Identification of FGF14 expansions from genome data
We used STRling to identify STR expansions in short-read genome data of 80 patients with neurological disorders (76 from Germany and 4 from Spain), including 48 patients with cerebellar ataxia from 39 independent families (Fig. 1). This analysis detected outlier values associated with significant q values indicating possible FGF14 AAG repeat expansions (chr13(hg38): 102,161,567 − 102,161,726) in 22 of the 48 patients with ataxia (19/39 families; 49%). Significant values were detected for both individuals of four families for which genome data were available. In contrast, only three of the 32 individuals with other neurological disorders (3/31 families; 10%) showed significant outlier values. In addition, STRling detected an expansion of another motif (AAGGAG) at the same locus in one family from Spain.
We set up LR-PCR and RP-PCR assays to validate these results. The existence of at least one large FGF14 AAG allele (PCR product ≥ 700 bp; triplet repeat number ≥ 180) was confirmed for 18 of the 26 individuals, all with cerebellar ataxia. The eight remaining individuals, including the three with another neurological disorder, had a larger allele below 650 bp (i.e. ≤160 repeats) and were therefore considered negative for FGF14 expansion/SCA27B. Of note, we also confirmed that all patients that had no outlier values detected by STRling only had small FGF14 alleles (Supplementary Fig. 1; Supplementary Data 1).
Screening of FGF14 expansions in a second cohort
We then used the LR-PCR and RP-PCR assays to analyze 106 additional patients (95 new index cases and 11 family members) with cerebellar ataxia (Fig. 1, Fig. 2, Fig. 3A and Supplementary Fig. 2). Twenty-seven of the 95 index cases (28%) and five family members had at least one AAG allele ≥ 700 bp. Taken together with the expansions detected from genome data, 49 individuals from 40 families out of 134 independent families analyzed (30%) had an expanded AAG allele (≥ 180 repeats) estimated from fragment size or gel analysis. The AAG expansions segregated with the disorder in all families including at least two affected individuals available for genetic analysis (Fig. 2; Supplementary Fig. 3A). Four patients from three families had an expansion of the AAGGAG motif (Supplementary Fig. 2B). The AAGGAG expansion segregated in two affected individuals from a Spanish family but was present in only the index case but not the affected mother of a German family (Supplementary Fig. 3B) and was further considered as non-pathogenic, as already reported6,7,17,18.
Targeted Nanopore Sequencing of FGF14 expanded alleles
Since LR-PCR and RP-PCR failed to provide the precise count of AAG repeats at the FGF14 locus, we developed a nanopore sequencing assay of LR-PCR amplicons. In parallel, we analyzed and compared the distribution of FGF14 alleles in a control cohort composed of 802 subjects. In total, we sequenced FGF14 alleles in 59 patients and 64 control individuals (Fig. 1; Fig. 3B-C). We observed a strong correlation between allele sizes determined by fragment analysis or gel electrophoresis and the median number of repeats calculated from nanopore data (Fig. 3D). Combining fragment analysis and nanopore sequencing of LR-PCR amplicons provided a comprehensive view of allele sizes in both patients and controls.
Overall, thirty-eight patients from 31 families (31/134; 23%) had at least one allele composed of pure AAG repeats exceeding the current established threshold for pathogenicity (250 repeats) and were considered as having SCA27B (Fig. 2). The median number of repeats calculated from nanopore reads ranged from 254 to 937 AAG repeats (Supplementary Data 2). Twenty-six patients had a repeat number above 300 repeats (Fig. 2A) whereas 12 patients had a repeat number between 250 and 299 (Fig. 2B). Nevertheless, we noted that the thresholds of 250 and 300 repeats appeared somewhat arbitrary. For instance, in one family, two sisters with similar symptoms and age at onset (AAO) had median repeat numbers of 401 and 281 repeats, respectively (Fig. 2A). Similarly, in another family, the index case had 258 repeats and his affected mother had 224 repeats (Fig. 2B).
Five patients from four independent families showed biallelic repeat expansions (Fig. 3A-B). In three families, the largest allele was below 300 repeats and the lowest allele below 250 repeats: 222/278, 196/284, 165/272. The fourth family included three affected siblings; two siblings had 204/311 and 196/319 repeats whereas their affected sister only had one large pathogenic allele (325 repeats). One patient repeatedly exhibited two large alleles, one with ≥ 300 repeats and another between 250 and 300 repeats, in addition to a small allele, suggesting somatic variability of the expanded allele (Fig. 3A-B). More generally, a high degree of somatic mosaicism around the mean value was detected for all individuals with repeat expansions, as highlighted by the positive correlation between the standard deviation and the allele size (Fig. 3E).
Distribution of FGF14 alleles in patients and controls
The 802 control subjects showed an overall different distribution of FGF14 alleles compared to patients with ataxia (Fig. 4A-B), especially when considering only the larger allele (Supplementary Fig. 5B; Mann-Whitney, p = 0.0002). Large alleles composed of pure AAG ≥ 180 repeats were enriched in patients with ataxia with alleles ≥ 250 repeats showing a more significant enrichment than intermediate alleles (Fig. 4C-D). Conversely, AAGGAG expansions were more frequent in controls (Fisher’s test: p = 0.02, OR 3.8; Fig. 4A). We observed true AAG interruptions (disrupting repeats in the middle) in smaller alleles only, while interruptions limited to 3’ or 5’ sides of the repeats were equally frequent in both alleles (Fisher’s test: p = 0.47) but more frequent in control subjects than patients with ataxia (Fisher’s test: p = 0.01, OR 4.2; Fig. 4; Supplementary Fig. 4). Out of 21 control subjects who had at least one allele above 250 repeats, 13 were composed of the non-pathogenic AAGGAG repeat motif (Supplementary Fig. 4; Supplementary Fig. 5). Eight control individuals had repeat expansions ≥ 250 AAG repeats and only two, aged 46 years old and 70 years old at the time of sampling, had a pure AAG repeat expansion above 300 repeats (313 and 319 repeats respectively; Supplementary Data 2).
Variability of the flanking region
Interestingly, the 5’ region flanking the repeats (3’ region in the context of the gene) was highly variable and drastically differed in expanded versus non-expanded alleles (Fig. 5A). We observed seven different sequences following a constant CTTTCT motif (chr13:102,161,558 − 102,161,563) upstream of the repeats. These 5’-flanking sequences were either directly followed by AAG repeats, or preceded by short AG-rich sequences (e.g., AAGAAAGAG or AAGAG) that we considered as ‘pre-repeat’ (Fig. 5B; Supplementary Fig. 6A-B). Pre-repeats were more frequently observed in larger (a2) alleles (Fisher’s test: p = 3.227×10− 8, OR 7.5; Fig. 5C; Supplementary Fig. 6C-D). The variable GTG sequence (present in the hg38 reference genome) was the most frequently associated with FGF14 expansions although the range of repeats observed was highly variable (range: 36–512). GG and GGG sequences were only detected in expanded alleles (326–937 repeats). Conversely, the GTTAGTCATAGTACCCC was strikingly associated with small alleles (9–21 triplet repeats) only. Interestingly, one nearly identical sequence, differing only in the final two nucleotides (GTTAGTCATAGTACCAG), was associated with 203/207 repeats in both affected individuals of family SPG-79. This suggests that the four consecutive cytosines at the end of the sequence play a key role in preserving the stability of the adjacent repeats.
Intermediate FGF14 alleles
Eleven patients exhibited FGF14 AAG repeat expansions ranging from 180 to 249 repeats (Supplementary Fig. 3A). According to the current threshold of 250 repeats for SCA27B diagnosis, these patients are classified as negative. For instance, in family E19-0805, the affected mother (M79607) had a median repeat value of 224, prompting further investigation into alternative genetic causes for her ataxia. Despite analyzing genome data from both family members, no other ataxia-associated variants were identified, suggesting the FGF14 expansion as the primary culprit. Our exploration revealed that alleles below 250 repeats could contribute to disease, as evidenced by segregation in a Spanish family with two affected individuals (SPG-79; Supplementary Fig. 3A). In another case (M91143), a female patient harbored a variant of unknown significance in PUM1 (NM_001020658.2: c.2180T > C; p.(Ile727Thr)) alongside 236 AAG repeats in FGF14. The remaining families had no identified variants in ataxia-associated genes. Among patients without genome data, one male patient (M87909) had a pathogenic SCA6 expansion (21 repeats) alongside 209 repeats in FGF14. This suggests the potential involvement of additional genetic or non-genetic factors in the disease manifestation, with FGF14 intermediate alleles likely acting as susceptibility factors.
Clinical comparisons
We divided patients with cerebellar ataxia into four distinct groups for clinical comparison: 1) patients with a median number of AAG repeats ≥ 300 (n = 26); 2) patients with 250–299 median repeats (n = 12); 3) patients with an intermediate allele (180–249; n = 10); and 4) patients negative for SCA27B (n = 90; Supplementary Data 3 and 4). Furthermore, we included the clinical data of a German family (affected father-daughter pair) with a novel pathogenic nonsense variant in FGF14 (SCA27A;NM_175929.3:c.239T > G; p.(Leu80*);NM_004115.4(MANE):c.224T > G; p.(Leu75*)) identified by routine exome sequencing (Fig. 2C; Fig. 8A-B).
Overall, most patients with FGF14 expansion ≥ 250 repeats (29/38; 76%) had a highly recognizable phenotype, characterized by the association of slowly progressive cerebellar signs accompanied by episodic symptoms of ataxia and/or downbeat nystagmus (DBN) that often present as first symptoms (Supplementary Data 3). Nine patients exhibited cerebellar symptoms without episodic features or downbeat nystagmus (DBN). Six patients displayed a phenotype with additional signs or a distinct disease course. Detailed case reports of these individuals (clinical outliers; #1–6 on Fig. 7F-G) are available in Supplementary data.
Patients with more than 300 repeats and patients with 250–299 repeats exhibited comparable clinical characteristics that slightly differed from FGF14-negative patients with ataxia. For example, we observed a significantly higher occurrence of early cerebellar oculomotors signs (95%; 92% and 96%, respectively, compared to 63% in FGF14-negative patients). In particular, DBN was present in 47% (50% and 46%) at the first examination, compared to only 3% in patients negative for FGF14 (Fig. 6A). Patients with SCA27B had a lower occurrence of dysarthria on the first examination (24%; 25% and 23%, respectively) compared to 62% in the FGF14 negative group; Fig. 6A). There was less cognitive impairment in patients with FGF14 expansions compared to the FGF14 negative group either at the first or last examinations (71% of negative patients had cognitive impairment versus 38% in the FGF14-positive group; Fig. 6B). These statistical differences remained even when removing the six patients behaving as clinical outliers.
The progression of SCA27B ataxia, assessed through SARA and ICARS scores, was generally slow, with a mean increase in SARA scores of 8.7 points over 30 years (Fig. 7F-G). Excluding the six patients considered as ‘clinical outliers’ resulted in even slower progression (4.5 points over 30 years; Supplementary Fig. 7D-E). However, variability was high in both SCA27B groups (250 ≤ repeats < 300 and ≥ 300 repeats). The linear regression model suggests that disease duration accounts for only 7.9% of observed variance, indicating that other factors exert a more substantial impact. Interestingly, a substantial proportion of SCA27B patients reported worsening symptoms in the mornings (67%; 71% and 64% in FGF14-positive groups compared to 25% in the FGF14-negative group; Fig. 6D). More than half of the patients who received 4-aminopyridine/fampridine reported symptom improvement (57%; 62% and 53%, respectively; Supplementary Fig. 7F-G). Treatment response to acetazolamide was also reported (29%; 0% and 50%, respectively).
The patient harboring the pathogenic nonsense variant in FGF14 (p.Leu80*; M96962) exhibited symptoms very similar to the groups of patients with ≥ 250 repeats. He had a slowly progressive cerebellar syndrome, episodic worsening of cerebellar symptoms and DBN. In addition, he also experienced episodic dystonia of the left hand. While he developed his first symptoms at the age of 36, his daughter had tremors at the age of 5 years and developed gait instability around 30 years.
Correlations between repeat number and age at onset (AAO)
The mean AAO in our combined cohorts was 55.2 years (range 33–76; Table 1; Fig. 7A) in patients with 250–299 repeats and 51.0 years (21–75 years) in patients with ≥ 300 repeats, whereas patients of the negative group had a mean AAO of 46.0 years (1–82 years). This difference was, however, not significant because variation in all groups was high. Seven patients presented with an earlier form of the disease, with an onset before 40 years old (21–38); all but two had ≥ 300 repeats. Accordingly, we observed a significant inverse correlation between the number of AAG repeats and the AAO (Fig. 7D; Supplementary Fig. 7A). Nevertheless, 81% of the variation is independent of the number of AAG repeats (R2 = 0.19). The variability of the AAO was illustrated by patient M90982 (258 repeats) who started to show symptoms at 34 years old whereas individual M91231 (259 repeats) only experienced the first symptoms at the age of 72 years.
Table 1
Mean age at disease onset stratified by FGF14 genetic status.
Patient group | This study | Meta-analysis |
n | Mean AAO | AAO Range | n | Mean AAO | AAO Range |
AAG expansion negative | 90 | 46 | 1–82 | - | - | - |
AAG expansion 180 ≤ exp < 250 | 10 | 53 | 9–72 | - | - | - |
AAG expansion 250 ≤ exp < 300 | 12 | 55.2 | 33–76 | 35 | 62.5 | 33–79 |
AAG expansion exp ≥ 300 | 26 | 51 | 21–75 | 104 | 59.1 | 21–78 |
AAG expansion exp ≥ 250 | 38 | 52.3 | 21–76 | 139 | 60.0 | 21–79 |
Truncating | 2 | 20.5 | 5–36 | 7 | 19.4 | 2–47 |
Missense (p.Phe150Ser) | - | - | - | 24 | 20.5 | 6–40 |
n: number of patients; AAO: age at onset |
We conducted a meta-analysis to assess the correlation between AAO and expansion size, pooling data from our study and four prior studies.7,19–21. We also included data of patients with truncating variants4,5,22,23 or the F150S (F145S in MANE isoform 1) missense variant in FGF143,24 in the comparison (Table 1; Supplementary Data 5). Overall, we confirmed a significant inverse correlation between AAO and expansion size (Fig. 7E; supplementary Fig. 7B). The aggregated data also showed a tendency for patients with FGF14 expansions between 250 and 299 repeats to be later affected on average (62.5 years; n = 37) than patients with ≥ 300 repeats (59.1 years; n = 110), but despite increased statistical power, the difference remains not significant due to high variation in both groups (Fig. 7C). Nonetheless, patients with expansions from both groups exhibited significantly later AAO compared to those with truncating (19.4 years, range: 2–47) or F150S (20.5 years, range: 6–40) variants (Fig. 7C).
Frequency of SCA27B in the German cohort
To assess the prevalence of SCA27B relative to other dominantly inherited ataxia subtypes, we compared the number of patients diagnosed in the ataxia outpatient clinic (Department of Neurology, University Hospital Essen). Among the diagnoses, SCA6 (CACNA1A) accounted for 40 patients, SCA27B for 36 patients, SCA3 (ATXN3) for 35 patients, SCA1 (ATXN1) for 17 patients, Episodic Ataxia 2 (point variant in CACNA1A) for 12 patients, SCA14 (PRKCG) for 7 patients, SCA2 (ATXN2) for 6 patients, and SCA8 (ATXN8OS) for 4 patients. Additionally, rarer forms such as SCA28 and SCA49 (two each), SCA7, SCA13, SCA15, and SCA27A (one each) were observed. This indicates that SCA27B ranks among the most prevalent SCA subtypes, representing approximately one-fifth of all diagnoses in patients with cerebellar ataxia. This finding independently supports the high frequency of SCA27B diagnoses observed in another German cohort.25
Secondary structures associated with FGF14 expansions
We used CD spectroscopy to assess the potential of secondary structure formation of the different FGF14 antisense repeat expansions AAG and AAGGAG as well as the complementary sense sequences CTT and CTCCTT on DNA and RNA. The AAG-DNA 25-mer formed an antiparallel homoduplex while CD spectroscopy of the AAGGAG-DNA oligo revealed formation of a parallel homoduplex26–29(Fig. 8C). The RNA counterparts CUU and CUCCUU did not obtain any secondary structure under the tested conditions, confirmed by a single positive band at 270 nm30 (Fig. 8C). Interestingly, the non-pathogenic AAGGAG-RNA oligo folded into a parallel guanine-quadruplex (G4) with a positive band around 260 nm, a negative band around 240 nm and positive values around 210 nm. Presence of a G4 was further confirmed by a G4-specific decrease in the stability detected, shifting from a parallel G4 structure (with 100 mM K) to a hairpin structure (with 100 mM Li)31,32 (Fig. 8C). Of note, for the pathogenic AAG-RNA repeat we detected a CD spectrum related to an A-form RNA structure, with a negative peak around 210 nm that reflects intra-strand interaction of RNA duplexes33.