FGFR1 expression is associated with bad clinical outcomes in NB patients
The association of FGFR1 expression with clinical outcomes was evaluated in three datasets that are deposited in the R2 microarray web tool (http://hgserver1.amc.nl/cgi-bin/r2/main.cgi): the Seeger dataset (102 patients with high-risk NB); the Versteeg dataset (88 patients) and the Asgharzadeh TARGET dataset (247 patients). Kaplan-Maier analysis showed that higher FGFR1 expression was significantly associated with inferior relapse-free survival in Seeger dataset (p-value = 3.1x10-5) and in Versteeg dataset (p-value = 0.057). In contrast, correlation of FGFR1 to overall survival was not significant in Asgharzadeh TARGET dataset (p-value = 0.061) and in Versteeg dataset (p-value=0.118) (Fig. 1A).
FGFR1 expression analysis in a dataset of 11 primary and 7 relapsed tumors showed a higher FGFR1 expression in relapsed NB samples without reaching the significance level (p-value=0.28), probably due to the limited number of samples (Fig. 1B).
Finally, we observed that FGFR1 mRNA levels in metastatic xenograft tumors were higher than those of NB primary tumors (p-value <0.001) but were similar to those of embryonic cells and neuronal crest cells (Fig. 1C).
FGFR1 somatic mutations and copy number variations in NB patients
We analyzed whole genome sequencing (WGS) data at FGFR1 locus (including 50 kb surrounding regions) from 19 paired diagnostic and relapse NBs. WGS data from 10 samples were obtained in our laboratory whereas 9 were downloaded from TARGET project repository.
We found the hotspot mutation FGFR1N546K in one tumor at diagnosis and relapse (Table 1). No other putative coding pathogenic mutations were found. We also wanted to investigate potential pathogenic function of noncoding point mutations. To this purpose, we annotated each mutation with DNase I hypersensitive sites, known to define active regulatory DNA elements, in SKNSH NB cells (ENCODE data). No potential pathogenic mutations located in DNA regulatory sites were found (Table 1).
Since FGFR1 amplifications have been associated with other cancers, we analyzed copy number variations in a public dataset of 381 NBs. No significant amplification of FGFR1 was found (Fig. S1).
Table 1. Coding and noncoding somatic mutations found at FGFR1 locus in 21 primary-relapse pairs NB tumors analyzed by whole genome sequencing.
Sample ID
|
Type
|
Position/Change
|
Location
|
Gene
|
Amino acid change
|
Band
|
CADD score
|
COSMIC ID
|
SNP ID
|
ENCODE annotation
|
TR008
|
R
|
chr8:38227325:G>T
|
intronic
|
WHSC1L1
|
.
|
8p11.23
|
2,48
|
.
|
.
|
Enhancerb
|
TR008
|
R
|
chr8:38236248:C>G
|
intronic
|
WHSC1L1
|
.
|
8p11.23
|
1,78
|
.
|
.
|
.
|
TR008
|
P
|
chr8:38246999:C>G
|
intronic
|
LETM2
|
.
|
8p11.23
|
1,77
|
.
|
.
|
.
|
TR008
|
R
|
chr8:38246999:C>G
|
intronic
|
LETM2
|
.
|
8p11.23
|
1,77
|
.
|
.
|
.
|
TR001
|
P
|
chr8:38267939:A>C
|
downstream
|
FGFR1,LETM2
|
.
|
8p11.23
|
1,03
|
.
|
.
|
TFP(CTCF)
|
SP_2_T
|
P
|
chr8:38268616:C>A
|
downstream
|
FGFR1
|
.
|
8p11.23
|
3,58
|
.
|
.
|
TFP(CTCF)
|
TR003
|
R
|
chr8:38273003:A>T
|
intronic
|
FGFR1
|
.
|
8p11.23
|
2,25
|
.
|
.
|
.
|
PATNKP
|
P
|
chr8:38274849:G>T
|
exonic
|
FGFR1
|
N457K
|
8p11.23
|
29,70
|
YESa
|
rs779707422
|
.
|
PATNKP
|
R
|
chr8:38274849:G>T
|
exonic
|
FGFR1
|
N457K
|
8p11.23
|
29,70
|
YESa
|
rs779707422
|
.
|
TR007
|
P
|
chr8:38282676:A>T
|
intronic
|
FGFR1
|
.
|
8p11.23
|
2,96
|
.
|
.
|
DHS(MCV-1)
|
TR001
|
P
|
chr8:38288403:G>C
|
intronic
|
FGFR1
|
.
|
8p11.23
|
0,22
|
.
|
.
|
DHS(MCV-2); TFP(SMARCC1)
|
TR008
|
P
|
chr8:38295809:T>C
|
intronic
|
FGFR1
|
.
|
8p11.23
|
0,23
|
.
|
rs975858205
|
.
|
PAUDDK
|
P
|
chr8:38296890:T>A
|
intronic
|
FGFR1
|
.
|
8p11.23
|
17,63
|
.
|
.
|
.
|
TR003
|
P
|
chr8:38301604:T>G
|
intronic
|
FGFR1
|
.
|
8p11.22
|
2,88
|
.
|
rs947373873
|
.
|
PATNKP
|
P
|
chr8:38311785:C>G
|
intronic
|
FGFR1
|
.
|
8p11.22
|
2,90
|
.
|
.
|
|
PATNKP
|
R
|
chr8:38311785:C>G
|
intronic
|
FGFR1
|
.
|
8p11.22
|
2,90
|
.
|
.
|
.
|
TR008
|
R
|
chr8:38319864:G>A
|
intronic
|
FGFR1
|
.
|
8p11.22
|
0,75
|
.
|
.
|
.
|
TR007
|
R
|
chr8:38324367:G>A
|
intronic
|
FGFR1
|
.
|
8p11.22
|
9,06
|
.
|
.
|
Enhancerb; TFP(SIN3A, TAF7, TCF12, YY1)
|
TR008
|
P
|
chr8:38337889:A>G
|
intergenic
|
FGFR1(dist=11537)
|
.
|
8p11.22
|
3,92
|
.
|
.
|
.
|
TR008
|
R
|
chr8:38338780:A>G
|
intergenic
|
FGFR1(dist=12428)
|
.
|
8p11.22
|
18,68
|
.
|
.
|
DHS(MCV-2)
|
TR008
|
R
|
chr8:38338784:C>G
|
intergenic
|
FGFR1(dist=12432)
|
.
|
8p11.22
|
17,75
|
.
|
rs201380585
|
DHS(MCV-2)
|
TR006
|
R
|
chr8:38349608:A>G
|
intergenic
|
FGFR1(dist=23256)
|
.
|
8p11.22
|
0,61
|
.
|
.
|
.
|
TR006
|
R
|
chr8:38350422:C>G
|
intergenic
|
FGFR1(dist=24070)
|
.
|
8p11.22
|
10,82
|
.
|
.
|
.
|
TR008
|
P
|
chr8:38350507:A>C
|
intergenic
|
FGFR1(dist=24155)
|
.
|
8p11.22
|
0,03
|
.
|
.
|
.
|
TR006
|
P
|
chr8:38350642:T>C
|
intergenic
|
FGFR1(dist=24290)
|
.
|
8p11.22
|
0,04
|
.
|
.
|
.
|
PATNKP
|
R
|
chr8:38357592:C>A
|
intergenic
|
FGFR1(dist=31240)
|
.
|
8p11.22
|
4,03
|
.
|
.
|
.
|
TR006
|
R
|
chr8:38369276:A>C
|
UTR3
|
C8orf86
|
.
|
8p11.22
|
0,45
|
.
|
rs565928745
|
.
|
aID=COSM3670398,COSM1284966,COSM1284968,COSM1284967,COSM19176;OCCURENCE=5(central_nervous_system),1(autonomic_ganglia)
P: Primary; R: Relapse.
bSegway/ChromHMM-predicted enhancer.
TFP: transcription factor binding peak.
DHS: DNase I hypersensitive sites.
MCV-1, MCV-2: Cell lines of the ENCODE catalog.
FGFR1 silencing impairs cells growth, invasion and colonigenicity in NB cells
We investigated the role of FGFR1 in two NB cell lines: SHSY5Y MYCN non-amplified and SKNBE2 MYCN-amplified cells.
We transduced SHSY5Y and SKNBE2 cells by lentiviral vectors encoding two independent short hairpin (sh)RNAs targeting FGFR1 (shFGFR1#A and shFGFR1#B) and a control shRNA (shCTR). Silencing efficiency was determined by western blotting and real time PCR (RT-PCR) (Fig. 2A).
Cell viability of both SHSY5Y and SKNBE2 shFGFR1 (shFGFR1#A and shFGFR1#B) significantly decreased compared to cell viability of shCTR at 48 and 72 hours (p-value ≤ 0.05) (Fig. 2B), suggesting that FGFR1 silencing impaired NB cell proliferation and cell growth.
Similarly, FGFR1 silencing affected NB cell ability to migrate through a matrigel-coated membrane (Fig. 2C and S2A) and the anchorage-independent growth, as shown by soft agar assay (Fig. 2D and S2B). Hence, colony numbers and invading cell numbers in shFGFR1 cells significantly decreased compared to shCTR cells in both SHSY5Y and SKNBE2 cell lines.
FGFR1N546K exhibits a nuclear localization
FGFR1 is constitutively found in cell membrane, cytoplasm and nucleus 44. Data samples contained in the Human Protein Atlas clearly show that FGFR1 can localize to the nucleus (https://www.proteinatlas.org/ENSG00000077782-FGFR1). FGFR1 nuclear localization in three-dimensional model of breast cancer and pancreatic cancer can influence the expression of hundreds of genes and contribute to migratory phenotype 45-47 (https://doi.org/10.1002/emmm.201302698). Additionally, in embryonic stem cells, FGFR1 nuclear localization may increase in developing brain cells during neuronal differentiation to Neuronal Progenitor Cells (NPC) 46;48.
In this study, we investigated FGFR1 localization in HEK293 cells and two NB cell lines, SHSY5Y and SKNBE2, overexpressing both FGFR1wt and FGFR1N546K.
HEK293, SHSY5Y and SKNBE2 cells were transiently trasfected with pCMV6 expressing FGFR1wt or FGFR1N546K proteins and pCMV6 empty vector.
In HEK293, we examined FGFR1wt and FGFR1N546K proteins localization by ImageStreamX Mark II Flow Cytometer (Fig. 3A). FGFR1 nuclear signal intensity was calculated for 1000 HEK293_FGFR1wt single cells and for 1000 HEK293_ FGFR1N546K single cells. Abundant nuclear localization of FGFR1N546K protein was statistically significant (p-value=0.0001). This observation was confirmed by immunofluorescence confocal microscopy assay showing FGFR1N546K protein mainly localized to nucleus, while FGFR1wt protein mainly localized to cytosol (Fig. 3B).
In SHSY5Y and SKNBE2 cell lines we observed a higher nuclear localization of the protein in FGFR1N546K overexpressing cells, compared to those overexpressing FGFR1wt (Fig. 3B).
These data were validated by western blot analysis on cytosol and nucleus fractions of HEK293, SHSY5Y and SKNBE2 trasfected cells (Fig. 3C).
FGFR1N546K establishes crosstalk pathway activation and induces an increase in NB cellular invasion and colonigenicity
Early studies reported FGFR1N546K mutation affects the conformational dynamics of the tyrosine kinase domain, resulting in gain-of-function and ligand-independent constitutive activation 26;49;50.
We performed western blotting analysis on total protein extracts from SHSY5Y and SKNBE2 transiently transfected with pCMV6 expressing FGFR1wt or FGFR1N546K proteins and pCMV6 empty vector to evaluate phosphorylated and total FGFR1, STAT3, ERK and AKT levels. The t-GFP protein level was used as trasfection control and β-Actin was used as loading control.
In NB cell lines, FGFR1N546K overexpression enhanced the receptor kinase activity resulting in higher FGFR1 auto-phosphorylation. In addition, we observed a higher ERK, AKT and STAT3 phosphorylation in FGFR1N546K compared to FGFR1wt overexpressing cells (Fig. 4A).
We then evaluated cell viability in both SHSY5Y and SKNBE2 overexpressing FGFR1wt and FGFR1N546K protein compared to empty vector. Cell viability at 24h, 48h and 72h significantly increased in FGFR1wt and FGFR1N546K overexpressing cells compared to pCMV6 empty vector (p-value ≤ 0.05) and FGFR1N546K overexpressing cells had the highest cell viability (p-value ≤ 0.05) (Fig. 4B).
The ability of transiently transfected SHSY5Y and SKNBE2 cells to invade and migrate through a matrigel-coated membrane support was evaluated. The number of invading FGFR1wt and FGFR1N546K overexpressing cells increased significantly compared to control pCMV6 cells. Interestingly, the number of invading FGFR1N546K overexpressing cells was even higher than the number of invading FGFR1wt overexpressing cells (p-value ≤ 0.05) (Fig. 4C and S2C).
In addition, we analyzed the capability of FGFR1wt and FGFR1N546K overexpressing cells to interfere with colonigenicity in SHSY5Y and SKNBE2 cell lines. FGFR1wt overexpression resulted in an increase of colony number and colony area compared to the empty vector in both cell lines (Fig. 4D and S2D). Moreover, FGFR1N546K overexpression was associated with the highest colony number and colony area in both cell lines (Fig. 4D and S2D).
N546K FGFR1 mutation may confer resistance to AZD4547 treatment in NB cell lines.
Since AZD4547 represents a small molecule inhibitor targeting FGFR1 aberrant activation currently used in clinical trial 51;52, we investigated the effects of this drug on FGFR1wt and FGFR1N546K in SHSY5Y and SKNBE2 cells.
Firstly, we evaluated AZD4547 potency against pCMV6-empty vector, FGFR1wt and FGFR1N546K overespressing cells (Fig. S3A).
We performed cell viability assays in both cell lines testing different AZD4547 concentrations (0.01 μM, 0.1 μM, 1 μM and 10 μM), and then we calculated the half maximal inhibitory concentration (IC50) for this drug, which resulted comparable in SHSY5Y and SKNBE2 (Fig. S3A).
Based on the IC50 results, we selected the lower concentration of AZD4547 (0.1 μM) able to decrease viability up to 20% in both cell lines (Fig. S3A). Specifically, we decided to not test 1 μM AZD4547 because the treatment with this concentration showed a reduction of 31% of cell viability in SKNBE2 pCMV6-empty vector compared to vehicle cells (DMSO) (Fig. S3A).
To investigate the early effect of the drug treatment on the inhibition of downstream pathways, cells overexpressing FGFR1wt and FGFR1N546K were incubated for 2 hours in serum-free medium in presence of AZD4547 (0.1 μM) or vehicle (DMSO).
Total protein extracts were analyzed by western blotting and phosphorylation levels of FGFR1, STAT3, ERK and AKT were evaluated in relation to their respective total protein quotas. β-Actin protein levels were used as loading control (Fig. 5A, B).
In both cell lines overexpressing FGFR1wt, AZD4547 0.1µM decreased phospho-FGFR1, phospho-ERK and phospho-AKT protein levels, while did not strongly decrease phospho-STAT3 protein levels (Fig. 5A, B).
In SHSY5Y overexpressing FGFR1N546K, AZD4547 did not show efficacy to decrease phospho-FGFR1, phospho-ERK, phospho-AKT and phospho-STAT3 protein levels, that remained aboundant in cells (Fig. 5A). In SKNBE2 overexpressing FGFR1N546K, although AZD4547 0.1µM decreased phospho-FGFR1 and phospho-ERK levels, phospho-AKT levels were not affected and phospho- STAT3 levels resulted even enhanced (Fig. 5B).
In line with western blotting results (Fig. 5A, B), AZD4547 0.1µM treatment, by impairing FGFR1 signaling, led to a reduction by almost 50% of invasive capacity (Fig. 5C and S4A) and colony number (Fig. 5D and S4B) in both SHSY5Y and SKNBE2 FGFR1wt overexpressing cells compared to untreated cells.
In SHSY5Y FGFR1N546K overexpressing cells AZD4547 0.1µM treatment, that increased phospho-ERK levels and unaffected phospho-FGFR1 and phospho-AKT levels as previously shown (Fig. 5A), did not strongly impair cellular invasion (Fig. 5C and S4A) and neurospheres formation capability (Fig. 5D and S4B). On the other hand, we observed an increase in cellular invasion capacity (Fig. 5C and S4A) and in colony number (Fig. 5D and S4B) in SKNBE2 FGFR1N546K overexpressing cells, probably due to STAT3 and AKT phosphorylation (Fig. 5B).
Altogheter, these data suggest that AZD4547 abolishes the pathway activation induced by FGFR1wt, but does not show a great effectiveness on those ehanced by FGFR1N546K. Hence, N546K mutation may establish a resistance to AZD4547 treatment through activation of AKT and STAT3 pathways.
Targeting of FGFR1N546K signaling by combination treatment with AZD4547 and GDC0941 decreases crosstalk pathways activation, invasion and neurosphere formation capability.
Since AZD4547 alone resulted non-effective in the abolishment of FGFR1N546K induced cross-pathways, we choosed to use it in combination with GDC0941, a PI3K inhibitor already used in clinical trials 53;54.
As previously done for AZD4547, we firstly tested different concentrations of GDC0941 (0.01 μM, 0.1 μM, 1 μM and 10 μM) alone in both cell lines transiently transfected with FGFR1wt and FGFR1N456K by performing cell viability assay (Fig. S3B). Differently from AZD3547, GDC0941 IC50 was higher in SKNBE2 cells (Fig. S3A, B).
Based on the IC50 results, we choosed to test the combination of AZD4547 (0.1 μM) and GDC0941 (0.1 μM and 1 μM) on cell viability, and we selected the lower concentrations able to decrease viability up to 20% (Fig. S3C). Particularly, we used two GDC0941 concentrations (0.1 μM and 1 μM) since GDC0941 has shown lower toxicity in SKNBE2 (Fig. S3B, C).
To investigate the early effects of the combination treatment on the inhibition of downstream pathways, cells overexpressing FGFR1wt and FGFR1N546K were incubated for 2 hours in serum-free medium in presence of AZD4547 (0.1 μM) and GDC0941 (0.1 μM and 1 μM) or vehicle (DMSO).
Our aim was to investigate if these combinations at low doses could be more effective than AZD4547 single treatment in NB cells overespressing FGFR1N546K.
The transfected cells were treated with single GDC0941 (0.1 µM or 1 µM) to test drug efficiency. In FGFR1N546K overexpressing cells treated with GDC0941 alone, we observed a significant decrease only in phospho-AKT protein levels (Figure 5A, B).
In cells overexpressing FGFR1wt, the combination treatment with AZD4547 (0.1µM) and GDC0941 (0.1µM or 1µM) was not effective to decrease both phospho-STAT3 and phospho-ERK protein levels, which in contrast showed an increase probably due to a compensation mechanism following the inhibition of FGFR1 signaling (Figure 5A, B). Of note, in both cell lines overexpressing FGFR1N546K, the combination of AZD4547 0.1µM and GDC0941 1µM showed the best in vitro efficacy for the inhibition of all the three examinated pathways, highlighted by the reduction of phosphorylated/total protein levels (Figure 5A, B).
In SHSY5Y cells overexpressing FGFR1wt protein, AZD4547 0.1 µM and GDC0941 1µM combination, compared to AZD4547 single treatment, showed a lower reduction in cell invasion capability (Fig. 5C and S4A) and a decrease of colony number higher than 50% (Fig. 5D and S4B), probably due to increment of phosoho-STAT3 and a strong decrease of phoshpo-AKT levels, respectively (Fig. 5A). In SKNBE2 overexpressing FGFR1wt, the combination and AZD4547 single treatment showed a similar effect on cell invasion (Fig. 5C and S4A) and colonigenic (Fig. 5D and S4B) capacity, as result of similar downstream pathways activation (Fig. 5B). Interesting to note, in both FGFR1N546K overexpressing cell lines treated with AZD4547 0.1 µM and GDC0941 1µM we observed a reduction of over 50% of invasion and neurosphere capacity, as consequence of downstream pathway impairments aboved mentioned (Fig. 5A-C and S4A, B).
Together, these results highlight that AZD4547 0.1 µM and GDC0941 1µM combination treatment was able to decrease the activation of downstream pathways, cell invasion and neurosphere formation abilities enhanced by FGFR1N546K overexpression in NB cells.
Therefore, AZD4547 and GDC0941 combination treatment may represent a promising therapeutic strategy to overcome the resistance mechanisms induced by FGFR1 N546K mutation under AZD4547 treatment alone.