Participants and clinical research studies

Seventeen individuals (age 12–76 years) with pathogenic or likely pathogenic FHL1 variants underwent detailed clinical evaluation between July 2014 and November 2018 at the NIH Clinical Center (n = 12) or their local clinic (n = 5). Medical records, muscle imaging (if available), muscle biopsy slides, cardiac studies (electrocardiogram, echocardiogram), pulmonary function testing, and serum creatine kinase values were retrospectively reviewed. Genetic testing (direct FHL1 sequencing or as part of diagnostic panels) was performed as part of standard clinical care. Variant classification was based on American College of Medical Genetics criteria¹⁷ (Table S1). Additional imaging studies (muscle ultrasound, muscle MRI including Dixon, and cardiac MRI) were performed as part of a research protocol at the NIH Clinical Center.

Ethical standards

Research studies were performed after informed consent (and assent when appropriate) as part of the following IRB-approved research studies: National Institute of Neurological Disorders and Stroke (NINDS) protocol #12-N-0095 and National Heart, Lung, and Blood Institute (NHLBI) protocol #02-H-0090. Clinical photographs were taken after a separate informed consent/assent for clinical photos was obtained.

Muscle ultrasound

Ultrasound scans were performed in 11/17 patients, using an Acuson S2000 upgraded to S3000 (Siemens Healthineers) with a 6–18 MHz linear array probe. Transverse ultrasound images were acquired by optimal perpendicular imaging, using the bone echo as reference, if visible. Tissue compression was avoided by observation of skin outline on the image. Axial scans of mid-thigh (rectus femoris, vastus lateralis, sartorius, gracilis and hamstring muscles), upper arm (deltoid, biceps and triceps muscles) and lower leg (tibialis anterior, soleus, and medial and lateral gastrocnemius) were obtained when feasible. The echogenicity was graded based on a modified Heckmatt scale¹⁸ with grade 0 for normal appearance, grade 1 for mild increase in echointensity with distinct bone echo, grade 2 with a moderate increase in muscle echointensity and reduced bone echo, and grade 3 with severe increase in muscle echointensity and loss of bone echo. Two independent raters trained in neuromuscular ultrasound graded each muscle and any discrepancies were resolved after a discussion. In muscles that had a sectorial or “geographical” pattern of involvement (i.e. adjacent areas with different degrees of echointensity), an overall score was given based on the total affected area and rater's overall judgment.

Skeletal muscle MRI

Muscle MRI was performed using conventional T1-weighted spin echo and short tau inversion recovery (STIR) of the lower extremities (N = 14/17) and upper extremities (N = 6) on different clinical scanners: Siemens Verio 3.0T (N = 6), Siemens Aera 1.5T (N = 5), Philips Achieva 3.0T (N = 2), and GE Healthcare 1.5T (N = 1). Non-contrast images were obtained from upper arms, forearms, pelvis, thighs, and lower legs in the axial plane. Slices were 5–10 mm thick. The gap between slices ranged between 8 and 16 mm thick. The lower extremity images included the hips, thighs, and lower legs to the ankles in all but one patient (F8P15 only tolerated imaging of hips to knees) who underwent these studies.

Cardiac MRI

Cardiac MRI was performed on six patients on a 3T Siemens Skyra scanner and reviewed by expert cardiologists. Late gadolinium enhancement images were acquired in five patients (F1P1, F1P2, F1P4, F3P6, F5P10) with an intravenous gadolinium-based contrast agent (gadobutrol 0.15 mmol/kg) while the sixth (F2P5) only underwent a non-contrast cardiac MRI. The heart was imaged in standard cardiac imaging planes, consisting of four, three, and two-chamber long axis as well as ventricular short axis views. ECG-gated steady-state free precession cine imaging (N = 5) or real-time cine imaging (N = 1) was used to assess left ventricular mass, volume, and systolic function by tracing epicardial and endocardial borders in end-systole and end-diastole. Native tissue characterization sequences included T1 mapping (N = 6) and T2-weighted black blood imaging (N = 4). In the five patients who received gadolinium contrast, T1-weighted inversion recovery early (i.e. less than approximately 10 min after injection) and late (i.e. starting approximately 10 min after injection) gadolinium enhancement (EGE, LGE) imaging was performed. Tissue characterization images were qualitatively evaluated to assess for the presence and location of myocardial fibrosis (LGE, T1 mapping) or myocardial edema (EGE, T1 mapping, T2-weighted black blood imaging).

Muscle MRI quantification

Fat fraction for each muscle image slice was plotted in violin plots indicating the mean and variability of fat fraction across the length of each muscle and in comparison to the contralateral side.

Clinical phenotype, genotype, and muscle histological findings

We identified 17 patients from nine independent families with genetically confirmed FHL1-related RBM (Fig. 1A). All affected individuals had missense variants in the FHL1 gene, exclusively situated in exon 5 (NM_001449.5), which encodes the LIM-2 domain of the protein (Fig. 1B, Table S1). Similar to prior reports,^{1, 19} substitution of the highly conserved cysteine or histidine residues of the LIM-2 domain were commonly pathogenic and associated with severe disease (Fig. 1C). In two independent families (F4 and F7), we identified a novel variant in the same exon [FHL1 NM_001449.5; c.356G>T; p.(G119V)] that was associated with a milder phenotype. Male patients with this variant (F4P9, F7P13) manifested with progressive scapuloperoneal weakness in late adulthood, and a female individual did not report subjective symptoms at 70 years of age (F4P8). She was found to have presymptomatic disease, however, with mild proximal and distal weakness upon detailed examination (Table 1, Text S1).

Clinical manifestations of RBM in our cohort were compatible with previous reports.¹⁹ All patients had normal development in the motor domain as infants and young children; however, age at symptom onset was variable, ranging from six to 60 years of age (Table 1, Text S1). The late adult-onset cases in our cohort were similar in presentation and progression to a previously reported scapuloperoneal presentation associated with the FHL1 p.W122S pathogenic variant.²⁰ Muscle weakness was focal and asymmetric at onset, often with an initial scapuloperoneal distribution. These symptoms were unrelenting in progression, ultimately resulting in a confluent pattern of weakness involving both distal and proximal muscles. Severe, often asymmetric and progressive large joint contractures in the elbows, neck, hips, knees, and ankles were a common clinical feature (Fig. 1D). Early onset and progressive respiratory insufficiency preceded loss of ambulation in most cases and typically occurred without severe scoliosis, reflecting the early involvement of diaphragm and other respiratory muscles (Table 1).

The phenotypic variability was in part FHL1-variant specific. For example, F1P3 and F2P5 are two women in their mid 30s with two different FHL1 variants involving the same residue (C126G for F1P3, C126S for F2P5). However, F2P5 used a motorized chair with a severely reduced forced vital capacity (23% predicted) while F1P3 was ambulatory and reported no subjective symptoms. The observed phenotypic variability also depended on the sex of the affected individuals: male individuals presented earlier and progressed faster than female individuals, even when carrying the same FHL1 variants, consistent with the X-linked dominant mode of inheritance of RBM. Illustrating this point, in contrast with their affected male relatives, three female individuals that carried FHL1 variants (F1P3, F3P7, and F4P8) did not report subjective symptoms and were found to have proximal and/or distal weakness only upon detailed clinical evaluation. EMG was obtained in a few patients and typically was consistent with an irritable myopathy with abnormal spontaneous activity in the form of fibrillation potentials and positive sharp waves, and early recruitment of small amplitude and short duration motor unit action potentials (Text S1). Serum creatine kinase levels were moderately elevated, typically 2.5–10 times the upper limit of normal range, including in one of the presymptomatic females (Table 1).

Muscle biopsies showed chronic myopathic changes with a distinct, focal pattern of protein aggregate formation. For example, fascicles of normal appearing muscle were found adjacent to severely diseased fascicles with severe myofiber atrophy and preponderance of cytoplasmic aggregates (Fig. 2A and B). Some of the cytoplasmic aggregates spontaneously reacted with the menadione-nitroblue tetrazolium (NBT) stain, consistent with sulfhydryl-rich reducing bodies (Fig. 2C), the pathologic hallmark of RBM. Ultrastructural studies identified electron-dense reducing bodies neighboring typical cytoplasmic bodies with a perinuclear and cytoplasmic distribution (Fig. 2D).

Muscle ultrasound

Muscle ultrasound was obtained in 11/17 patients and showed widely variable changes in echogenicity of skeletal muscle, partly reflecting the range of clinical severity and disease duration in our cohort (Fig. 3A). Illustrating the high sensitivity of muscle ultrasound, two clinically asymptomatic females with FHL1 variants who underwent ultrasound (F1P3 and F4P8) had mildly abnormal patterns of muscle echogenicity (Fig. 3A and B). Consistent with the focal histologic findings, the echogenicity pattern was also focal and “geographical” in most affected muscles, with neighboring areas within the same muscle showing strikingly different echointensities. In the biceps muscle, the increased echogenicity appeared more severe in deeper fascicles (that is, nearer to the humerus). This pattern was consistent with findings on upper extremity T1 axial MRI images (F1P1, F1P2) (Fig. 3C). This conspicuous differential involvement within defined areas of muscles was also noted in the tibialis anterior muscle and was a common finding across the cohort (Figure S1). In addition, in two patients with repeat ultrasound studies ~2.5 years apart, we noted increased echogenicity in a few muscles over time, in correlation with overall disease progression and functional deterioration (Fig. 3D).

Muscle MRI and fat fraction quantification

Similar to ultrasound studies, T1 axial MR imaging of the lower extremity muscles including the hips, thighs, and lower leg to the ankles, obtained in 13/17 patients (F8P15 unable to tolerate lower leg imaging and F3P7, F9P16, F9P17 without any lower extremity imaging), demonstrated variable degrees of contractile tissue remodeling, as reflected by areas of T1 hyperintensity, in correlation with overall clinical disease severity (Fig. 4A, Figure S2). When assessing selective involvement of muscles, we observed a relative sparing of gluteus maximus in the pelvis, the anterior thigh compartment (except for vastus intermedius) compared to hamstrings, gracilis muscle compared to hamstrings and adductors, and lateral gastrocnemius compared to medial gastrocnemius muscle (Fig. 4A). Side-to-side asymmetry, focal signal change, and heterogeneity within individual muscles were conspicuous features. In addition, short tau inversion recovery (STIR) sequence identified areas of increased signal without corresponding T1 hyperintensity (Fig. 4B). While the exact histopathologic correlate of these areas is not clearly understood, they are thought to represent areas of disease activity, tissue edema, or inflammation in myopathies.²¹ Two individuals (F1P1, F1P2) had serial MRIs obtained ~2.5 years apart that showed expanding areas of T1 hyperintensity in several muscles in parallel with worsening weakness and loss of motor function during this period (Fig. 4C).

The heterogeneity and focal pattern of T1 signal hyperintensity within individual muscles and along their length as a characteristic feature of the disease made a qualitative grading of fat fraction on conventional T1 images difficult and unreliable. Thus, we resorted to Dixon imaging, albeit only in a subset of patients that had this modality available, to analyze fat fraction along the length of each muscle. Thigh muscle fat fraction quantification was performed in 6/17 patients using Dixon images of rectus femoris, vastus intermedius and vastus lateralis muscles in the thigh and lateral gastrocnemius, medial gastrocnemius, and soleus in the lower leg (Fig. 5). Side-to side asymmetry of muscle involvement and intramuscular variability of fat fraction in individual muscles was notable and consistent with the focal onset and progression of the disease.

Cardiac function assessments

Echocardiographic (N = 17) and electrocardiographic (N = 15) studies were largely without clinically important findings in this cohort (Table 2) except for one patient (F7P13) with left bundle branch block on ECG and left ventricular hypertrophy with mildly reduced global systolic function (left ventricular ejection fraction of 47%) on echocardiogram. This patient did not undergo cardiac MRI and later passed away at ~80 years of age. Six patients (F1P1, F1P2, F1P4, F2P5, F3P6, and F5P10) underwent cardiac MRI studies. Global left ventricular systolic function was normal in all six patients. Five patients (all except for F2P5) also underwent post gadolinium contrast cardiac MRI studies, and among these, one (F1P4) had possible myocardial edema based on a combination of early gadolinium enhancement, T1 mapping, and T2-weighted black blood images. However, there was no evidence of cardiac fibrosis on late gadolinium enhancement images (Figure S3), and these findings did not have any clinical correlates. None of the other patients had evidence of myocardial disease or myocardial fibrosis on late gadolinium enhancement images.