Retrospective Multicentric Study on Non-Optic CNS Tumors in Children and Adolescents with Neurofibromatosis Type 1

The natural history of non-optic central nervous system (CNS) tumors in neurofibromatosis type 1 (NF1) is largely unknown. Here, we describe prevalence, clinical presentation, treatment, and outcome of 49 non-optic CNS tumors observed in 35 pediatric patients (0–18 years). Patient- and tumor-related data were recorded. Overall survival (OS) and progression-free survival (PFS) were evaluated. Eighteen patients (51%) harbored an optic pathway glioma (OPG) and eight (23%) had multiple non-optic CNS lesions. The majority of lesions (37/49) were managed with a wait-and-see strategy, with one regression and five reductions observed. Twenty-one lesions (42.9%) required surgical treatment. Five-year OS was 85.3%. Twenty-four patients progressed with a 5-year PFS of 41.4%. Patients with multiple low-grade gliomas progressed earlier and had a lower 5-year PFS than those with one lesion only (14.3% vs. 57.9%), irrespective of OPG co-presence. Non-optic CNS tumors are common in young patients with NF1. Neither age and symptoms at diagnosis nor tumor location influenced time to progression in our series. Patients with multiple lesions tended to have a lower age at onset and to progress earlier, but with a good OS.


Study Population
Thirty-five patients were enrolled in the study. Twenty-eight patients were followed at the Neurofibromatosis Referral Center of the "Luigi Vanvitelli" University of Campania, Italy, and seven at Dana Children's Hospital, Tel Aviv Sourasky Medical Center, Israel. Some of the patients described in the present report were included in other previously published studies [6,[22][23][24]. Patient demographics and clinical data are summarized in Table 1. An MRI scan was obtained for 12 patients (34.2%) with symptoms of intracranial hypertension (4 cases); persistent headache (3 cases); neurological signs, including hemiparesis, right-side pyramidal signs, cerebellar signs, absence, and dizziness (4 cases); and endocrine disorder (growth hormone hypersecretion; 1 case). MRI was performed as a follow-up of eight patients with OPG and one with moyamoya syndrome, and to screen 13 patients for OPG and one for intellectual disability (Table 1).
In addition, eight out of 31 patients with histological or radiological diagnosis of LGG (25.8%) had multiple metachronous lesions other than OPG: four patients had two lesions, two had three lesions, and two had four lesions ( Table 2). Age at diagnosis of CNS tumor was lower in patients with multiple non-optic lesions than in patients with one lesion only (median of 7 years, interquartile range (IQR) of 5.4-10.6, range 3.2-15.8 vs. median 10.8, IQR 9.3-14.6, range 3.8-18, p = 0.048). There was no difference in age at diagnosis of CNS tumor in patients with or without OPG (median 9.5 years, IQR 6.4-14.3, range 3.8-16.2 vs. median 11.1 years, IQR 9.4-14.4, range 3.2-18, p = 0.186).  .08, at a median time from diagnosis of 0.7 years, IQR 0.11-1.81, range 0.01-2.08. Two patients with high-grade glioma (HGG) died because of tumor progression, occurring 4 months and 2.1 years from diagnosis. A third patient with intracerebral malignant peripheral nerve sheath tumor (MPNST) died as a result of post-surgical complications shortly after diagnosis, while a fourth affected by diffuse astrocytoma of corpus callosum died three years after diagnosis from another tumor, a progressive metastatic MPNST involving thoracic nerve roots from T4 to T12.

Tumor Characteristics and Treatment Strategies
We recorded a total of 49 tumors. Histology was available for 23 lesions (19 patients; Table 1), 18 of which (15 patients) were World Health Organization (WHO) grade 1 gliomas (14 PA and 4 gangliogliomas), one was a WHO grade 2 glioma (low-grade diffuse astrocytoma), two (2 patients) were grade 4 gliomas (glioblastoma), one was an intracranial MPNST, and one was an intracerebral schwannoma ( Figure 1). The remaining 26 lesions were considered LGGs based on MRI characteristics.  .08, at a median time from diagnosis of 0.7 years, IQR 0.11-1.81, range 0.01-2.08. Two patients with highgrade glioma (HGG) died because of tumor progression, occurring 4 months and 2.1 years from diagnosis. A third patient with intracerebral malignant peripheral nerve sheath tumor (MPNST) died as a result of post-surgical complications shortly after diagnosis, while a fourth affected by diffuse astrocytoma of corpus callosum died three years after diagnosis from another tumor, a progressive metastatic MPNST involving thoracic nerve roots from T4 to T12.

Tumor Characteristics and Treatment Strategies
We recorded a total of 49 tumors. Histology was available for 23 lesions (19 patients; Table 1), 18 of which (15 patients) were World Health Organization (WHO) grade 1 gliomas (14 PA and 4 gangliogliomas), one was a WHO grade 2 glioma (low-grade diffuse astrocytoma), two (2 patients) were grade 4 gliomas (glioblastoma), one was an intracranial MPNST, and one was an intracerebral schwannoma ( Figure 1). The remaining 26 lesions were considered LGGs based on MRI characteristics. Thirty-six lesions (73.4%) showed enhancement after contrast injection on T1-weighted MRI sequences, while 13 lesions did not show any enhancement in post-gadolinium MRI images.
Anatomical distribution of tumors according to histology are reported in Table 3. Most lesions (18, 36.7%) were located in the brainstem. The cerebellum was the second most common tumor location (10, 20.4%), followed by cerebral hemispheres (7,14.3%), with other locations accounting for the remaining 28.6%. Thirty-six lesions (73.4%) showed enhancement after contrast injection on T1-weighted MRI sequences, while 13 lesions did not show any enhancement in post-gadolinium MRI images.
Anatomical distribution of tumors according to histology are reported in Table 3. Most lesions (18, 36.7%) were located in the brainstem. The cerebellum was the second most common tumor location (10, 20.4%), followed by cerebral hemispheres (7,14.3%), with other locations accounting for the remaining 28.6%. Table 3. The initial approach to lesions based on histology and tumor location. CT, chemotherapy; RT, radiotherapy; HGGs high grade gliomas; MPNST, malignant peripheral sheet tumor; LGGs, low grade gliomas.
The initial approaches to lesions based on histology and tumor location are summarized in Table 3. A total of 25 lesions (21 patients) were treated at diagnosis or because of clinical or radiological progression. Specifically, 21 out of 49 tumors (42.9%) were treated with surgery (19 patients): six at diagnosis (12.2%), 12 during wait-and-see management (24.5%), and three after initial chemotherapy (6.1%). Thermal ablation of a PA of the right frontal lobe was performed in an 11-year-old girl.

Gliomas (HGGs and LGGs)
The two HGG lesions were treated with surgery, radiotherapy (60 Gy), and temozolamide (75 mg/m 2 /d). The first patient was a 6-year-old boy with HGG located in the right cerebellar hemisphere, while the second was a 17-year-old girl with HGG in the basal ganglia. The former died four months after diagnosis, while the latter 2.1 years after diagnosis, after having received cranial irradiation for an OPG 14 years previously.
Initial approaches adopted for suspected LGG lesions are listed in Table 3. Thirty-seven lesions were treated with a wait-and-see approach for a median time of 5.68 years, IQR 3.26-7.71 (range 0.14-16.27). The evolution of these patients and lesions, together with their outcome, is schematically reported in Figure 2. Specifically, 12 tumors (12 patients) required surgery after an initial observation and were located in cerebellar hemispheres (4), cerebral hemispheres (4), the basal ganglia (1), brainstem (1), corpus callosum (1), and ventricle (1). progressed 11 months after surgery and was therefore resected again; the residual disease has remained stable for eight years. The cerebellar lesion progressed three months after the first surgery and was eventually completely resected, with no further relapse for three years.
A total of 16 LGGs were treated by surgery, with a curative gross total resection (GTR) performed in eight of them. Of the remaining eight lesions that were not completely removed at the first attempt, three required further surgery and five remained stable.   Two lesions (2 patients) were surgically treated at diagnosis. One was a brainstem glioma causing hydrocephalus and the other was located in an anatomically critical site (cerebellar hemisphere). Both patients underwent subtotal resection (STR). The residual brainstem lesion progressed 11 months after surgery and was therefore resected again; the residual disease has remained stable for eight years. The cerebellar lesion progressed three months after the first surgery and was eventually completely resected, with no further relapse for three years.
A total of 16 LGGs were treated by surgery, with a curative gross total resection (GTR) performed in eight of them. Of the remaining eight lesions that were not completely removed at the first attempt, three required further surgery and five remained stable.   One patient with OPG and multiple brain lesions developed a third PA within a lesion initially considered an unidentified bright object (UBO) of the occipital hemisphere ( Figure 4). Post-operative image (E). The tumor histology was confirmed as a pilocytic astrocytoma.

Schwannoma and MPNST
The intracerebral parietal schwannoma was asymptomatic and diagnosed by screening MRI (Figure 1). The lesion was completely removed and did not show any signs of relapse during a follow-up of 6.75 years.
The MPNST, arising from the brainstem and occupying the occipital foramen, was surgically treated at diagnosis because of intracranial hypertension and severe brainstem compression. The patient died following a stormy post-operative period complicated by hydrocephalus, swallowing problems, tracheostomy, and cerebrospinal fluid mycotic infection.

Survival Analyses and Predictive Factors
Whole population 5-year OS was 85.3%, irrespective of tumor histology and treatment ( Figure  5). Twenty-four patients out of 35 progressed during follow-up with a 5-year PFS of 41.4% ( Figure  5). No significant difference was found in median time to progression for the following binary variables: gender (p = 0.24), age at diagnosis less or more than 10 years (p = 0.92), tumor location (posterior fossa vs. others, p = 0.77; brainstem vs. others, p = 0.60; midline supratentorial vs. others, p = 0.84). Three patients required chemotherapy as a first-line treatment because of symptomatic, unresectable tumors [25], with a total of six lesions, three of which were located in the brainstem, two in the thalamus, and one in the right temporal lobe. Patients were treated with vincristine and carboplatin according to International Society of Pediatric Oncology (SIOP) LGG 2004 protocol [25]. Three lesions (2 patients) subsequently required surgery because of progression; a subtotal resection was achieved for a lesion of the temporal lobe, whereas lesions of the brainstem and thalamus were partially removed. All tumors remained stable following surgery.
Two lesions (2 patients) were surgically treated at diagnosis. One was a brainstem glioma causing hydrocephalus and the other was located in an anatomically critical site (cerebellar hemisphere). Both patients underwent subtotal resection (STR). The residual brainstem lesion progressed 11 months after surgery and was therefore resected again; the residual disease has remained stable for eight years. The cerebellar lesion progressed three months after the first surgery and was eventually completely resected, with no further relapse for three years.
A total of 16 LGGs were treated by surgery, with a curative gross total resection (GTR) performed in eight of them. Of the remaining eight lesions that were not completely removed at the first attempt, three required further surgery and five remained stable.

Schwannoma and MPNST
The intracerebral parietal schwannoma was asymptomatic and diagnosed by screening MRI (Figure 1). The lesion was completely removed and did not show any signs of relapse during a follow-up of 6.75 years.
The MPNST, arising from the brainstem and occupying the occipital foramen, was surgically treated at diagnosis because of intracranial hypertension and severe brainstem compression. The patient died following a stormy post-operative period complicated by hydrocephalus, swallowing problems, tracheostomy, and cerebrospinal fluid mycotic infection. When considering LGGs, patients with multiple lesions (outside the optic pathway) progressed earlier than those with one lesion only, with a median time to progression of 1.52 years (95% confidence interval (CI) 0.11-2.94) vs. 5.76 years (95% CI 2.5-9.0), respectively (p = 0.03). Fiveyear PFS was 14.3% for patients with multiple non-optic lesions and 57.9% for patients with one lesion only ( Figure 6).

Survival Analyses and Predictive Factors
No significant difference was found in median time to progression in patients with and without OPG (4.99 years (95% CI 0-10.23) vs. 3.66 years (95% CI 0-7.37), respectively, p = 0.63). A similar 5-year PFS was found in both groups (42.8% for patients with OPG vs. 48.2% for patients without OPG). None of the LGG patients received chemotherapy for OPG during the observation time.
The presence of multiple non-optic LGG lesions proved to be the only independent predictor of progression in the univariate analysis (Hazard Ratio 2.81 (CI 95% 1.06-7.46), p = 0.038), while the other explored variables were not found statistically significant (Table S1). When considering LGGs, patients with multiple lesions (outside the optic pathway) progressed earlier than those with one lesion only, with a median time to progression of 1.52 years (95% confidence interval (CI) 0.11-2.94) vs. 5.76 years (95% CI 2.5-9.0), respectively (p = 0.03). Five-year PFS was 14.3% for patients with multiple non-optic lesions and 57.9% for patients with one lesion only ( Figure 6).
No significant difference was found in median time to progression in patients with and without OPG (4.99 years (95% CI 0-10.23) vs. 3.66 years (95% CI 0-7.37), respectively, p = 0.63). A similar 5-year PFS was found in both groups (42.8% for patients with OPG vs. 48.2% for patients without OPG). None of the LGG patients received chemotherapy for OPG during the observation time.
The presence of multiple non-optic LGG lesions proved to be the only independent predictor of progression in the univariate analysis (Hazard Ratio 2.81 (CI 95% 1.06-7.46), p = 0.038), while the other explored variables were not found statistically significant (Table S1). Cancers 2020, 12, x 10 of 19 Figure 6. Kaplan-Meier progression-free survival (PFS) curves by non-optic multiple lesions in patients with low-grade gliomas. Please note that just one patient was censured among those with multiple lesions and the relative hyphen is in light grey.

Genetic Findings
NF1 mutations were identified in 18 patients, of which 17 carried truncating mutations. The mutation type and effect together with the inheritance pattern are reported in Table 4 for each patient included in the study.

Discussion
To the best of our knowledge, our series represents the largest cohort of NF1 children and adolescents affected by extra-optic CNS tumors [4,5,8,12,[26][27][28][29][30][31]. In our series, almost half of lesions were incidentally detected (23/49, 46.9%). MRI was requested mainly (13) for OPG screening. Although the efficacy of routine brain MRI screening is still debated in NF1, clinicians should be aware of the opportunity to detect and manage an incidental brain tumor in this population. However, a negative scan does not exclude the possibility that a patient may later develop a brain tumor.
In our cohort, overall prognosis was good, with a 5-year OS of 85.3%, in line with previous reports [8]. Nevertheless, despite an excellent survival rate, 5-year PFS was 41.4% ( Figure 5). Contrary to the findings of previous studies, neither age and symptoms at diagnosis nor tumor location influenced progression timing in our series [17,[32][33][34].
Our population included a large number of LGGs (91.8% of all lesions).
LGGs are in fact the most common CNS tumors in NF1, with PA representing the most prevalent histological subtype [5,30]. We also recorded rarer histologies in our population, including HGGs, MPNST, and schwannoma.
We reported a prevalence of 26% of metachronous extra-optic LGGs in our pediatric and adolescent population with NF1 (8/31). In 2017, Sellmer et al. found 20% of patients with multiple non-optic gliomas in their series, including both adults and children [4]. In 2003, Guillamo et al. reported the same percentage of multiple CNS tumors, but they included OPGs [8]. Furthermore, they found that patients with multiple lesions were not affected by higher mortality. In agreement, none of our patients with metachronous lesions died.
In our study, patients with multiple lesions had a greater risk of progression and a lower 5-year PFS than patients with a single lesion, whereas the co-presence of OPG was not a risk factor for progression (Table S1). Although nearly all multifocal lesions radiologically worsened, patients tended to remain asymptomatic. These findings support the evidence that multiple CNS tumors, especially LGGs, are a hallmark of NF1 [8,31,35], and even in the case of multiple lesions and radiological progressions, patients could still be asymptomatic [7,34]. Tumor molecular biology may be different in this category of patients. It was in fact speculated that a patient is more likely to develop additional non-optic tumors if they have already developed one [4]. Our patients with multiple non-optic lesions have a lower CNS tumor onset age than those with a single non-optic lesion, strengthening this hypothesis.
Here, 13 out of 45 LGG lesions (28.9%) did not show any enhancement on post-contrast MRI images.
LGG in NF1 children might be enhanced heterogeneously [36] or not at all [37][38][39]. This observation makes the differentiation between LGGs and UBOs even more challenging. UBOs are non-enhancing T2 hyperintensities typical of NF1 found in the brain and medulla [39,40], which are described as potential precursors of brain tumors [41]. In agreement, a boy with OPG and multiple lesions in our population developed a third PA within a lesion initially considered as a UBO of the occipital hemisphere (Figure 4). The absence of contrast enhancement could, therefore, be misleading in diagnosis of LGG [37].
In terms of therapy, 60% of patients in our series received treatment for a CNS tumor. In the case of LGG, chemotherapy was the treatment option for symptomatic children with unresectable lesions [25], while GTR was confirmed curative in eight out of 16 tumors treated by surgery (50%), demonstrating its key role in managing LGG [25,42]. One patient previously treated for OPG in the first year of life and for aqueductal stenosis at the age of 7 developed a PA of the right frontal lobe at the age of 11 and underwent thermal ablation (LITT procedure) of this lesion. Although the long-term effects of laser ablation in patients with NF1 are not yet known, this technique is more precise, less invasive, does not involve ionizing radiations in terms of volume of tissue ablation compared to other methods, and allows complete resection, avoiding collateral damage [43][44][45][46].
A wait-and-see approach was adopted as a first-line strategy for almost 75% of patients and lesions in our series. Although one-third of tumors required further surgery, 16% did not progress, 13% showed radiological reduction, and one disappeared ( Figure 3). Cases of spontaneous resolution of brain gliomas in NF1 patients are described in the literature [47][48][49][50][51]. Therefore, when managing an asymptomatic lesion without radiological signs of high-grade histology, a wait-and-see policy should be taken [25].
From the genetic perspective, the number of genotype-phenotype associations in NF1 continues to increase. NF1 gene has one of the highest spontaneous mutation rates and more than 3000 mutations have been recognized to date [68]. The following variants are associated with specific phenotypes: missense variants at the NF1 codons p.Arg1809, p.Met1149, p.Arg1276, and p.Lys1423 and the in-frame deletion c.2970-2972 delAAT [69][70][71]. In our series, 19 out of 35 patients underwent molecular testing for NF1 and an NF1 mutation was identified in 18 cases ( Table 4). Seventeen of these mutations were predicted to produce a truncated neurofibromin.
NF1 germline mutations were shown to result in different levels of neurofibromin expression in NF1 patient fibroblasts, ranging from 25% to 75% [72]. This finding supports the hypothesis that not all NF1 mutations are equivalent and that residual amounts of functionally active neurofibromin might be linked to the phenotype [73]. Genotype-phenotype associations and the parent-of-origin effect overall or by patient sex were investigated and subsequently excluded for OPGs in NF1 [74][75][76]. It might be interesting to extend such studies and test this hypothesis in brain tumors. Any attempt to associate observed genotypes and brain tumors in our cohort of patients was limited by the very low number of cases examined. Although somatic NF1 gene inactivation is required for NF1-related tumorigenesis [77], we could speculate that: (1) the observed higher rate of truncating mutations in NF1 may suggest that at least 50% loss-of-function is necessary to initiate tumorigenesis; (2) additional genetic modifiers unlinked to the NF1 locus might play a role in brain tumors, as we proposed for NF1-related moyamoya vasculopathy [78].
Major limitations of the present study are its retrospective design and the lack of a molecular tumor profile. Our molecular and immunological understanding of NF1-associated gliomas is rapidly evolving, and biopsy is increasingly indicated not only for histologic confirmation, but also for molecular-targeted therapy [42]. In terms of new therapeutic options, The MEK1/2 inhibitor selumetinib seems to be active in NF1-related pediatric LGG [79]. The efficacy of selumitinib versus standard chemotherapy in both newly diagnosed and progressive or recurrent tumors, irrespective of biopsy, is currently under evaluation in a phase III trial (clinical trial.gov number NCT03871257).

Materials and Methods
This retrospective multicentric study involved patients (0-19 years of age) affected by NF1, according to NIH NF1 criteria (1988) [80] and diagnosed with non-optic CNS tumors from January 2000 to January 2019 at two centers.
Brain MRI images of all enrolled patients were reviewed. Any measurable area in at least two dimensions of high signal intensity on T2-weighted images was considered a tumor if it had one of the following characteristics: gadolinium enhancement, mass effect, peripheral edema, or a mural nodule associated with a cystic or necrotic component [8,38,39].
Histological confirmation of tumors was not mandatory for patient inclusion. Specifically, lesions with typical characteristics of LGGs did not undergo biopsy [7,37,42]. In contrast, histological examination was performed for lesions with radiological features of high-grade histology (high degree of tumor heterogeneity and contrast enhancement, restricted diffusion on diffusion-weighted MRI, and increased relative cerebral blood volume on perfusion-weighted MRI) [36,38,42]. Tumors with aggressive behavior and those requiring surgery were also subjected to histological examination.
Collected data included demographic characteristics (gender, age at diagnosis, and inheritance of NF1), age at diagnosis of CNS tumors and NF1, and clinical signs or symptoms at diagnosis of CNS tumors. Indications for brain MRI scan at diagnosis were also recorded. When a patient presented multiple lesions, data on all lesions were collected. We documented the anatomical site, enhancement of all lesions, and histology when available [81], as well as the co-presence of OPG. We also recorded the management of tumors, including wait-and-see, surgery, chemotherapy, and radiotherapy approaches. We collected data on the type of surgical resection (gross total-and subtotal-resection (respectively GTR and STR) [82,83], duration of follow-up, recurrence, progression, or resolution. Ethical approval from local committees was obtained.

Statistical Analysis
Continuous non-parametric variables are presented as the median, IQR, and range, whereas categorical variables are expressed as number and percentage. Mann-Whitney U test was used to compare continuous non-parametric variables. Kaplan-Meier analysis was run to determine OS and PFS [84]. OS was calculated from date of diagnosis until death from any cause. PFS was measured from date of diagnosis until radiological or clinical progression date. Specifically, for multiple lesions, the first radiological progression of one lesion was considered as the progression date, irrespective of the evolution of any others. For both OS and PFS analyses, patients were censored at last available follow-up time if no event occurred.
Factors that may influence time to progression (gender, age at tumor diagnosis, location of tumor, symptoms at diagnosis) were tested by comparing PFS curves with the log-rank test.
Exclusively for patients with histological or radiological diagnosis of LGG, log-rank test was used to compare PFS curves of patients with and without associated OPG, and those of children with single and multiple lesions. The Cox regression model was used to explore predictors of progression in patients with LGG. For all analyses, p values < 0.05 were considered statistically significant. IBM SPSS Statistics 22 Software for Windows was used for statistical analysis.

Genetic Findings
NF1 mutations were recorded in terms of the genomic and protein location and type of mutation. NF1 germline mutations were obtained from a review of clinical records and no further genetic analyses were performed during the study.

Conclusions
Extra-OPG CNS tumors are a relatively common malignancy in children and adolescents with NF1. Although other histological types do rarely occur, LGGs are the most common CNS tumors in this population. In our experience, LGG lesions tend to be and remain asymptomatic in young patients, regardless of radiological progression. In addition, patients with multiple metachronous LGGs have a lower age at onset and tend to progress earlier than patients with one lesion only, irrespective of the co-presence of an OPG. Hence, in line with the consensus very recently published by Packer et al. in 2020 [42], our experience confirms that a wait-and-see approach is initially advised for asymptomatic lesions with radiological characteristics of low-grade tumors. Conversely, surgery remains the best therapeutic option in symptomatic LGGs, with the aim of achieving complete resection.

Conflicts of Interest:
The authors declare no conflict of interest.