Pretreatment blood NfL indicates response to cellular therapies in cerebral adrenoleukodystrophy

Abstract

Background

The neuroinflammatory cerebral form of X-linked adrenoleukodystrophy (CALD) is among the most severe neurological diseases affecting children. Early intervention by hematopoietic stem cell transplantation (HSCT) or gene therapy halts CALD, justifying its inclusion in newborn screening programmes. Currently, eligibility for treatment is assessed using MRI-based Loes and neurological scoring; however, grading poorly differentiates advanced stages or atypical lesion patterns. We recently identified blood neurofilament light chain (NfL) and glial fibrillary acidic protein (GFAP), indicative of axonal and glial damage, as valid biomarkers reflecting CALD activity.

Methods

Here, we investigated how pre-treatment biomarker levels relate to outcome of HSCT or gene therapy in a retrospective cohort of 14 paediatric CALD patients aged 5–13 years. Patients were monitored longitudinally before and for up to 5.5 years following HSCT or gene therapy. Disease progression was assessed using MRI-based Loes scoring and neurological evaluations. Plasma NfL and GFAP were quantified using single molecule array (Simoa) technology.

Results

Here we show that blood NfL and GFAP transiently increase due to myeloablative conditioning and gradually normalize post-treatment. Pre-treatment NfL ≤ 113 pg/ml, but not GFAP, correlate with CALD stabilization or, for Loes >9, only minor progression. In contrast, NfL >243 pg/ml associates with major progression irrespective of baseline Loes score. In three boys with atypical lesions or advanced disease, NfL outperforms Loes-scoring in predicting outcome.

Conclusions

Blood NfL can complement clinical decision-making, particularly in patients with advanced CALD or atypical lesions. These findings are especially relevant for clinical management in countries that have not yet implemented X-linked adrenoleukodystrophy into newborn screening.

Plain language summary

Cerebral X-linked adrenoleukodystrophy (CALD) is a rare but severe brain disease that mainly affects boys. If treated early with a stem cell transplant or gene therapy, the disease can be stopped. However, deciding who will benefit from treatment is difficult, especially when MRI scans show unusual patterns or the disease is already advanced. In this study, we measured two blood biomarkers, neurofilament light chain (NfL) and glial fibrillary acidic protein (GFAP), indicative of nerve and glial injury. We followed 14 boys with CALD for up to 5.5 years after treatment. We found that the blood NfL level before treatment predicted whether the disease would stabilize or continue to worsen. These findings are especially relevant for doctors in countries that have not yet implemented X-linked adrenoleukodystrophy into newborn screening.

Weinhofer et al. examine whether pre-treatment blood neurofilament light chain (NfL) can predict treatment effectiveness in cerebral X-linked adrenoleukodystrophy (CALD). Findings show that blood NfL can complement clinical decision-making, particularly in patients with advanced CALD or atypical lesions.

Introduction

X-linked adrenoleukodystrophy (X-ALD, OMIM #300100) is the most common monogenetically inherited brain white matter disease and has been recently added to newborn screening programmes in several countries^1,2. X-ALD is caused by ABCD1 mutations that lead to impaired peroxisomal degradation and accumulation of saturated very long-chain fatty acids^3,4. The core syndrome of X-ALD in adulthood is adrenomyeloneuropathy (AMN), a slowly progressive dying-back axonopathy in the spinal cord⁵. In more than half of all male X-ALD patients, with peak onset in boyhood, yet unknown genetic, epigenetic or environmental triggers induce a life-threatening progressive leukodystrophy (cerebral ALD, CALD) that cannot be predicted based on the genotype^6–8. CALD is characterized by rapidly advancing inflammatory destruction of myelin and axons with disruption of the blood-brain barrier (BBB) and infiltration of peripheral immune cells⁶. To detect the onset of CALD presymptomatically, patients undergo a burdensome, frequent MRI screening, often under sedation, to visualize inflammatory brain white matter lesions and opening of the BBB by extravasation of gadolinium (Gd)-based contrast agents⁹.

The standard of care for paediatric CALD is hematopoietic stem cell transplantation (HSCT), either allogenic with a suitable donor or autologous with ex vivo lentiviral gene therapy¹⁰. However, the latter approach has recently been associated with an increased risk of hematologic cancer¹¹. Exactly how HSCT halts CALD is not yet completely understood, but strong evidence points to the cerebral engraftment of donor-derived or corrected peripheral macrophages that replace dysfunctional microglial cells¹². Following HSCT, restoration of the BBB is usually evident within 100 days¹³. In contrast, myelin destruction and axonal injury, as semi-quantitatively assessed by conventional MRI, may progress for up to 18 months after the HSCT intervention¹⁴. Accordingly, patients with higher pre-treatment MRI severity scores, indicating a more advanced disease status, generally show higher neurocognitive decline after transplant¹⁵.

Defining the eligibility of CALD patients for treatment is key to enable survival free of major functional deficits^14,16. Currently, CALD progression is graded using MRI-severity Loes scoring, ranging from 0 to 34, whereby higher values correlate with an increasing number of demyelinated structures and atrophy¹⁷. However, this grading system poorly reflects the size and dynamics of demyelinating lesions or patients with atypical patterns^18,19. Therefore, it may be challenging to decide on the Loes score alone who is eligible for cellular therapies, especially in countries without a newborn screening programme. Moreover, Loes' scoring was not intended to monitor treatment success due to its continued increase in successfully transplanted patients with local brain gliosis and subsequent atrophy¹⁰. As such, there is an unmet need for alternative biomarkers to assist clinicians in facilitating treatment decisions, predicting disease outcome, and monitoring disease activity.

We and others recently demonstrated that the blood-derived biomarkers neurofilament light chain protein (NfL), a neuronal scaffolding protein indicating axonal damage, and glial fibrillary acidic protein (GFAP), indicating astrocyte activation or damage, reflect the inflammatory activity and progression of CALD^20–23. So far, NfL shows the best correlation with neuroaxonal damage and brain white matter lesion volume in CALD patients, including decreasing levels upon halt of the disease by HSCT^{20–22,24,25}. However, no data are available on how pre-treatment blood NfL associates with treatment outcome in CALD. Here, we investigated the prognostic potential of pre-treatment blood NfL and GFAP levels for post-treatment response and outcome in transplanted children with CALD. We show that pre-treatment blood NfL, but not GFAP, is associated with treatment response in children undergoing HSCT or gene therapy for CALD. Lower NfL levels are linked to subsequent disease stabilization, whereas higher levels identify patients who remain at risk of progression despite intervention. We also observe that NfL provides useful prognostic information in cases with atypical lesions or advanced MRI severity, outperforming Loes-scoring in these patients. Together, these findings suggest that pre-treatment blood NfL adds clinically relevant information beyond current imaging-based assessments and may support decision-making for children with CALD.

Methods

Study design and participants

The longitudinal retrospective single-centre cohort study was approved by the Ethical Review Board of the Medical University of Vienna (EK1613/2019) and local biobanking by the Ethical Review Board of the University Hospital Leipzig (EK 370/21). Plasma samples were collected after informed consent was obtained from legal guardians and participants. With X-ALD being an X-linked disorder and CALD very rarely affecting females, we included male paediatric X-ALD patients (n = 18, median age = 8.5 years) who were diagnosed based on elevated plasma VLCFAs by gas chromatography—mass spectroscopy and/or (likely) pathogenic ABCD1 mutations. CALD in X-ALD children was defined as progressive demyelinating lesions with or without Gd-enhancement. The brain lesion severity was assessed by MRI scoring according to the scale of Loes¹⁷. The neurologic function score (NFS; range, 0 to 25 points)²⁶ was used to quantify neurological deficits, whereas, the ALD disability rating scale (ALD-DRS; range, 0 to IV levels) was used to assess requirements for assistance²⁷. The longitudinal HSCT group consisted of 11 children (median age = 7 years; median Loes score = 4) and the gene therapy group of 3 children (median age = 9 years; median Loes score = 2). In addition, plasma samples from 4 CALD children up to 9.5 years post-allogeneic HSCT were assessed. Baseline characteristics of paediatric CALD patients are shown in Table 1 and Supplementary Table 1.

Table 1.

Demographics of paediatric CALD patients at baseline and post-treatment

Patient ID	#1	#2	#3	#4	#5	#6	#7	#8	#9	#10	#11	#12	#13	#14
Age range at study inclusion (yrs)	9–13	6–8	6–8	6–8	9–13	6–8	9–13	<6	6–8	6–8	6–8	<6	9-13	9-13
MRI severity (Loes score)	3	3	1	4	2	2	12	6.5	8	15	13	1.5	2	6
Gd-enhancement	Neg.	Neg.	Neg.	Pos.	Pos.	Pos.	Pos.	Pos.	Pos.	Pos.	Pos.	Pos.	Pos.	Pos.
Lesion pattern	PO	PO	PO	PO	BS	PO	PO+BS	PO+BS	PO	PO	PO	PO	BS	FT
NFS	0	0	0	0	0	0	2	0	1	1	2	0	0	0.
ALD-DRS	0	0	0	0	0	0	2	1	1	2	2	0	0	2
Treatment	HSCT	HSCT	HSCT	HSCT	HSCT	HSCT	HSCT	HSCT	HSCT	HSCT	HSCT	GT	GT	GT
Source	BM	BM	BM	BM	PBSC	PBSC	BM	PBSC	BM	BM	BM	PBSC	PBSC	PBSC
Match	10/10	Ident	9/10	10/10	10/10	10/10	9/10	10/10	10/10	Ident	9/10	Auto	Auto	Auto
Follow-up duration (mths)	24	24	6	24	12	36	2/6a	24	66	6	6	60	42	24
NfL (pg/ml)
pre-t	9.9	25.2	6.3	43.6	16.8	113.1	112.1	201.2	104.9		573.3	N.A.	8.3	243.7
post-t (12 m)	7.5	N.A.	N.A	25.3	16.5	109.3	N.A.	140.0	N.A.	797.7	N.A.	N.A.	25.7	332.4
post-t (≥12 m)	7.1	19.7	N.A	6.9	N.A.	9.2	N.A.	16.0	2.9	┼	N.A.	2.8	4.7	134.8
GFAP (pg/ml)
pre-t	129.6	129.0	195.3	394.6	90.8	314.2	N.A.	610.3	265.2	969.5	270.5	N.A.	49.7	179.2
post-t (12 m)	364.7	N.A.	N.A.	227.9	101.4	228.2	N.A.	448.7	N.A.	┼	N.A.	N.A.	227.4	460.7
post-t (≥12 m)	196.4	236.1	N.A.	92.8	N.A.	518.6	N.A.	193.3	322.6		N.A.	512.2	69.2	616.3

MRI magnetic resonance imaging, Gd Gadolinium-enhancement; brain white matter lesions, PO parieto-occipital, FT frontal, BS brain stem, NFS neurologic function score, ALD-DRS adrenoleukodystrophy disability rate scale, HSCT allogeneic hematopoietic stem cell transplantation, GT autologous gene therapy, Source of hematopoietic stem cells, BM bone marrow, PBSC peripheral blood stem cells, human lymphocyte antigen (HLA-)donor match, 10/10 10/10 HLA match (unrelated), 9/10 9/10 HLA match (unrelated), Ident HLA identical sibling, Auto autologous, pre-t pre-treatment; post-t (12 m), 12 months post-treatment, post-t (≥12 m) ≥12 months post-treatment, ┼ death of patient post-treatment.

^aThe follow-up time for patient #7 was 2 months for NfL measurement and 6 months for clinical assessment.

Allogeneic HSCT and gene therapy procedures

Children of the HSCT group (#1 to #11) received allogeneic hematopoietic stem cells (bone marrow n = 8, partially T-cell depleted peripheral stem cells n = 2, unmanipulated peripheral stem cells n = 1) from a ≥ 9/10 HLA-matched donor (sibling n = 2, unrelated n = 9). Myeloablative conditioning consisted of fludarabine (180 mg/kg) and busulfan IV with therapeutic drug monitoring (targeted AUC 75 mg*h/L), together with anticonvulsive prophylaxis over 7 days. For prophylaxis of graft-vs-host disease, patients received serotherapy (ATG Grafalon®), tacrolimus, and mycophenolate mofetil. Patient #11 additionally received cyclophosphamide (50 mg/kg for 2 days) post-transplant. Children of the gene therapy group received LentiD (bluebird bio) in the study ALD-102 (#12) or in the study ALD-104 (#13 and #14) with the respective conditioning protocol²⁸.

MR imaging protocol

For acquiring the scans, a 3.0-tesla MRI system (Prisma^fit; Siemens, Erlangen, Germany) was used. The protocol included T1-weighted, T2-weighted, and fluid-attenuated inversion recovery (FLAIR) sequences covering all three dimensions. Gd was added as contrast enhancement at baseline examination and after treatment, at least until unequivocal resolution of Gd enhancement was confirmed in a follow-up scan.

Blood sampling and biomarker measurement

Blood samples were collected at the University of Leipzig, Germany, into standard polypropylene EDTA collection tubes and processed at room temperature within two h. Plasma was generated by centrifugation at 3.500 × g for 5 min and stored aliquoted, pseudonymized at −80 °C. NfL and GFAP levels were assessed by a blinded investigator using Simoa^® NF-light Advantage Kit (#103400, Quanterix), GFAP Simoa^® Discovery Kit (#102336, Quanterix) and the Simoa® Neurology 2-Plex B (NfL, GFAP) assay (#103520) on an SR-X Analyzer (Quanterix).

Statistics and reproducibility

For statistical analysis, the two-sided non-parametric Wilcoxon matched-pairs signed rank test was used. For association analysis, the Spearmans´s correlation coefficient was calculated. P-values below 0.05 were regarded as indicating statistical significance. Graphs were produced, and statistical results were calculated using GraphPad Prism 8. Boxplots indicate the median ± interquartile range, while whiskers show minimum and maximum. Bar graphs show means with individual data points.

Results

Demographics of the CALD patients

Our longitudinal dataset consisted of plasma samples derived from 14 boys with CALD (median age = 7.5 years, median Loes score = 3.5) with samples taken 29–10 days (median = 12 days) before and at various time points up to 5.5 years after allogeneic HSCT (#1 to #11) or gene therapy (#12 to #14). In addition, plasma was collected from 4 CALD children (median age = 14 years) at 4–9.5 years after allogeneic HSCT (#15 to #18). In all patients, MRI analysis confirmed lesion progression, whereas Gd-enhancement using a standard single-dose condition was not observed in three (#1, #2 and #3) of the 14 longitudinal patients. Baseline characteristics of the participants are shown in Table 1 and Supplementary Table 1.

Pre-treatment levels of NfL but not GFAP correspond with treatment outcome

To explore whether pre-treatment NfL and/or GFAP could complement MRI in predicting treatment outcome for individual CALD patients, we determined NfL and GFAP before and longitudinally up to 5.5 years post allogeneic HSCT or gene therapy in plasma samples of 14 boys with CALD including 4 patients with uncommon lesion patterns (#7 and #8, additional brain stem lesion; #13, brain stem lesion only; #14, frontal involvement) and 3 with advanced demyelination (#7, 10 and #11, Loes scores ≥12, respectively) (Fig. 1). We categorized the treatment outcome as “stabilization”, “minor progression” or “major progression”, according to the NFS after treatment. “Stabilization” refers to patients who do not develop any additional neurological deficits after treatment, “minor progression” refers to the development of new deficits but show survival without major functional deficits (MFDs); and “major progression” was classified as MFDs post-therapy. Before treatment, all CALD patients exhibited elevated NfL levels when compared to median plasma NfL previously reported for 20 asymptomatic boys with X-ALD of similar age²¹ (Fig. 1A). Notably, elevated NfL levels were also observed in patients #1, #2 and #3, who did not show Gd-enhancement on MRI under standard single dose conditions despite ongoing brain lesion progression. When we compared pre-treatment NfL levels with treatment outcomes for individual patients, we found that pre-treatment NfL stratified CALD patients into groups corresponding to later treatment efficacy (Fig. 1A, B). Whereas all children with Loes score <9 and pre-treatment NfL levels of ≤113 pg/ml stabilized after treatment (patients #1 to #6, #9 and #13), patient #8 with a Loes score of 6.5 but pre-treatment NfL of 201 pg/ml experienced minor progression. In contrast, patient #7 with advanced disease (Loes score = 12) but baseline NfL of 112 pg/ml displayed only minor progression within 6 months post bone marrow transplantation. All patients with a pre-treatment NfL > 243 pg/ml (patients #10, #11 and #14) developed MFD irrespective of baseline Loes scores. Six months post treatment, NfL levels in all investigated patients remained at or above pre-treatment levels. In patients with minor or major progression, levels stayed elevated above 100 pg/ml one year after the treatment (Supplementary Fig. 1A, B).

Fig. 1. Longitudinal assessment of blood NfL and GFAP in paediatric CALD.

A, B NfL and C, D GFAP were determined by Simoa digital ELISA in plasma samples of paediatric CALD patients (n = 14, median age = 7.5 years, median Loes score = 3.5) before (pre-t) and up to 5.5 years after receiving HSCT or gene therapy. Treatment outcome (“stabilization”, “minor progression” and “major progression”) was evaluated based on neurological function score. Pre- and post-treatment median NfL and GFAP levels for the age-matched cohort of asymptomatic X-ALD children (NfL, n = 20; GFAP, n = 19) have been described previously in ref. ²¹. Post-treatment time points: D day, M months, Y year. B Pre-treatment (pre-t) NfL and D pre-t GFAP associated with pre-t Loes scores in the paediatric CALD group. The vertical dashed line indicates the recommended MRI-based cut-off for treatment eligibility (Loes score ≤ 9). The associations between NfL or GFAP and the Loes score were quantified using Spearmans´s correlation coefficient. P-values below 0.05 were regarded to indicate statistical significance.

In contrast to NfL, pre-treatment GFAP levels were neither consistently higher in CALD patients when compared to the median GFAP value of the asymptomatic X-ALD group nor were these levels associated with treatment outcomes (Fig.1C, Supplementary Fig. 1C, D). Moreover, there was a better correlation between pre-treatment NfL and MRI severity score than for GFAP levels (Fig. 1B, D), thus confirming our previously reported data²¹.

Transient increase in blood NfL and GFAP indicates neurotoxicity of the myeloablative conditioning in CALD patients

Patients with advanced CALD have been described to deteriorate more rapidly under aggressive chemotherapeutic preconditioning²⁷. Chemotherapy has also been associated with a transient increase in neurotoxicity in multiple sclerosis (MS) patients treated with HSCT²⁹. Thus, we assumed that the CALD patients with marked acute axonal damage, as reflected by high plasma NfL levels (≥201 pg/ml; #8, #10, #11 and #14), may be especially vulnerable to the neurotoxic effects of the conditioning regimen. Accordingly, we compared the NfL levels of these more advanced patients before and 1 month after HSCT or gene therapy to those of earlier-stage CALD children with lower plasma NfL (≤113 pg/ml; #1 to #7, #9, #13, Fig. 2A, B). Intriguingly, we found a significant transient increase in NfL upon conditioning 1 month post-treatment relative to baseline only for the CALD group with relatively low pre-treatment NfL (Fig. 2A). No significant further elevation of NfL was observed 1 month post treatment in the CALD group already presenting high pre-treatment NfL levels (Fig. 2B). Similarly, plasma GFAP was significantly stronger increased 1 month post-treatment in the group of early-stage CALD when compared to the group of more advanced CALD children (Fig. 2C, D). Of note, for both groups of CALD patients, the median transient increase from baseline due to myeloablative conditioning was higher for GFAP than for NfL.

Fig. 2. Plasma NfL and GFAP transiently increase after allogeneic HSCT and gene therapy.

A, B Plasma NfL and C, D GFAP before (pre-t) and 1 month after (M1) receiving HSCT or gene therapy in the cohorts of paediatric CALD patients that stabilized or progressed post-treatment. In (A), patients with pre-treatment NfL levels ≤ 113 pg/ml and in (B), patients with pre-treatment NfL levels ≥ 201 pg/ml are shown. Median NfL and GFAP levels for the cohort of age-matched asymptomatic X-ALD children (NfL, n = 20; GFAP, n = 19) have been described previously²¹. For statistical analysis, the two-sided non-parametric Wilcoxon test was used, and p-values below 0.05 were regarded to indicate statistical significance. Boxplots indicate the median±interquartile range, while whiskers show minimum and maximum.

To assess the long-term stabilization of neuronal injury and astrocyte damage in paediatric CALD, we measured NfL and GFAP in 4 CALD patients who were successfully treated by HSCT 4 to 9.5 years before sampling of plasma (#15, #16, #17 and #18). When compared to plasma NfL and GFAP of the cohort of 20 (NfL) or 19 (GFAP) asymptomatic boys with X-ALD²¹, we found that both biomarkers stabilized at slightly elevated levels 4–9.5 years post-treatment when compared to the age-matched asymptomatic X-ALD cohort (Fig. 3A, B).

Fig. 3. Long-term stabilization of neuronal injury and astrocyte damage in paediatric CALD.

Association of plasma NfL A and GFAP B with age in samples from 4 children with CALD (#15 to #18) at 4–9.5 years post-HSCT (median age = 14 years). For comparison, the values of age-matched asymptomatic paediatric X-ALD patients (NfL: n = 20, total sample number = 40; GFAP: n = 19, total sample number = 38) as reported in ref. ²¹ are shown.

Discussion

With neuroinflammatory CALD being one of the most severe brain disorders affecting children, X-ALD has been added to newborn screening programmes in several countries. Lifesaving allogeneic HSCT or gene therapy can effectively halt CALD progression, but only if applied at early disease stages. Thus, prompt diagnosis of CALD onset and assessment of eligibility for treatment is crucial. We compared in the present study the prognostic value of pre-treatment blood NfL and GFAP, as measured in plasma samples using Simoa technology, in a cohort of 14 childhood CALD patients who received allogeneic HSCT or gene therapy. The data revealed for patients with baseline Loes score <9 that pre-treatment plasma NfL levels of ≤113 pg/ml correlate with stabilized CALD course post-treatment. In contrast, pre-treatment NfL of > 243 pg/ml was associated with major progression of the inflammatory cerebral demyelination, i.e. development of major functional deficits, independent of the baseline Loes score. Further, blood NfL only gradually normalized in CALD patients, a pattern similarly observed in MS^29,30 but contrasting with the more rapid decline seen within 6 months of pharmacological fingolimod or tocilizumab treatment in MS or neuromyelitis optica spectrum disorder, respectively^31,32.

Evaluating treatment eligibility through MRI-based Loes scoring is especially demanding in CALD children with atypical lesion patterns or more advanced disease stages. The latter aspect is particularly relevant for countries where X-ALD newborn screening has not been implemented, and CALD is often diagnosed at later symptomatic stages. With the limitation of a small sample set, we compared the prognostic value of pre-treatment plasma NfL levels and Loes scores for our paediatric CALD cohort. Applying Loes score ≤9 as threshold for treatment eligibility and pre-treatment NfL ≤113 pg/ml to indicate probable stabilization post-treatment, we found that higher pre-treatment NfL correlated with disease progression post-treatment in two CALD patients (#8 and #14) with more advanced disease status and unusual lesion pattern, despite seemingly favourable Loes scores 6.5 and 6, respectively. In contrast, lower pre-treatment NfL of 112 pg/ml in patient #7 with atypical lesions and a high Loes score of 12 resulted in only minor progression post-treatment.

Our finding that myeloablative conditioning using the chemotherapeutic agent busulfan, combined with either cyclophosphamide or fludarabine, transiently exacerbates axonal damage and astrocyte activation, elevating the systemic release of NfL and GFAP in paediatric CALD aligns with previous observations in adult patients with severe MS^29,30. The potential relationship between this transient increase in neurotoxicity and neurocognitive impairment in the already compromised brains of CALD patients remains unexplored. Additionally, it remains unclear whether the accumulation of saturated very long-chain fatty acids, which have been previously linked to pro-inflammatory skewing of immune cells^33,34, also potentiates the susceptibility of brain cells to the neurotoxic effects of chemotherapeutic conditioning. However, the transient increase in NfL may be in part be characteristic of busulfan, since paediatric leukemia patients conditioned with busulfan also display some elevation, albeit to a lesser extent (Weinhofer/Kühl, unpublished data). Therefore, reduced-intensity or non-myeloablative conditioning protocols omitting busulfan might attenuate the chemotherapy-associated neurotoxicity reflected by the transient rise in NfL and GFAP^19,35. As all patients in our cohort received busulfan-based myeloablative regimens, we cannot compare biomarker dynamics between conditioning strategies. However, while myeloablative conditioning may be associated with further injury to the CNS in childhood CALD patients, correction of the defective monocyte/macrophage system by engraftment of the allogeneic donor or corrected autologous myelopoiesis is pivotal. Therefore, the benefits of myeloablation far outweigh the risks by preventing the devastating and ultimately fatal outcome of CALD progression. Furthermore, in early-stage patients, the transitory increase in plasma NfL or GFAP is without clinical correlation.

With X-ALD being a rare disorder (combined male and female incidence 1:14,700) and about 30% of boys below the age of 12 developing CALD⁵, the major limitation of our study is its small sample size, comprising longitudinal samples from 14 children with CALD including 3 patients with advanced demyelination, which reflects the rarity of paediatric CALD and the restricted availability of longitudinally sampled patients undergoing cellular therapies. Moreover, our study included only 3 patients who received gene therapy, while 11 children underwent allogeneic HSCT. It is possible that NfL levels indicating stabilization post-treatment differ between these two approaches, similar to how efficacy appears to be somewhat different³⁶. Detailed neurocognitive follow-up data were not available for all patients, limiting our ability to relate chronic low-grade biomarker elevation to long-term functional outcomes. The persistent low-grade elevation of NfL and GFAP levels observed years after HSCT likely reflect residual effects of prior neuroinflammation but cannot be correlated with neurocognitive trajectories in our dataset. Further, the exhibited heterogeneity in lesion distribution, Gd-enhancement status and MRI severity within this small cohort of patients limits conclusions about biomarker performance across the full spectrum of CALD presentations. Therefore, we could not perform multivariable modelling to adjust for confounders such as baseline Loes score, lesion pattern, age, or conditioning regimen; thus, we cannot establish NfL as an independent predictor of treatment response. Accordingly, the limited cohort size restricts the generalizability and applicability of our findings, which may not fully capture variability in the relationship between pre-treatment NfL levels and treatment outcome in larger populations. Further, the retrospective, single-centre design introduces potential sources of selection and measurement bias, including non-uniform MRI timing and variability in clinical sampling intervals. As a consequence of these limitations, the proposed plasma NfL threshold ranges should be interpreted with caution and validated in larger, multicentre cohorts. Nevertheless, our data provide valuable insights relevant for the clinical management of childhood CALD, especially in countries that have not implemented X-ALD newborn screening and a relevant proportion of patients present with CALD in rather advanced stages, with questionable benefits of cellular therapies.

Author contributions

Conceptualization: I.W. and J.S.K.; Methodology and Investigation: I.W., J.S.K., M.P., P.R., and J.B.; Patient recruitment, sample collection and interpretation of clinical data: J.S.K., W.K., A.M., D.G., S.S., and C.G.B.; writing—original draft: I.W. and J.S.K.; writing- review and editing: I.W., J.S.K., S.F.P., and J.B.

Peer review

Peer review information

Communications Medicine thanks Takashi Matsukawa, Yuri Matteo Falzone and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data are included in the manuscript or in the Supplementary Material and are available from the corresponding authors upon request with the publication of the manuscript. The source data to generate Figs. 1–3 and Supplementary Fig. 1 are provided in Supplementary Data 1.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s43856-026-01525-6.