RESULTS
Assay validation with cDNA Titrations: To further validate our cDNA and assay system (Wallis et al., 2018), we performed titration experiments to show a dose-response effect utilizing 0.25 - 1000ng/well of WT mouse Nf1 cDNA in NF1 null HEK293 cells with 500,000 cells balanced with empty vector (EV) control. Neurofibromin/tubulin levels (Figure 2A), GTP-Ras levels (Figure 2B), and pERK/ERK ratios (Figure 2C) all respond in a dose-response manner to the amount of cDNA transfected into the cells. Neurofibromin levels become detectible via Western blot analysis between 1-4 ng/500,000 cells transfected (Figure 2A). GTP-Ras levels were lowered, beyond that of empty vector (0ng WT cDNA), starting at 1-4 ng/500,000 cells transfected, with statistically significant differences observed at 15 ng and above (Figure 2B). Lowered pERK/ERK ratios, a marker for MAPK signaling activity, were also detected at 1-4 ng/500,000 cells transfected, with statistically significant differences from EV at 250 ng (Figure 2C).
Variant Selection and groupings: We utilized a full gene-encompassing panel of mNf1 cDNAs transfected into anNF1 null (-/-) cell line to evaluate the functional effects of unique variants (Figure 3 and Table 1). Variants, in addition to WT and EV, were selected based on the following criteria: controls, genotype-phenotype correlations, occurrences in different NF1 domains, and type of mutation. Benign variants, both within and outside of the GRD, were selected due to lack of pathogenicity and as “controls”: E1327G, Q1336R, and P2782L. Variants with published genotype phenotype-correlations were prioritized. Variants associated with “mild” phenotypes include delM992, R1038G, M1149V, and R1809C (Koczkowska et al., 2019; Koczkowska, Callens, et al., 2018; Rojnueangnit et al., 2015; Trevisson et al., 2019; Upadhyaya et al., 2007). Variants associated with “severe” phenotypes include L847P, G848R, R1276Q, and K1423E (Koczkowska et al., 2019; Koczkowska, Chen, et al., 2018; Korf, Henson, & Stemmer-Rachamimov, 2005). We also included variants from multiple domains, despite having “unknown” phenotype associations. C379R falls within the 5’ region with no described domain function, W784R falls within the putative CSRD domain, L1490P falls within the SPRED1 interaction domain, D1623 falls within the Sec14 domain, and L1957P, S1997R, and L2317P all fall within the 3’ region of the protein but not in well-described domains. Additional variants were selected based on the formation of cryptic “splice” sites: Y489C (Messiaen et al., 1999) and G629R. While splicing is not affected in the cDNA system, assessment of functional effects of the subsequent missense variant is a critical first step in the development of antisense therapeutics that might restore normal splicing but leave the variant intact. Additional “nonsense” variants were prioritized based on incidence as well as location throughout the protein.
NF1 levels First we evaluated neurofibromin levels after transfecting cells with equal plasmid concentrations from each representative cDNA. All cDNAs were assayed for NF1 levels by transfecting a consistent 1 ug cDNA into a 6 well plate with 500,000 cells and harvesting cells 48 hours post transfection. Figure 4 demonstrates quantification of NF1/tubulin ratios for all cDNAs and indicates differential variant-specific effects on neurofibromin levels. For example, while some variants remain stable with levels similar to WT (e.g., R1809C), others show much lower levels of neurofibromin; L1490P and D1623G lead to approximately 20% WT levels. We interpret NF1 levels to reflect protein stability, which may be dependent on mutation-targeted proteasomal degradation. All cDNAs should have similar transfection, transcription, and translation efficiencies. Each experiment included both WT and EV control cDNAs. Samples were normalized to tubulin as a load control and WT/tubulin ratios were set at 1.0 in each experiment. All other cDNA/tubulin ratios were reported relative to WT levels. Normalization allowed comparison across experiments. Western blots of select MS variants show varying full-length NF1 levels (Figure 4B). Western blots of nonsense variants show varying NF1 levels (Figure 4C); note the truncation products run at the anticipated sizes; however, NF1 levels did not correlate with length of prematurely terminated mutant proteins.
Ras Activity: Next we assayed all Nf1 variant cDNAs for their ability to repress Ras signaling. We evaluated both GTP-Ras levels and pERK/ERK ratios (Figure 5). All cDNAs were assayed by consistently transfecting in 1 ug of cDNA into a 6 well plate with 500,000 cells/well and harvesting protein lysate 48 hours post transfection. Each experiment included both WT and EV control cDNAs. For GTP-Ras levels (blue bars; Figure 5A), each sample was normalized to EV GTP-Ras levels, which were set at 1.0 in each experiment, and all other GTP-Ras levels were reported relative to EV levels. Each cDNA’s GTP-Ras level was statistically compared via Student’s t-test to the WT cDNA’s GTP-Ras level to determine if the variant negatively impacted NF1 ability to repress GTP-Ras levels. Blue asterisks indicate that a variant has statistically significant impaired ability to inhibit GTP-Ras levels; these include: delM992, M1149V, L847P, R1276Q, K1423E, C379R, L1490P, D1623G, S1997R, L2317P, R192X, R461X, R681X, R816X, R1276X, and R1306X.
For pERK/ERK ratios (black bars; Figure 5A), each experiment included both WT and EV control cDNAs, with each sample normalized to the EV cDNA pERK/ERK ratio, which was set at 1.0. Each cDNA’s pERK/ERK ratio was statistically compared via Student’s t-test to the WT pERK/ERK ratio to determine if the variant negatively impacted its ability to repress pERK activity with black asterisks indicating statistically significant impaired ability; these include M1149V, L847P, R1276Q, K1423E, L1490P, D1623G, S1997R, L2317P, R192X, R461X, R681X, R816X, R1276X, R1306X, and R1947X.
Activity as a function of stability: Neurofibromin functionin vivo relies on numerous factors, including stability (abundance of protein available), cellular localization, ability to bind interacting proteins (such as Ras), and ability to stimulate Ras GTPase activity. As we measured two of these factors, neurofibromin levels and GTPase activity, we wanted to determine if the combined factors can lead to variant functional insights. To achieve this, we plotted ourNf1 WT cDNA titration data (derived from Figure 2) such that neurofibromin/tubulin levels at 1000 ng cDNA was set to a maximum of 1 and plotted on the x-axis and corresponding GTP-Ras activity levels were plotted on the y-axis (Figure 6, gray dots). A trend line was generated (Figure 6, blue dotted line). To evaluate this multidimensional concept, we overlaid variant data onto this plot and categorized variants as we had in Figures 4 and 5 as “Control” (green dots), “Splice” (yellow dots), “Mild” (orange dots), “Severe” (pink dots), and “Unknown” (teal dots). The control variants clustered such that NF1/tubulin ratios were > 0.75 and GTP-Ras levels were < 0.52 (Figure 6 green oval). The cryptic splice variants also clustered at NF1 > 0.89 and GTP-Ras < 0.66. While the mild variants didn’t cluster together as tightly, three of the four variants had stable protein levels > 0.8 and two had low GTP-Ras levels < 0.66. These loosely clustered mild variants also cluster with the splice variants (Figure 6 orange ovals). Severe and unknown variants did not form a single cluster. Severe variants R1276Q and K1423E that interact with Ras do maintain stability but cannot suppress Ras and are clustered in the top right (Figure 6 pink oval). Outside of those clusters, we find multiple variants that hug the trend line (blue oval): “Severe” variants L847P and G848R and unknown variants L2317P, C379R, W784R, and L1957P (these have been individually labeled in the plot). This suggests that given a certain abundance of neurofibromin, the variants can suppress Ras signaling.