Genotype-phenotype Correlations and assay results
Genotype-phenotype correlations indicate allele-specific effects forNF1 . There are multiple MS and small in-frame deletion variants
that are associated with unique NF1 phenotypes that may be utilized for
further functional analysis. A mild phenotype is associated with the
c.2970-2972 delAAT (delM992) single amino acid deletion, consisting of
café-au-lait macules (CALMs) and skinfold freckling and lack of
neurofibromas (Koczkowska, Callens, et al., 2018; Upadhyaya et al.,
2007). Variants involving R1809 are the most frequent recurrentNF1 variants and present with multiple CALMs, with or without
freckling and Lisch nodules, but externally visible plexiform
neurofibromas, symptomatic optic pathway glioma or cutaneous or
subdermal neurofibromas are not found (Rojnueangnit et al., 2015). Mild
phenotypes are also associated with R1038G (Trevisson et al., 2019) and
M1149V (Koczkowska et al., 2019). All of these genotypes are associated
with Noonan-like facial features. In fact, 31.1%, 29% and 11.5% of
individuals with variation at R1809, M1149, and M992, respectively, show
Noonan features, in comparison to only 3.4% of “classic” NF affected
individuals (Koczkowska et al., 2019). The R1038G cohort, consisting of
two families, is too small to meaningfully compare, but Noonan features
are also reported in both families (Trevisson et al., 2019). Overall,
NF1 affected individuals with delM992, R1038G, M1149V, and R1809C
associated with mild phenotypes lack clinically suspected plexiform,
cutaneous, or subcutaneous neurofibromas and are not at risk for
malignancy.
In contrast, other MS variants are associated with more severe
phenotypes, including increased risk for malignancy and/or spinal
neurofibromas. Constitutional MS variants affecting one of five
neighboring NF1 codons—Leu844, Cys845, Ala846, Leu847, and
Gly848—located in the cysteine-serine-rich domain (CSRD), typically
result in a large number of plexiform and symptomatic spinal
neurofibromas, symptomatic optic nerve gliomas, skeletal abnormalities,
and malignant neoplasms (Koczkowska, Chen, et al., 2018). R1276Q has
been identified in some individuals affected with spinal neurofibromas
at all levels and also has been associated with a more severe phenotype
(Koczkowska et al., 2019; Korf et al., 2005). K1423E has been associated
with a severe phenotype (Koczkowska et al., 2019). Some of these
variants (affecting residues G848 and R1276) are observed in individuals
with a distinctive phenotype, referred to as “spinal NF.” The “spinal
NF” phenotype includes few or no cutaneous neurofibromas (cNfs) and a
very mild pigmentary phenotype (Burkitt Wright et al., 2013; Ruggieri et
al., 2015). These individuals may suffer from a massive internal tumor
burden, with neurofibromas at each spinal nerve root and extreme
enlargement of most peripheral nerves. These individuals are at great
risk of spinal cord compression, pain, and malignant change, and the
extreme number of tumors makes surgical treatment difficult or
impossible (Koczkowska et al., 2019; Korf et al., 2005). Certain
variants correlate with increased incidences of cancer. A greater risk
of malignancy for MS variants in codons 844-848 has been reported
(Koczkowska, Chen, et al., 2018). Codon 847 is recurrent in NF1 patients
with breast cancer (Frayling et al., 2019). A lack of large deletions
with an excess of NS and FS variants has been observed with breast
cancer (Frayling et al., 2019; Zheng et al., 2020). Thus, the type ofNF1 variant matters when genotype-phenotype correlations can be
made and can influence clinical care.
We had hoped that our functional assays might predict which genotypes
could result in specific phenotypes, but a simplistic interpretation is
not readily available. Our assays indicate that most of the variants
associated with mild phenotypes are hypomorphic alleles. All mild
variants are relatively stable and produce neurofibromin at levels above
60% of WT levels. It is not surprising that a certain threshold of
neurofibromin must be achieved to have a mild phenotype; however, Ras
signaling may be altered depending on the variant. Ras signaling for
R1038G and R1809 is elevated, but not statistically different than for
WT cDNA. The delM992 variant is unable to completely suppress GTP-Ras or
pERK activity; this is statistically significant for GTP-Ras. Finally,
M1149V has statistically significant increased Ras signaling. These data
suggest that delM992, R1038G, and R1809C act as hypomorphs. In contrast,
M1149V appears to have lost the ability to inhibit Ras signaling and an
explanation for this unanticipated result for this “mild” variant is
not available.
Other genotypes are associated with severe phenotypes: L847P, G848R,
R1276Q, and K1423E. cDNAs with these genotypes have highly variable NF1
levels, ranging from 38% - 136% that of WT cDNA. L847P and G848R are
located in the CSRD; R1276Q and K1423E are located within the GRD and
interact directly with Ras. R1276Q and K1423E are completely unable to
suppress Ras signaling and are statistically different from WT cDNA, as
would be expected for variants that are critical for Ras binding and
GTP-hydrolysis. L847P is also unable to repress Ras signaling; however,
G848R can repress Ras signaling. Its lack of stability likely explains
why it is unable to function properly and causes a phenotype.
Given the interdependency of stability and function, we wanted to
evaluate NF1 stability as a function of Ras activity and plotted
neurofibromin levels with GTP-Ras levels (given that this assay had less
variability than the pERK/ERK assay) and drew a trend line. We noted
clustering of controls and cryptic splice variants, and that genotypes
associated with mild phenotypes also are loosely clustered (orange
circle). The presence of the “unknown” variant S1997R within this
cluster suggests that individuals with the genotype may have a mild
phenotype. In fact, the Leiden Open Variation Database (LOVD) has
classified this variant as a variant of uncertain significance (VUS);
however, we have identified an individual that meets NF1 diagnostic
criteria with this de novo variant. Though still adolescent, no
cutaneous or plexiform neurofibromas have been identified and the
phenotype is thus-far “mild”. In addition, we find multiple variants
that hug the trend line: variants L847P and G848R (associated with
severe phenotypes) and variants L2317P, C379R, W784R, and L1957P. This
suggests that given a certain abundance of neurofibromin, the variants
are partially able to repress Ras signaling. If the protein could be
stabilized in vivo , Ras might be repressed and the phenotype
rescued. Thus, we demonstrate two distinct clusters of genotypes
associated with severe phenotypes, one (pink oval) indicating loss of
GTPase function results in pathogenicity and the other indicating loss
of stability leads to pathogenicity.
Six NF1 phenotypic subtypes have recently been proposed, and while
genotypic data were inadequate to make statistically significant
conclusions, particular variants were noted to be consistent with three
of the six clusters (Tabata, Li, Knight, Bakker, & Sarin, 2020).
delM992 was consistent with the mild subtype (cluster 1); R1809C was
consistent with the freckling-predominant subtype (cluster 2), and L847P
and was consistent with the early-onset neural severe (cluster 6)
subtype. Ideally, combining genotype and functional data with such
phenotypic clustering would be a powerful tool in understanding the
phenotypic heterogeneity of NF1. Unfortunately for this cohort (derived
from a self-reported registry) only 61 of 2051 participants provided
molecular diagnostic data (though ~50% reported that a
molecular diagnosis had been made).
Study Limitations: There are several factors that limit our
study. First, HEK293 cells are very different from Schwann cells (one of
the cell types primarily affected in individuals with NF1). HEK293 cells
are derived from human embryonic kidney cells (not neural crest cells;
but they maybe neuronal as they share similarities with embryonic
adrenal precursor cells (Lin et al., 2014) transformed by incorporation
of 4.5 kb of adenovirus 5 genome into human chromosome 19 and carries a
modal chromosome number of 64 in 30% of cells. Notably, this increased
chromosome number does not affect any of the RAS or RASGAP genes.
HEK293s have all three wild type Ras isoforms. Even though the Ras
pathway is remarkably conserved, there could be modifiers in this cell
line not present in Schwann cells, melanocytes, neurons, and
osteoblasts/osteoclasts. Second, while Ras assays are commonly employed
to evaluate NF1 variants, determination and evaluation of
alternative functions is critical. Little information is available
regarding how variants might affect NF1 dimerization, nuclear
localization (or cellular localization in general) or even how they
might interact with other binding partners. Recently, a new mechanism
whereby NF1 binds the estrogen receptor (ER) and acts as a
transcriptional corepressor has emerged; this ER activity is
functionally independent of GAP activity (Zheng et al., 2020). Thus, NF1
is a dual repressor for both Ras and ER signaling. Defining how variants
affect these functions will aid our understanding of neurofibromin
structure-function. Finally, our assay is an over-expression assay and
hence does not reflect endogenous expression levels.
To date, NF1 cannot be cured. While MEK inhibitors can block Ras
signaling regardless of mutation type, therapeutics that address the
underlying cause of the disease by restoring neurofibromin function to a
level that leads to a non-pathogenic phenotype do not yet exist. Various
gene and mRNA targeting strategies have been proposed and are being
evaluated for their therapeutic potential in NF1 (Leier et al., 2020).
Compounds such as proteosome inhibitors could be used to stabilize
neurofibromin levels if the protein is being targeted for degradation.
Potentiators (analogous to those utilized for CFTR) might be used to
directly bind NF1 and stabilize it or prevent it from being degraded.
NF1 mimetics might be developed to stimulate Ras GTPase activity. Thus,
molecular diagnostics and determination of a variant’s stability and
function are increasingly relevant to guide clinical care for those with
known genotype-phenotype correlations and may also have implications for
both classification of variants of uncertain significance (particularly
those found in breast cancer) and developing therapeutics. Once variants
and their effects are established and categorized, new classes of
therapeutics become possible.