Falling levels of 4 proteins in CSF may predict response to Spinraza
Changes in biomarkers seen as key in choosing, maintaining an SMA therapy
A drop in the levels of four proteins in the cerebrospinal fluid (CSF) may be a marker of response to treatment with Spinraza (nusinersen) in people with spinal muscular atrophy (SMA), a real-world data study suggests.
The predictive value of the four-protein panel — which includes neurofilament light chain, a known marker of nerve cell damage — was maintained regardless of patients’ age, motor impairment at treatment start, and type of SMA.
This exploratory study in a range of patients “showed that most proteins had a reduction in CSF concentration after 6 months of [Spinraza] treatment,” with four likely to “predict motor improvement at 2 years,” the researchers wrote in the study, “Cerebrospinal Fluid Proteomic Changes after Nusinersen in Patients with Spinal Muscular Atrophy,” published in the Journal of Clinical Medicine.
Spinraza treats SMA by increasing levels of a working SMN protein
SMA is caused by a lack or insufficient levels of the survival motor neuron (SMN) protein, essential for motor neuron health.
Spinraza, by Biogen, is approved to treat all types of SMA and works to restore SMN levels. It is given by intrathecal (spinal canal) injection directly into the CSF, the liquid that surrounds the brain and spinal cord.
Shorter disease duration and better function before starting with Spinraza are associated with better outcomes, especially when treatment begins before SMA symptoms develop.
“This underscores the importance of timely diagnosis, but also the emerging need for predictive tools to monitor disease activity and response to treatment, and to identify which patients are likely to respond to additional treatment in order to reduce motor neuron loss and optimize outcomes,” the researchers wrote.
Analyzing proteins in the CSF can help in understanding how Spinraza influences disease progression and in spotting indicators early that mirror disease activity and treatment impact. Since the CSF directly interacts with affected motor neurons, changes in proteins there may align with likely progression and patient response to disease treatment.
But “in many SMA patients, [Spinraza’s] therapeutic benefits are documented by slowing or halting their rate of motor decline, rather than by improved function, making it impossible to differentiate ‘responders’ from ‘non-responders,’” the team noted.
Recent studies identified some promising CSF biomarkers of Spinraza response, highlighting their potential to predict motor gains with treatment. However, real-world clinical studies across patients of various ages, disease types, and functional levels are needed to capture more effectively CSF protein changes with treatment, and to pinpoint meaningful markers of response.
Study into Spinraza responders in range of patients by age and SMA type
Researchers at Stanford University’s School of Medicine conducted a retrospective analysis of CSF samples from 49 SMA patients, ranging from 3 months to 65 years old, treated with Spinraza between 2017 and 2018 at their center.
A majority had later-onset SMA — 18 with SMA type 2 and 21 with type 3 disease — and 10 had infantile-onset or SMA type 1.
CSF samples were taken immediately before starting treatment (a baseline measure) and again before each Spinraza injection (four loading doses in the first two months, then one injection every four months). CSF protein content was analyzed in samples taken at baseline and six months after starting treatment.
In parallel, patient’s motor function was assessed at baseline and repeatedly during up to two years of follow-up. The Children’s Hospital of Philadelphia Infant Test of Neuromuscular Disorders (CHOP-INTEND), the Hammersmith Functional Motor Scale-Expanded (HFMSE), the 6-Minute Walking Test (6MWT), and the Revised Upper Limb Module (RULM) all were used to assess motor abilities.
For each of these scales, higher scores indicate better motor function.
A supervised machine learning program — a form of artificial intelligence that allows a computer to adapt, analyze, and infer from patterns in data — was used to identify proteins with a potential to predict patient response to Spinraza.
A total of 31 patients showed motor function improvements after two years of Spinraza treatment. Among them, 18 showed an increase in CHOP-INTEND scores, rising from a median of 23.5 at baseline to 30.5 at two years. Ten patients showed improvements in HFMSE scores, from a median of 11 at baseline to 13; 19 saw improvements in RULM scores and two in the 6MWT.
Proteomics analysis reliably identified 595 peptides, small chains of amino acids, which are the building blocks of proteins. Levels of a majority of the proteins detected in the CSF were lower after six months of Spinraza’s use, with 43 found to be significantly lower.
All of these proteins were linked with nerve cells processes in the CNS, and a significant reduction at six months of treatment was seen in four particular proteins: arylsulfatase B (ARSB), ectonucleoside triphosphate diphosphohydrolase 2 (ENTPD2), neurofilament light chain (NfL), and interferon-gamma-inducible protein 30 (IFI-30).
Determining response important with different treatments available
According to the machine learning program, changes in the levels of these four proteins predicted with 79.6% accuracy those patients with motor improvements after two years of treatment. This prediction had a sensitivity of 80.6% and a specificity of 77.8%.
Sensitivity (true positive results and false negative ones) and specificity (true negatives and false positives) are used to determine the likelihood of a given test’s usefulness.
However, the levels of each protein — examined individually — did not associate significantly with motor improvement.
NfL has been reported to be a marker of response to Spinraza in children with SMA in early disease stages. The protein is released into body fluids following nerve cell damage.
“Neurofilaments have received considerable interest as markers of disease severity and treatment response in SMA,” the researchers wrote. “In our study, [NfL] concentration was significantly decreased in the whole cohort after 6 months of treatment, and [NfL] was identified within the 4 top proteins with the highest predictive ability for motor improvement at 2 years.”
ARSB is a protein found in lysosomes, cellular compartments responsible for breaking down and recycling molecules.
IFI-30 is an enzyme believed to facilitate the degradation of proteins in lysosomes, and to be involved in the processing of antigens, proteins that can trigger an immune response.
ENTDP2 helps to regulate chemical communication between neurons, using signaling molecules called purine nucleotides.
Overall, “the use of machine learning predictive models that can integrate biomarker and clinical data is a promising avenue to guide treatment decisions in SMA and to optimize individual patient outcomes,” the researchers wrote.
With three disease-modifying SMA treatments available — Zolgensma and Evrysdi, in addition to Spinraza — “patients and clinicians are faced with unprecedented therapeutic dilemmas as to which medication to start and how long to continue before concluding that additional treatment is needed,” they noted. “Hence, the prospect of using CSF data for 6 months to help guide ongoing management has broad potential.”
Further study that validates this work’s findings in “in larger, independent, and diverse datasets will enhance the model’s predictive ability and help confirm the results presented here,” the scientists concluded.
About the Author
