural orextradural components, or both (so-called dumbbelltumours), with a large extraspinal component causingforaminal enlargement. Histologically they prove to beclassical neurofibromas (Fig. 3a,b). The biopsy needsto have enough tissue of high quality to enable theclinical pathologist to make the determination of tumourtype, and there is no role of thin needle biopsy. If thediagnosis of NF1 or NF2 cannot be established then abiopsy is indicated to determine if the tumours wereneurofibromas or schwannomas. Schwannomas havebeen recorded, occasionally, along with neurofibromaswithin the SNF phenotype (Tables S1 and S2) (5).

404

Spinal neurofibromatosis

Fig. 2. Coronal (a) and sagittal (b, c) T2-weighted magnetic resonance images of the spine in an affected individual with spinal neurofibromatosisreveal multiple, bilateral neurofibromas of the spinal roots that leave no intact segment [patient III-1 in Nicita et al. (19)].

Fig. 3. Coronal (a) and axial (b) T2-weighted magnetic resonance images of the sacral spine and lower limbs showing the extension of the multipletumours along the lower nerve branches leaving no intact segment (patient 2, Table 1).

The MRI appearance of neurofibromas may start assimple nerve thickening (Table S1) or is that of anovoid lesions, which show slightly greater signal inten-sity than skeletal muscle on T1-weighted sequencesand high intensity periphery with variable intensitycentre on T2-weighted sequences (‘target sign’) (21,22). On computerized tomography (CT), the paraspinalneurofibromas have low attenuation when comparedwith the muscle. Intraspinal (intrathecal or extrathecal)

neurofibromas may displace the spinal cord or the nerveroots of the cauda equina to the contralateral side of thecanal. When bilateral neurofibromas are present, overtime, at a single level, the cord can be compressed intoa narrow, central band of tissue that is elongated in theanterior–posterior direction (22).

Neurofibromas, presenting as subcutaneous nodularlesions of one or more limbs, are frequent (Fig. 4): theselesions can involve, over time, the entire major (and/or

405

Ruggieri et al.

Fig. 4. Surgical field in the leg area showing a large mass affecting one of the major nerve branches (a) in a 14-year-old twin girl [case 2 in Ruggieri et al.(16)] with spinal neurofibromatosis; (b) histological analysis of the lesion removed in (a) reveals a myxoid matrix (black arrow) containing neoplasticcells and dense collagen bundles typical of the neurofibroma (hematoxylin and eosin staining, ×200).

less frequently the minor) nerve trunks, leaving no intactsegment [this is diagnostically important as differentiatesSNF from the mosaic forms of neurofibromatosis (23,24) and from schwannomatosis (25)]. Less frequently,neurofibromas can develop in or extend to the mediasti-nal region, the abdomen or the pelvis.

Associated involvement of all the major peripheralnerve branches in the four limbs could not always bepresent (Table S1): for this reason we preferred to leavesuch phenotypic aspect in brackets, in the definition ofSNF, and to add the word ‘eventually’. As stated before,however, we could record subcutaneous neurofibromas(relentlessly involving all the major peripheral nervebranches) in all the nine cases followed-up at our institu-tions (Tables 1 and S4) (16, 19).

Associated NF1-features

Skin manifestationsThe spectrum of skin manifestations in SNF may consistin café-au-lait spots and/or freckling in specific places,and cutaneous neurofibromas (Tables S1 and S2).

Pigmentary manifestations: café-au-lait spotsand freckling. The typical café-au-lait spots in SNFhave been recorded in 67% of the analysed cohort (vs56% in MNFSR); these were multiple (i.e. >6) in 60%(SNF) vs 83% (MNFSR). Freckling in specific places(mostly distributed over the axillae) was reported in18% (vs 16% in MNFSR). These data indicate thatcafé-au-lait spots in SNF can be less frequent andfreckling much less frequent than in classical NF1, thusgenerating difficulties in diagnosis unless the involve-ment of peripheral nervous system is apparent (17).We could not record detailed information regarding theshape and colour of café-au-lait spots in most publishedreports on SNF (Tables S1 and S2); the description of‘light-brown’ cafè-au-lait spots was recorded only oncepreviously (2) and only one illustration of a typical skinpigmentary lesion has been published previously (14).In our personal experience (16, 19) (Tables 1 and S4)the café-au-lait macules, within the context of SNF,were paler in colour, larger in size and with less regularmargins as compared with the typical NF1 pigmentary

lesions and were somewhat closer in size, shape, andcolour to the hyperpigmented macules typically encoun-tered in (children but also in adults) with NF2 (26, 27)(Fig. 1a,c). Thus, we proposed (16, 19, 20) that multiple(sometimes fewer than six but also in larger numbers),large, light-brownish café-au-lait spots with irregularmargins (as those shown in Fig. 1) seen in children,irrespective of their age, could be a marker of or couldraise the suspicion of SNF.

Cutaneous neurofibromas. Cutaneous neurofibromaswere recorded in 24% of SNF vs 37% of MNFSRpatients. Thus, whatever group we consider, these figuresare lower than classical NF1 where multiple dermal neu-rofibromas affect more than 90% of affected individuals.Notably, cutaneous neurofibromas were not seen in thegroup of SNF patients followed at our Institutions, irre-spective of their ages.

Lisch nodulesLisch nodules were seen in 24% of SNF patients and in33% of MNFSR patients, and if present most patientsin either cohort had few Lisch nodules. Of interest, werecorded iris nodules in one of nine patients seen at ourinstitutions (16, 19) (Tables 1 and S4). These figuresare low as compared with patients with classical NF1(approximately 90%) and could reflect a localized phe-notype, which is the hallmark of SNF. Lisch nodules areseen in less that 10% of patients with mosaic/segmentalNF1 (23, 24, 28, 29).

Other NF1-associated features and complicationsSkeletal involvement. Bone involvement (e.g. longbone dysplasia, vertebral or chest anomalies, cystic-likelesions in long bones), which is typically encountered inNF1 (30), has been reported only occasionally in SNF.Of interest, even though neurofibromas involve everyspinal nerve root, only 18% of SNF patients developedscoliosis, of mild to moderate degree (i.e. with a cur-vature of <10 in about 50% of cases). In our experi-ence (16,19; Tables 1 and S4), scoliosis, coupled withNF1-pigmentary manifestations was the only findingduring early childhood in some of the SNF patients

406

Spinal neurofibromatosis

followed at our institutions; only after years (i.e. aroundpuberty: between 8 and 10 years of age), the appearanceof spinal involvement was recorded [see for example,case 3 (16)]. Thus, we advise to obtain full spinal MRIscanning, to identify spinal nerve root involvement, afterthe appearance of Tanner stage I in every child withNF1-pigmentary manifestations coupled with other NF1stigmata.

Cognitive profile. Neuropsychiatric manifestations havebeen rarely reported in SNF: learning difficulties wereonly occasionally recorded (7%, which is below thepopulation background prevalence) and only one patienthad associated attention deficit hyperactivity disorder(ADHD) (9) (Table S1). Overall, in the SNF cohorthereby analysed the cognitive profile was normal (TablesS1 and S2). This is different from classical NF1 whosefigures for learning difficulties are usually higher than50%.

Tumours. Only few affected individuals developed opticpathway gliomas (OPGs, 10%), which is lower than inclassical NF1. Other low-grade gliomas were recordedin 4% and phaeochromocytoma in 2%, which is in linewith the prevalence figures of classical NF1. Of interest,none developed these tumours in MNFSR.

Plexiform neurofibromas. Only 6% of SNF patientsdeveloped classical PNFs (histologically proven PNF,not just neurofibromas encompassing multiple nerveroots or extending to major peripheral nerves) vs higherpercentages in classical NF1.

Malignant peripheral nerve sheet tumours. Malignantperipheral nerve sheet tumours (MPNSTs), arising ordeveloping de novo from peripheral nerves (not fromspinal nerve roots) have been recorded in 6% of SNFpatients (Tables S1 and S2). Notably, all the SNF patientswho developed these tumours died from complications ofMPNST.

Other features. Only few SNF patients manifested otherNF1-associated features, including large head [i.e. headcircumference either above the 97th percentile (17)or between the 75th and 97th percentile (18)], shortstature (unspecified) (9, 18) and Noonan-like features(unspecified) (17). This is in line with the other recordedfigures for NF1 manifestations in SNF: i.e. lower than inclassical NF1 and possibly reflecting the mosaic natureof the SNF phenotype.

Central nervous system involvementOverall, brain involvement has been observed in 24%of SNF patients including (Tables S1 and S2): (1) T2hyperintensities in the basal ganglia in 5 of 49 (10%)patients (2, 16, 18), 2 OPGs in 5 of 49 (10%) patients(4, 9, 14), and (3) non-optic pathway tumours in 3 of 49(6%) patients including localization in the frontal lobe(2), the cerebellum (9) and the cerebral peduncles (18).

We had previously emphasized, for the first time, onthe presence of bilateral, typical T2 hyperintensities inthe basal ganglia in three children with SNF (16). Simi-lar findings have been previously (2) and more recentlyrecorded (18). Interestingly, these brain lesions in thethree children seen at our institutions (16) were similarfor anatomic localisation to the other NF1 cases in the lit-erature, but appeared later than usually recorded in clas-sical NF1 (21). These brain high signal lesions, besidescurrently not included in the NF1 criteria or regardedas typical NF1 stigmata, had features reminiscent of thetypical hyperintense on T2-weighted lesions recorded inchildren with NF1 by the age of 4–6 years (21, 22).

Phenotypic variability within SNF families

Overall, in the cohort of SNF patients (Tables S1, S2 andS4) we could identify: (i) 9 SNF families (for a total of21/49 SNF patients), (ii) a single SNF patient within afamily with classical NF1, and (iii) 27 of 49 sporadicSNF patients (including the 5 previously unpublishedsporadic SNF patients; Tables 1 and S4).

In the cohort of MNFSR patients (Table S3) wecould identify: (i) 6 MNFSR families (for a total of20/32 MNFSR patients), (ii) 6 of 32 MNFSR patientsbelonging to four SNF families, and (iii) 6 of 32 sporadicMNFSR patients.

In addition, we could also record (Tables S1–S3) agroup of 14 patients (out of the 98 reported patientslabelled under the term ‘spinal neurofibromatosis’), hav-ing NF1 stigmata with no spinal neurofibromas, whobelonged either to SNF families (8/49) or to MNFSRfamilies (6/32).

Within the SNF spectrum we could record ‘pure’ SNFfamilies (i.e. where, after rigorous application of SNFcriteria, all affected members had SNF) (5, 9) or a singleSNF family with SNF twins (16). In most recordedSNF families, however, some or most affected membershad SNF and other family members had either MNFSRor classical NF1 or NF1 stigmata (Tables S1, S2, S4and S5). Some additional multi-generation families hadonly members with MNFSR (or mixed SNF/MNFSRmembers, as outlined above; Tables S1–S5).

Thus, there is a phenotypic variability within the SNFspectrum; SNF members can coexist aside MNFSR orNF1 members within families.

Molecular genetics

Molecular testing for the NF1 gene was performed in 40of 49 SNF patients (Tables S1, S2 and S5) and 24 of 32MNFSR patients (Tables S3 and S6).

NF1 gene abnormalities were recorded in 37 of 40SNF [including a double mutation in case B of Fauthet al. (11)] and 24 of 25 MNFSR patients. A total of24 different mutations were identified in the SNF cohort(Table S5).

At present, the restricted number of published reportson SNF precludes one from drawing firm conclu-sions about genotype–phenotype correlations in SNF.

407

Ruggieri et al.

Nonetheless, by comparing the NF1 mutations recordedin SNF (Table S5) vs MNFSR (Table S6), missensemutations were significantly more frequent in the SNFcohort (p= 0.0271). More interestingly, by comparingthe spectrum of NF1 gene mutations recorded in SNFvs the recorded spectrum in classical NF1 we wereable to calculate an increased risk of having SNF vsclassical NF1 (http://www.hgmd.org) in individualsharbouring missense mutations (p= 0.04; OR= 6.88;CI= 3.5–13.45), whereas for other mutations weobtained a (non-significant) p value >0.5. Overall, theanalysis of phenotype in the 16 SNF patients harbouringmissense mutations recorded café-au-lait spots in 62%(with only 2/16 patients having >6 macules), frecklingand Lisch nodules in 12%, cutaneous neurofibromas in43% (with only 1/16 having multiple skin tumours), andsubcutaneous neurofibromas in 25%.

A hallmark of NF1 is the high phenotypic variabilitybetween affected individuals, even among membersof the same family. The only phenotype–genotypecorrelations described so far in NF1 are large deletionsencompassing the whole NF1 gene (31, 32) and the 3-bpdeletion of exon 22. One of the striking characteristicsof SNF is the overall paucity of associated NF1-featuresand in particular the absence or low number of cuta-neous neurofibromas (Tables S1, S2 and S4), which isin contrast with the high number of early-onset cuta-neous neurofibromas seen in patients with the NF1micro-deletion syndrome (30, 31). Even though thenumbers hereby analysed (Table S5) are too few to drawgenotype–phenotype correlations we compared SNF vssevere classical NF1 (i.e. NF1 gene microdeletion syn-drome) by outlining the phenotype in the SNF patientswho had NF1 gene deletions (Table S7); interestingly,these four SNF deletion patients had few NF1-associatedfeatures, including few cutaneous neurofibromas. It mustbe noted, however, that one patient in this group haddysmorphic features (18) [never encountered in SNF(Tables S1 and S2)] and an additional patient died ofMPNST complications (11) (Table S7).

The SNF phenotype might derive from cooperationbetween the NF1 gene and other factors: e.g. modifiergenes. Some evidences seem to support such premisesincluding the identification of: (i) similar NF1 genemutations in both classical NF1 and SNF (Tables S5 andS6), (ii) SNF patients belonging to families with classi-cal NF1 who harboured identical NF1 gene abnormal-ities (Table S1), and (iii) families with either SNF andMNFSR affected members all harbouring identical NF1gene abnormalities (Tables S1 and S2). Further geneticstudies are needed to clarify these unsolved questions.

Natural history of disease

First appearance of spinal neurofibromas

Spinal neurofibromas, according to our experience(Tables 1 and S4), usually appear early in life (in somecases as early as the first years of life; Tables S1 and S2)(4, 9, 14, 16). Their detection rate, however, can be lowat young ages, as most (if not all) of these lesions can be

initially asymptomatic, becoming symptomatic later inlife (e.g. at or after age 20 or 30 or up to 50 years of age;Table S1), or they can be asymptomatic until older ages.When they are symptomatic early in life, this is usuallysecondary to massive involvement of both spinal rootsand major nerve trunks (4, 9, 14).

Age at onset of spinal neurofibromas in SNF rangedfrom 4 to 52 years and was lower in males (mean= 16;mode= 15) vs females (mean= 26; mode= 24) (TableS1). Interestingly, age at onset of spinal neurofibromas inMNFSR had a wider range from 2 to 60.5 and was higherthan SNF, irrespective of gender (males: mean= 31;mode= 31; females: mean= 33.5; mode= 30) (TableS3), thus reflecting the true onset of typical neurofibro-mas in classical NF1.

The initial localization was in one segment (by litera-ture review) or at multiple sites (by literature review andby personal experience) and the lesions could appear assimple nerve thickening (in some recorded cases nervethickening of all the spinal roots persisted until olderages; Table S1) or be tiny. Over the years the nodulesprogressed and multiplied over the affected and adjacentroot segments and nerve trunks.

The diagnosis of SNF, unless the suspected individ-ual belonged to SNF or NF1 families, was usually madeby chance because of (Table S1): (i) the appearanceof a subcutaneous nodule (a ‘lump’), (ii) the removalof a nodule, which turned out to be a neurofibromaor a PNF, (iii) the presence of café-au-lait spots, (iv)the appearance/progression of pain or motor signs, (v)the occasional detection of NF1-like bone dysplasia, or(vi) the occasional detection of nodules at MRI becauseof radiological investigation for other reasons. Some-times, despite the appearance of signs, which should haveraised the suspect of a form of neurofibromatosis (e.g.café-au-lait spots, histologically proven neurofibromas),the diagnosis was delayed until the appearance of moresevere signs/symptoms (e.g. large neurofibromas, para-paresis) (Table S1).

Signs/symptoms at onset and progression of disease

Among the most frequent neurological signs/symptomsrecorded at onset or bringing affected individuals to theirfirst referral are sensory deficits (neuropathy), and motorsigns (e.g. weakness or limb paresis/paralysis). Thesecan be accompanied by (electric-shock) pain. Due tothe predominantly extra-axial location of neurofibromas,an extensive tumour burden may be present in asymp-tomatic SNF patients. Symptomatic and asymptomaticSNF patients may have cutaneous or, more often, sub-cutaneous ‘lumps’ (and/or NF1 pigmentary anomalies)noticed earlier than the appearance of first neurologicalmanifestations.

Usually, signs and symptoms, relentlessly progress,with the appearance of increasing pain, gait distur-bances and paraparesis or, eventually, signs of moder-ate to severe cord compression due to enlargement ofthe paraspinal masses. Neurosurgical outcomes may beimproved by pre-symptomatic intervention in affectedindividuals.

408

Spinal neurofibromatosis

We recorded no symptomatology progression insome of the affected individuals seen at our institution(Tables 1 and S4).

Recommendations for diagnosis and managementand differential diagnosis

The diagnosis of SNF relies on visualization of neurofi-bromas on imaging (i.e. MRI) involving every nerve root,bilaterally. Irrespective to age, when a given individualpresent with multiple neurofibromas of many spinal seg-ments, serial imaging studies should be obtained so asto get (eventual) later confirmation of (bilateral) involve-ment of all spinal roots. It cannot be excluded a priori thatMNFSR can transition to SNF over time. Thus, unlessthe SNF phenotype is overt from the first imaging study,serial imaging studies should always be obtained to con-firm diagnosis: serial studies would serve for follow-upin confirmed cases.

Other NF1 stigmata may be present, but are oftenlacking. Café-au-lait spot to a lesser extent than classicalNF1, and intertriginous freckling, Lisch nodules andcutaneous neurofibromas are seen in a minority of cases.Thus, the affected status of SNF could be assignedirrespective of the presence or lack of NF1-associatedmanifestations or complications (33, 34), but requiresimaging of the spine. The presence of attenuated NF1manifestations without cognitive impairment or lack ofcutaneous neurofibromas in a pubertal toddler or anolder adolescent with NF1 may lead one to performwhole spinal imaging to detect SNF. One important issueis that many of the affected SNF (but also MNFSR)patients have endured several years of symptoms beforethe diagnosis was suspected: this may be unavoidable ifno stigmata of NF1 are evident (17).

MRI of the spine, limbs and abdomen (e.g. full bodyMRI) is pivotal for visualization of the tumours. BrainMRI and ophthalmologic evaluation should be obtainedin each patient. Electromyography and nerve conductionvelocities could confirm the peripheral nervous systemdamage and may help in localizing the affected site andthe extension of involvement.

There is a possibility that someone is referred with afew café-au-lait spots or a few cutaneous (or subcuta-neous) neurofibromas or with isolated Lisch nodules. Inthese cases attention should be paid to the specific fea-tures of the pigmentary anomalies and a biopsy might beindicated to determine if a tumour was a neurofibroma,schwannoma or something else. Other specific NF1 fea-tures/complications should be carefully checked clini-cally or by other means (e.g. ultrasonography, x-ray) asdictated by clinical findings. The usual clinical assess-ment in childhood (32) cannot be relied upon to confirmor refute the diagnosis of SNF (17). We usually recom-mend MRI of the whole spine and brain when the sus-picion of a related form of NF1 (or NF2) is raised (seebelow). Visualization of tumours on MRI involving everynerve root would lead to an assessment for other man-ifestations of NF1 or NF2. Likewise, absence of painwould steer one away from the diagnosis of schwanno-matosis. Due to the predominantly extra-axial location

of neurofibromas, an extensive tumour burden may bepresent in asymptomatic SNF patients.

Molecular NF1 gene testing should be alwaysobtained. The apparent higher relative-risk for SNF toharbour missense mutations needs future confirmationby larger studies.

Neurofibromas in SNF affect massively the spine andtheir multiplicity and progression, which ultimatelyleave no intact spinal segment, is important in that thesefeatures help in differentiating SNF from similar con-ditions including 33: (i) classical NF1, whose affectedpatients have usually a (even massive but) limited num-ber of nerve sheath tumours scattered at multiple (butnot all) levels of the spine (unilaterally or bilaterally)(21, 22). The presence/absence of typical NF1 stigmatacould also help in differentiating the two phenotypes; (ii)‘mosaic/segmental NF1’, in which patients have eithermultiple neurofibromas in localized (small or large)segments of their body or extensive PNFs involvinglocalized regions of their body (23, 24); (iii) classicalNF2, whose tumours are schwannomas, which areisointense to neural tissue on T1-weighted images andhyperintense (usually) on T2-weighted images andenhance uniformly after intravenous administration ofparamagnetic contrast. In addition to that these patientshave different NF2-associated stigmata (e.g. vestibularschwannomas, meningiomas, cataract, and peripheralschwannomas) (21, 22, 26, 27); (iv) ‘multiple isolatedneurofibromas’ (35), who usually do not have lesionsin all the spinal roots; and (v) ‘schwannomatosis’, whousually have paraspinal nerve sheath tumours (mostlyschwannomas but sometimes also some neurofibromas)or schwannomas along the major nerves of all the fourlimbs and thoracic region but do not (usually) manifestat any stage of the disease a massive involvement of thespinal roots as it occurs in SNF (25).

The anatomical lesion (through initial MRI suspicionand later histological confirmation, when needed) willhelp at least in differentiating between the two majorforms of neurofibromatosis (i.e. NF1 vs NF2; in thisrespect between SNF vs schwannomatosis, even thoughit must be kept in mind that, rarely, schwannomas can berecorded in the context of SNF; Table S1). The extensionof the area(s) involved and the natural history of diseasewill help in differentiating between the NF1-relatedforms. It is important to keep in mind that only aminority of cases with SNF could entirely satisfy the NF1diagnostic criteria (34); in these cases the diagnosis ofSNF vs classical NF1 will be based on the presence of thehallmark of disease. Therefore, besides the true familialcases of SNF or the cases of SNF within families withclassical NF1 or with MNFSR, and given the wide rangeof manifestations in SNF, assignment of the affectedstatus to a sporadic case can be achieved on the basisof the solely presence of neurofibromas of all the 38spinal roots. A new term (MNFSR) introduced in thisstudy could provide an avenue for clinical judgement andthe expectation that every spinal nerve root is involvedin SNF might be a bit restrictive. However, individualswith multiple spinal root tumours and few café-au-laitspots, without post-pubertal cutaneous neurofibromas,

409

Ruggieri et al.

and without cognitive impairment should be vieweddifferently than individuals with classical NF1.

Supporting Information

Additional supporting information may be found in the onlineversion of this article at the publisher’s web-site.

References1. Pulst SM, Riccardi VM, Fain P, Korenberg JR. Familial spinal neurofi-

bromatosis: clinical and DNA linkage analysis. Neurology 1991: 41:1923–1927.

2. Poyhonen M, Leisti EL, Kytola S, Leisti J. Hereditary spinal neurofibro-matosis: a rare form of NF1? J Med Genet 1997: 34: 184–187.

3. Ars E, Kruyer H, Gaona A et al. A clinical variant of neurofibromatosistype 1: familial spinal neurofibromatosis with a frameshift mutation inthe NF1 gene. Am J Hum Genet 1998: 62: 834–841.

4. Pasqual-Castroviejo I, Pascual-Pascual SI, Viano J, Martinez V. General-ized nerve sheath tumors in neurofibromatosis type 1 (NF1) a case report.Neuropediatrics 2000: 31: 211–213.

5. Kaufmann D, Müller R, Bartelt B et al. Spinal neurofibromatosis withoutcafé-au-lait macules in two families with null mutations of the NF1 gene.Am J Hum Genet 2001: 69: 1395–1400.

6. Wimmer K, Muhlbauer M, Eckart M et al. A patient severely affected byspinal neurofibromas carries a recurrent splice site mutation in the NF1gene. Eur J Hum Genet 2002: 10: 334–338.

7. Messiaen L, Riccardi V, Peltonen J et al. Independent NF1 mutations intwo large families with spinal neurofibromatosis. J Med Genet 2003: 40:122–126.

8. Kluwe L, Tatagiba M, Fünsterer C, Mautner VF. NF1 mutations andclinical spectrum in patients with spinal neurofibromas. J Med Genet2003: 40: 368–371.

9. Pascual-Castroviejo I, Pascual-Pascual SI, Velazquez-Fragua R, BotellaP, Viaño J. Familial spinal neurofibromatosis. Neuropediatrics 2007: 38:105–108.

10. Upadhyaya M, Spurlock G, Kluwe L et al. The spectrum of somaticand germline NF1 mutations in NF1 patients with spinal neurofibromas.Neurogenetics 2009: 10: 251–263.

11. Fauth C, Kehrer-Sawatzki H, Zatkova A et al. Two sporadic spinal neu-rofibromatosis patients with malignant peripheral nerve sheath tumour.Eur J Med Genet 2009: 52: 409–414.

12. Tsuji M, Harada S, Ueno Y, Osaka N. Neurological pictures. Famil-ial spinal neurofibromatosis: three generations of identical level symp-tomatic tumours. J Neurol Neurosurg Psychiatry 2010: 81: 1382.

13. Pizzuti A, Bottillo I, Inzana F et al. Familial spinal neurofibromatosis dueto a multiexonic NF1 gene deletion. Neurogenetics 2011: 12: 233–240.

14. Pascual-Castroviejo I, Pascual-Pascual SI, Viaño J, Velazquez-FraguaR, López-Gutierrez JC. Bilateral spinal neurofibromas in patients withneurofibromatosis 1. Brain Dev 2012: 34: 563–569.

15. Nunez-Farias AC, Bortzusky A, Morales PS, Zunino R, del Villar S.Severe spinal neurofibromatosis in a child. J Pediatr 2012: 161: 368.

16. Ruggieri M, Polizzi A, Salpietro V et al. Spinal neurofibromatosis withcentral nervous system involvement in a set of twin girls and a boy.Further expansion of the phenotype. Neuropediatrics 2013: 44: 239–244.

17. Burkitt Wright EM, Sach E, Sharif S et al. Can the diagnosis of NF1 beexcluded clinically? A lack of pigmentary findings in families with spinalneurofibromatosis demonstrates a limitation of clinical diagnosis. J MedGenet 2013: 50: 606–613.

18. Carman KB, Yakut A, Anlar B, Ayter S. Spinal neurofibromatosisassociated with classical neurofibromatosis type 1: genetic charac-terization of an atypical case. BMJ Case Rep 2013: Feb 14; doi:10.1136/bcr-2012-008468.

19. Nicita F, Torrente I, Spalice A et al. Spinal neurofibromatosis in a familywith classical neurofibromatosis type 1 and novel NF1 gene mutation.J Clin Neurosci 2014: 21 (2): 328–330.

20. Nicita F, Spalice A, Ruggieri M. Spinal neurofibromatosis in children.J Pediatr 2013: 162: 217.

21. Ruggieri M, Upadhyaya M, Di Rocco C, Gabriele A, Pascual-CatroviejoI. Neurofibromatosis type 1 and related disorders. In: Ruggieri M,Pascual-Castroviejo I, Di Rocco C, eds. Neurocutaneous disorders:phakomatoses and hamartoneoplastic syndromes. Wien, New York, NY:Springer-Verlag, 2008: 51–151.

22. Vezina G, Barkovich AJ. The Phakomatoses. In: Barkovich AJ, RaybaudC, eds. Pediatric neuroinaging, 5th edn. Philadelphia, PA: LippincottWilliams and Wilkins, 2012: 569–636.

23. Ruggieri M, Polizzi A. Segmental neurofibromatosis. J Neurosurg 2000:93: 530–532.

24. Ruggieri M, Huson SM. The clinical and diagnostic implications ofmosaicism in the neurofibromatoses. Neurology 2001: 56: 1433–1443.

25. Plotkin SR, Blakeley JO, Evans DG et al. Update from the 2011International Schwannomatosis Workshop: from genetics to diagnosticcriteria. Am J Med Genet A 2013: 161: 405–416.

26. Ruggieri M, Iannetti P, Polizzi A et al. Earliest clinical manifestationsand natural history of neurofibromatosis type 2 in childhood: a study of24 patients. Neuropediatrics 2005: 36: 21–34.

27. Ruggieri M, Gabriele AL, Polizzi A et al. Natural history of neurofibro-matosis type 2 (NF2) with onset before the age of 1 year. Neurogenetics2013: 14: 89–98.

28. Ruggieri M, Pavone P, Polizzi A et al. Ophthalmological manifestationsin segmental neurofibromatosis type 1. Br J Ophtalmol 2004: 88:1429–1433.

29. Nicita F, Iannetti L, Spalice A et al. Unilateral Lisch nodules in a47-year-old woman without other stigmata of neurofibromatosis type 1:an example of segmental neurofibromatosis? Ophtalmic Genet 2013: 34:178–179.

30. Ruggieri M, Pavone V, De Luca D, Franzò A, Tiné A, Pavone L.Congenital bone malformations in patients with neurofibromatosis type1. J Pediatr Orthop 1999: 19: 301–305.

31. Kayes LM, Burke W, Riccardi VM et al. Deletions spanning theneurofibromatosis 1 gene: identification and phenotype of five patients.Am J Hum Genet 1994: 54: 424–436.

32. Mautner VF, Kluwe L, Friedrich RE et al. Clinical characterisation of 29neurofibromatosis type-1 patients with molecularly ascertained 1.4 Mbtype-1 NF1 deletions. J Med Genet 2010: 47: 623–630.

33. Ruggieri M. The different forms of neurofibromatosis. Childs Nerv Syst1999: 15: 295–308.

34. National Institutes of Health Consensus Development Conference. Neu-rofibromatosis conference statement. Arch Neurol 1988: 45: 575–578.

35. Blakley P, Louis DN, Short MP, MacCollin M. A clinical study ofpatients with multiple isolated neurofibromas. J Med Genet 2001: 38:485–488.

410