Reversible white matter changes following a 4‐week high phenylalanine exposure in adults with phenylketonuria

Abstract

Alterations in brain structure are frequently observed in adults with early‐treated phenylketonuria (PKU) compared to healthy controls, with cerebral white matter (WM) being particularly affected. The extent to which temporary elevation of phenylalanine (Phe) levels impacts WM remains unclear. We conducted a double‐blind, randomised, placebo‐controlled crossover trial to investigate the effects of a 4‐week high Phe exposure on cerebral WM and its relationship to cognitive performance and metabolic parameters in adults with PKU. In this study, 27 adults with early‐treated classical PKU (aged 19–48 years) underwent diffusion tensor imaging (DTI) before and after the 4‐week Phe and placebo interventions. Fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) were analysed using tract‐based spatial statistics. Neuropsychological examinations at each timepoint evaluated executive functions and attention. Additionally, brain Phe levels were measured using MR spectroscopy, and blood levels of Phe, tyrosine, and tryptophan were assessed after an overnight fast. Following the Phe period, significant decreases in AD, MD, and RD were observed compared to the placebo period, particularly in the posterior corona radiata and optic radiation. Notably, these WM changes were reversible in patients who first received Phe (n = 13). Cognitive performance and metabolic parameters were not significantly related to DTI scalars after the Phe period. In conlcusion, a 4‐week Phe elevation induced reversible microstructural alterations in cerebral WM. Further investigation is necessary to determine the clinical implication of these changes.

1. INTRODUCTION

Phenylketonuria (PKU), a rare inherited metabolic disorder, disrupts the conversion of the amino acid phenylalanine (Phe). This dysfunction elevates Phe levels in the blood and brain, causing severe intellectual disabilities if dietary management during childhood is inadequate or insufficiently stringent. The primary objective of the PKU treatment is to prevent the manifestation of its debilitating consequences during childhood. This is achieved by lowering Phe concentrations to non‐harmful levels through a low‐protein diet and compensating for the resulting shortage of other amino acids. However, adherence to this strict and complex diet tends to decrease with age. ¹ The currently available pharmacological treatments, sapropterin dihydrochloride (BH4) and pegvaliase, are often either not tolerated or ineffective for most patients. While promising new treatments, such as sepiapterin, ² are under investigation, the low‐protein diet combined with amino acid supplementation remains the cornerstone of PKU treatment.

Although early detection and treatment of PKU prevent cognitive impairments, subtle cognitive alterations may persist. ³ , ⁴ Such cognitive alterations are most frequently noticed in the domains of executive functions and attention. ⁴ It is, however, important to emphasise that cognitive performance in adults with early‐treated PKU is still within the range of normative samples. ⁵

Investigations have indicated structural brain changes among individuals with early‐treated PKU, particularly in cerebral white matter (WM). These imaging findings are commonly observed in the periventricular WM and appear as increased signal intensities (hyperintensities) on FLAIR and T2‐weighted images. ⁶ The WM alterations, prevalent in over 90% of patients, ⁷ are a recognised hallmark of PKU, yet their clinical implications remain uncertain. ⁸ WM abnormalities are observable not only at the macrostructural level, but also quantifiable at the microstructural level. Numerous cross‐sectional studies have demonstrated alterations in the WM microstructure in both children and adults with PKU compared to healthy controls. ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ WM alterations are often widespread and include commissural fibres (e.g., the corpus callosum), ¹⁰ projection fibres (e.g., the corona radiata), ¹¹ , ¹³ and association fibres (e.g., optic radiation). ¹⁰ Moreover, research indicates that WM pathology likely progresses over the lifespan of patients with PKU as blood Phe levels remain chronically elevated. ¹⁴ However, WM alterations have been shown to be reversible with the strict low‐protein diet, ¹⁵ but also with aforementioned treatments such as sapropterin ¹⁶ and pegvaliase. ¹⁷

Our recent investigation on the short‐term impact of high Phe levels on the grey matter (GM) of the brain revealed transient cortical thickness decreases after a 4‐week intake of Phe. ¹⁸ The extent to which this temporary elevation of Phe also affects WM microstructure has not been investigated in a randomised, placebo‐controlled, cross‐over trial. Therefore, longitudinal analyses are critically needed to assess the full extent of WM changes over time and their relationship with sustained high Phe exposure.

Diffusion tensor imaging (DTI) allows the detection of subtle alterations in the integrity of WM fibres, providing insights into the impact of metabolic alterations on WM microstructure and the cognitive consequences in adults with PKU. Additionally, understanding whether potential microstructural changes are reversible after returning to baseline Phe levels is vital for evaluating the brain's plasticity in response to metabolic fluctuations. Recognising that individuals with PKU may respond differently to high Phe levels highlights the importance of considering each patient's unique response in clinical practice.

This study aimed to investigate how elevated plasma and brain Phe levels over a 4‐week period affect the WM microstructure in adults with early‐treated classical PKU. A further goal of the study was to explore potential links between WM microstructure, cognitive function, and metabolic markers in PKU after the high Phe intervention compared to placebo.

2. METHODS

2.1. Participants

The PICO (Phenylalanine and Its Impact on Cognition) study investigated the impact of high Phe levels on cognitive performance, brain structure, and function in adults with PKU. ¹⁹ , ²⁰ Using a randomised, double‐blind, placebo‐controlled crossover design, the effects of experimentally increased Phe levels versus placebo were examined in adults with early‐treated PKU. The trial, conducted between July 2019 and June 2022 at the University Hospital Bern, Switzerland, received approval from the local Ethics Committee (2018‐01609), adhering to the ethical principles of the Declaration of Helsinki, and was registered on clinicaltrials.gov (NCT03788343). Patients were recruited from metabolic centres in Switzerland, Germany, and Austria, and written informed consent was obtained from all participants. Eligibility criteria were extensively described in a previous publication. ²⁰ In short, individuals needed to have received a PKU diagnosis after newborn screening, initiated early dietary treatment, be aged ≥18 years, and demonstrate the ability to comply with all study procedures. Exclusion criteria encompassed factors such as a lack of adherence to a Phe‐restricted diet in the past 6 months, Phe levels exceeding 1600 μmol/L in the previous 6 months, conditions influencing outcomes (e.g., untreated B12 deficiency, substance abuse), medications affecting cognition, pregnancy or lactation without adequate contraception, noncompliance to the study design, inability to follow procedures (e.g., language barriers, psychological disorders, dementia), concurrent participation in an investigational drug study, and MRI contraindications.

2.2. Study design and procedures

The intervention involved administering either Phe or placebo capsules for 4 weeks each, to simulate a temporary discontinuation of the dietary Phe restriction. To maintain a double‐blind setting without altering patients' customary low‐protein diets, capsules containing Phe or placebo were used. Treatment assignment remained blinded for participants, study investigators, and data analysts until after initial analysis. The investigational product consisted of capsules containing 250 mg of Phe and visually identical placebo capsules (pregelatinised starch), provided by Laboratorium Dr. G. Bichsel AG, Switzerland, in accordance with applicable regulations. Mirroring the approach of ten Hoedt et al. ²¹ the daily Phe and placebo capsule dosages were adjusted based on sex and body weight to model the typical dietary Phe intake for a healthy adult or unrestricted levels expected in an individual with PKU going completely ‘off‐diet’. Patients were randomly assigned to initially receive either Phe or placebo capsules for 4 weeks and then switched to the other capsules for the next 4 weeks. The two intervention periods were separated by a 4‐week washout period.

Participants completed four study visits at the University Hospital Bern, Switzerland, occurring before and after the interventions. Assessments included blood collection, neuropsychological examination, structural and functional MRI, and MR spectroscopy. For primary analysis results regarding cognitive outcomes, see Trepp et al. ²⁰ Grey matter findings are reported by Muri et al. ¹⁸ Task‐based fMRI and arterial spin labelling data are published in Maissen‐Abgottspon et al. ²²

3. CLINICAL OUTCOME MEASURES

3.1. Metabolic parameters

To evaluate metabolic control, venous blood samples were collected following an 8–12 hour overnight fast prior to MRI scans. Plasma Phe, tyrosine, and tryptophan concentrations were quantified through high‐performance ion‐exchange liquid chromatography paired with post‐column ninhydrin‐derivatisation and photometric detection.

3.2. Cognitive assessment

Cognitive assessments were performed four times before and after each intervention and as predefined in the PICO study protocol utilising validated instruments. Attention was examined via the Test of Attentional Performance (TAP), ²³ including the subtests of alertness (median reaction time in milliseconds), divided attention (total number of omissions), and sustained attention (standard deviation of reaction time in milliseconds). Executive functions, guided by the Miyake model, ²⁴ were evaluated using the TAP n‐back task (working memory accuracy in percent) and Delis‐Kaplan Executive Function System (D‐KEFS) ²⁵ conditions 3 (inhibition) and 4 (cognitive flexibility) of the colour‐word interference task (seconds to complete). Baseline general intelligence was evaluated using the Matrix Reasoning, Vocabulary, Arithmetic, and Symbol Search subtests from the Wechsler Adult Intelligence Scale‐Fourth Edition (WAIS‐IV). ²⁶ , ²⁷

3.3. Neuroimaging

Structural MRI and MR spectroscopy data were acquired following an 8–12 hour overnight fast on a 3T Siemens Prisma MRI with a 64‐channel head coil featuring a mirror to display calming nature videos during scans to minimise motion. Imaging was conducted for all participants at the Translational Imaging Center of the Swiss Institute for Translational and Entrepreneurial Medicine in Bern, Switzerland.

3.4. Structural MRI acquisition

High‐resolution T1‐weighted images were collected using an MPRAGE sequence (repetition time TR = 1950 ms, echo time TE = 2.26 ms, inversion time TI = 900 ms, acquisition time TA = 4:34 min, flip angle = 9°, 1 × 1 mm² in‐plane resolution, 1 mm slice thickness, 176 slices, 256 × 256 mm² field of view).

Diffusion‐weighted images were acquired with a spin‐echo echo planar imaging sequence using 122 non‐collinear directions preceded by a b = 0 reference volume (TR = 3700 ms, TE = 87 ms, TA = 7:55 min, slice thickness = 2.2 mm (isotropic), number of slices = 56, phase encoding direction = anterior–posterior, acceleration factor = 2, q‐space weightings = 3, q‐space max. b‐value = 3000 s/mm², full q‐space coverage).

3.5. ¹H spectroscopy acquisition

Proton magnetic resonance spectroscopy (¹H‐MRS) was performed using a semi‐LASER sequence with VAPOR water suppression localising a large supraventricular volume. The acquisition details were described in previous publications. ¹⁸ , ²⁸

3.6. DTI analysis

The FSL toolbox (Oxford Centre for Functional MRI of the Brain, version 6.0.5) was utilised for preprocessing of the DTI data, following the pipeline proposed by Maximov et al. ²⁹ Briefly, this pipeline incorporated skull stripping with FSL‐BET, corrections for noise, ³⁰ Gibbs ringing artefacts, ³¹ eddy currents and motion (FSL eddy), bias field correction ³² incorporated in Advanced Normalisation Tools (ANTs), and 1 mm spatial smoothing (FSL fslmaths). Thereafter, diffusion tensor model at each voxel was fitted using DTIFIT on the preprocessed data.

Tract‐based spatial statistics (TBSS) facilitated an exploratory examination of interventional changes in DTI scalars within our patient cohort. This method involved generating a WM skeleton on fractional anisotropy (FA) maps for each participant's longitudinal DTI data. Subsequently, FA, mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) values were projected onto the individualised skeletons for longitudinal analysis. Longitudinal changes per DTI scalar were assessed using paired statistical tests within FSL randomise, incorporating 5000 permutations and threshold‐free cluster enhancement with family‐wise error correction. Resulting p‐values were additionally Bonferroni‐adjusted for the four DTI scalars.

WM tracts were extracted using the Johns Hopkins University (JHU) ICBM‐DTI‐81 atlas ³³ , ³⁴ and the XTRACT HCP Probabilistic Tract atlas. ³⁵ We focused on regions of interest (ROIs) previously identified as differing between patients with PKU and healthy controls. ¹⁰ , ¹¹ , ¹³ , ³⁶ , ³⁷ These ROIs included various segments of the corpus callosum (genu, body, and splenium), as well as specific tracts such as the optic radiation, corona radiata (anterior, superior, and posterior), inferior and superior longitudinal fasciculus, internal capsule (anterior and posterior limbs), and the external capsule. For bilateral structures, tracts were extracted from the left and right hemispheres separately. In total, 21 WM tracts were selected for ROI analysis.

3.7. ¹H spectroscopy analysis

¹H spectra were processed in jMRUI (including frequency alignment, eddy correction, averaging, and water removal) and Phe was quantified in analogy to Kreis et al. ³⁸ Downfield fitting in FiTAID included a fixed background model verified by difference spectra. N‐acetylaspartate, homocarnosine, Phe, and background amplitudes were adjusted per subject. Quantification assumed literature values for WM and GM water content, and unaltered relaxation times. For more details, see Muri et al. ²⁸ and Muri et al. ¹⁸

3.8. Statistical analysis

Analyses were conducted in R version 4.1.2. ³⁹ DTI parameters, cognitive, and metabolic outcome measures were analysed following the statistical protocol used in the previous cognitive data analysis. ²⁰ A linear mixed model (LMM) with restricted maximum likelihood estimation was applied, using post‐intervention values (timepoints 2 and 4) as dependent variables. Fixed effects in the model included the intervention (Phe or placebo), period (first or second), sex, and randomisation centre. Baseline values (timepoints 1 and 3) and patients' age were included as covariates. Participant ID was treated as a random intercept to account for within‐subject correlations. LMM results are presented as point estimates with two‐sided 95% Wald confidence intervals (CIs). Additionally, estimated marginal means with standard errors (SE) were calculated from the LMM for each intervention (Phe and placebo). These marginal means estimate the average response for each intervention level while adjusting for other covariates in the model.

To assess the reversibility of the DTI alterations, a similar LMM was applied with the DTI scalars as dependent variable, the timepoint (T1, T2, T3, and T4), sex and centre of randomisation as fixed effects, age as covariate and participant ID as random intercept.

For each DTI scalar, cognitive, and metabolic variable, the fitted values were extracted from the model for the Phe and placebo period separately. These fitted values represent the model's estimates of the dependent variable based on the independent variables and their estimated effects. In LMMs, fitted values are conditional predictions that account for both fixed effects (group means) and random effects (individual variability). This allows analysis of data with hierarchical structures, such as repeated measures within subjects. Spearman correlations (r _s ) were used to assess relationships between DTI scalars, cognitive, and metabolic fitted values.

We report uncorrected, two‐sided p‐values with a significance threshold of p < 0.05. We additionally indicate whether these p‐values survive correction for multiple comparisons using the false discovery rate (FDR) by Benjamini and Hochberg. ⁴⁰ The p‐values for each hypothesis were FDR corrected. Effect sizes were interpreted as suggested by Funder and Ozer ⁴¹ : very small r _s  ≥0.05, small r _s  ≥0.10, medium r _s  ≥0.20, large r _s  ≥0.30 and very large r _s  ≥0.40.

4. RESULTS

4.1. Characteristics of participants

From 71 patients screened, 30 were enrolled in the randomised trial (116/120 planned visits were completed). One patient withdrew after the first visit, and another was excluded after the third visit for not following safe contraception protocols. Out of the 116 measurements, seven were done without MRI because of either technical issues (n = 1), claustrophobia (n = 3), or because they were performed at home visits (n = 3). The characteristics of randomised patients with at least three MRI assessments (n = 27) are displayed in Table 1. None of these characteristics were different between the Phe‐placebo and the placebo‐Phe subgroups, except for baseline blood tryptophan levels.

TABLE 1.

Baseline characteristics of participants.

	Overall (n = 27)	Phe‐placebo (n = 13)	Placebo‐Phe (n = 14)	p‐value
Age a
Median	35.66	35.27	36.14	>0.05
Interquartile range	10.26	8.96	19.57
Sex b
Female	13 (48%)	7 (54%)	6 (43%)	>0.05
Male	14 (52%)	6 (46%)	8 (57%)
Education b
High school	1 (4%)	0 (0%)	1 (7%)	>0.05
College/job training	22 (81%)	11 (85%)	11 (79%)
Graduate school	4 (15%)	2 (15%)	2 (14%)
Intelligence quotient (IQ)
Median	97	100	96	>0.05
Interquartile range	17	14	18
Blood Phe levels c
Median	733	726	751	>0.05
Interquartile range	373.5	333	381
Brain Phe levels d
Median	0.143	0.140	0.150	>0.05
Interquartile range	0.030	0.031	0.032
Blood tyrosine levels c
Median	38	44	37	>0.05
Interquartile range	12	9	9
Blood tryptophan levels c
Median	37	38	32	0.046
Interquartile range	9	3	8

^a Data are shown in years.

^b Data are numbers of participants, with percentages in parentheses.

^c Data are displayed as μmol/L.

^d Data are displayed as mmol/L.

4.2. Changes in WM microstructure

Whole brain TBSS analysis showed significantly altered WM structure across the brain after the Phe period compared to placebo (Figure 1). MD was most altered, followed by RD and AD, whereas FA showed no significant differences between the Phe and placebo periods. ROI analysis revealed that the posterior corona radiata and optic radiation were particularly affected by the high Phe intervention (see Table 2). The non‐significant results can be found in Supplementary Material S1.

FIGURE 1.

Voxel‐wise difference in WM microstructure between Phe and placebo periods indicated by t‐values, with larger values reflecting stronger changes in Diffusion Tensor Imaging (DTI) scalars. Regions with significantly decreased DTI scalars after the Phe period are marked by the colour continuum (the lighter (yellower), the stronger the decrease). Blue regions represent the mean fractional anisotropy (FA) skeleton. Voxel‐wise differences are displayed for mean diffusivity (MD, upper row), axial diffusivity (AD, middle row), and radial diffusivity (RD, bottom row). Fractional anisotropy did not show any significant differences between Phe and placebo periods after correction for multiple comparisons and is therefore not shown here.

TABLE 2.

Differences in DTI scalars across selected WM tracts between the Phe and placebo periods.

DTI scalar	ROI	Phe mean (SE)	Placebo mean (SE)	Estimate (SE)	95% CI	p‐value
AD	Anterior limb internal capsule right	0.760 (0.00232)	0.767 (0.00243)	0.00634 (0.00260)	0.00127, 0.0116	0.020
	Anterior corona radiata right	0.682 (0.00220)	0.690 (0.00233)	0.00751 (0.00302)	0.00190, 0.0131	0.014
	Body corpus callosum	0.891 (0.00390)	0.906 (0.00413)	0.0152 (0.00528)	0.00513, 0.0252	0.009
	Genu corpus callosum	0.897 (0.00368)	0.911 (0.00389)	0.0139 (0.00510)	0.00445, 0.0234	0.009
	Optic radiation left*	0.746 (0.00264)	0.758 (0.00281)	0.0124 (0.00368)	0.00556, 0.0192	<0.001
	Optic radiation right*	0.738 (0.00303)	0.751 (0.00322)	0.0128 (0.00422)	0.00502, 0.0207	0.002
	Posterior corona radiata left*	0.646 (0.00327)	0.663 (0.00348)	0.0166 (0.00456)	0.00819, 0.0251	<0.001
	Posterior corona radiata right*	0.641 (0.00398)	0.661 (0.00423)	0.0200 (0.00554)	0.00973, 0.0303	<0.001
	Splenium corpus callosum	0.909 (0.00382)	0.919 (0.00404)	0.0101 (0.00501)	0.000288, 0.0197	0.050
MD	Body corpus callosum	0.420 (0.00202)	0.427 (0.00214)	0.00736 (0.00269)	0.00218, 0.0125	0.008
	Genu corpus callosum	0.422 (0.00183)	0.428 (0.00194)	0.00580 (0.00254)	0.00109, 0.0105	0.029
	Optic radiation left	0.437 (0.00192)	0.443 (0.00204)	0.00590 (0.00252)	0.000984, 0.0108	0.026
	Optic radiation right	0.429 (0.00217)	0.437 (0.00230)	0.00745 (0.00279)	0.00200, 0.0129	0.009
	Posterior corona radiata left*	0.391 (0.00205)	0.401 (0.00218)	0.0103 (0.00286)	0.00504, 0.0156	<0.001
	Posterior corona radiata right*	0.387 (0.00244)	0.399 (0.00259)	0.0125 (0.00339)	0.00620, 0.0188	<0.001
	Splenium corpus callosum	0.392 (0.00202)	0.397 (0.00213)	0.00528 (0.00250)	0.000411, 0.0101	0.044
RD	Posterior corona radiata left	0.263 (0.00161)	0.270 (0.00171)	0.00708 (0.00223)	0.00294, 0.0112	0.010
	Posterior corona radiata right*	0.260 (0.00190)	0.269 (0.00202)	0.00881 (0.00264)	0.00392, 0.0137	0.002

Note: Estimated marginal means of DTI scalars with standard errors (SE) in brackets. AD, MD, and RD were scaled by 10⁻³ and are displayed as mm²/s ×10⁻³ to improve the readability of the table. Results that survived FDR correction are marked in bold and with an asterisk (*).

Abbreviations: AD, axial diffusivity; CI, confidence interval; MD, mean diffusivity; p, p‐value; Phe, phenylalanine; RD, radial diffusivity; ROI, region of interest; SE, standard error.

However, all WM changes were reversible within 4 weeks after cessation of the high Phe intervention (see Supplementary Material S2).

4.3. Changes in metabolic parameters

Blood Phe levels significantly increased after the Phe period (Phe period = 1433.0 μmol/L (46.0), placebo period = 876.0 μmol/L (48.8); point estimate = 557.45, 95% CI [438.71, 676.19], p < 0.001). Similarly, brain Phe levels significantly increased after the Phe period compared to placebo (Phe period = 0.276 mmol/L (0.008), placebo period = 0.172 mmol/L (0.009); point estimate = 0.104, 95% CI [0.084, 0.124], p < 0.001).

Blood tyrosine levels were significantly higher after the Phe period (Phe period = 52.6 μmol/L (2.33), placebo period = 43.6 μmol/L (2.50); point estimate = 9.08, 95% CI [2.92, 15.25], p = 0.009), although still within the reference range. Blood tryptophan levels did not show a significant change after the Phe period (point estimate = −1.52, 95% CI [−4.10, 1.16], p > 0.05).

4.4. Changes in cognitive performance

The cognitive performance results for the current sample (n = 27) are consistent with those previously reported (n = 28, Muri et al. ¹⁸ ). Specifically, the differences in cognitive performance between the Phe and placebo periods remained significant for inhibition and sustained attention, as observed in the prior study.

4.5. Relationship between WM microstructure, cognition, and metabolic control

4.5.1. WM microstructure and cognitive performance

At the ROI level, correlation analyses initially revealed significant associations between DTI scalars and cognitive performance after both the Phe and placebo periods (Supplementary Material S3a and S3b, respectively). However, following FDR correction, only the correlations after the placebo period between RD in the right superior corona radiata and performance in working memory and sustained attention remained statistically significant.

4.5.2. WM microstructure and metabolic parameters

After both the Phe and placebo periods, blood and brain Phe levels showed significant correlations with DTI scalars (Supplementary Material S4a and S4b, respectively), but only the correlation after placebo between MD in the left posterior limb of the internal capsule and blood Phe levels survived FDR correction.

4.5.3. Cognitive performance and metabolic parameters

Although some correlations were initially observed, they did not survive FDR correction. For instance, blood tyrosine levels were correlated with performance in working memory after the Phe period (r _s  = −0.48, p = 0.015, 95% CI [−0.73, −0.14]), and cognitive flexibility after the placebo period (r _s  = 0.42, p = 0.038, 95% CI [0.05, 0.69]); however, neither of these correlations remained significant after applying FDR correction.

5. DISCUSSION

5.1. Summary

This randomised, placebo‐controlled, cross‐over trial involving 27 adults with PKU investigated the impact of a 4‐week high Phe intervention on the WM microstructure and its relationship to cognitive performance and metabolic parameters. We observed a reduction in DTI scalars including AD, MD, and RD, following the Phe period compared to placebo. Notably, these alterations were particularly evident in the posterior corona radiata and the optic radiation. However, these changes in the WM appear to be reversible, as evidenced in those patients who first received the Phe intervention. Neither cognitive performance nor metabolic parameters was significantly associated with DTI scalars following the Phe period.

5.2. Decreases in DTI scalars

The reductions in AD, MD, and RD after the Phe period compared to the placebo period suggest that elevated Phe levels over 4 weeks induces microstructural alterations in the WM of adults with PKU. Among these, MD appears to be the most affected DTI scalar according to the whole‐brain TBSS maps. This finding aligns with Vermathen et al. ³⁶ who also reported decreased apparent diffusion coefficient (ADC) values and a reduced myelin water fraction in WM lesions. Both MD and ADC quantify the overall rate of water diffusion within tissues and have been found to decrease in the presence of cytotoxic oedema. ⁴² Intramyelinic oedema, a specific form of cytotoxic oedema, is hypothesised to be the underlying cause of the hyperintense WM lesions observed on T2‐weighted and FLAIR images in early‐treated patients with PKU. ⁷ , ³⁶ Vacuole formation within the myelin sheaths could explain the observed changes in diffusivity scalars. Small vacuoles of a few micrometres, with very small effective ADC due to water being restricted to within the vacuoles and longer T2 relaxation times, may contribute to the signal in a manner which results in apparent decreases in AD and RD without altering the actual tissue diffusivity properties. This signal contribution from vacuoles reinforces the hypothesis of intramyelinic oedema. FA values, which reflect the directionality of water diffusion, were not significantly different between the Phe and placebo periods, indicating overall preserved fibre organisation after 4 weeks of high Phe.

It is important to emphasise that changes in the WM microstructure were reversible within 4 weeks in patients who first received Phe. On a macrostructural level, Cleary et al. ¹⁵ previously demonstrated that WM abnormalities improved 3–12 months after a reintroduction of the diet in adolescents and adults with PKU, although these abnormalities did not completely resolve within that period. On a microstructural level, Clocksin et al. ¹⁶ have shown reversibility of WM abnormalities after 6 months of sapropterin treatment in individuals with PKU. Our findings contribute to this body of evidence by demonstrating reversibility of WM microstructure after a shorter, 4‐week high Phe exposure. In maple syrup urine disease, lower ADC values were found in the periventricular WM and cerebral cortex (among other regions) after metabolic decompensation, which were also reversible after treatment. As is the case for PKU, the hypothesis was that lower ADC values in the WM were an indication of cytotoxic or intramyelinic oedema. However, it remains unclear when these changes become permanent.

After the Phe period, changes in the WM were primarily observed in the posterior WM tracts, especially in the bilateral posterior corona radiata and optic radiation. These findings are consistent with our previous cross‐sectional study using the same patient sample, in which we compared white matter integrity to age‐, sex‐, and education‐comparable healthy controls. ²⁸ In that study, we observed significant alterations in these same posterior WM tracts, suggesting that individuals with PKU exhibit baseline WM alterations in comparison to healthy controls. Given that the optic radiation and corona radiata intersect with various other fibres, MD values are recommended for interpretation as they are more robust to the presence of multiple fibre populations. ⁴³ Since MD values in the posterior WM tracts showed significant differences between the two periods, it is reasonable to conclude that these tracts are particularly susceptible to short‐term high Phe levels in adults with PKU. Notably, the location of these alterations overlaps with areas that frequently exhibit white matter lesions on a macrostructural level.

5.3. Associations

The correlation pattern indicates that the relationship between DTI scalars and cognitive performance was more extensive and had higher effect sizes after the placebo period compared to the Phe period. The correlation coefficients ranged from |0.39| to |0.65| after the Phe period, and from |0.41| to |0.72| after the placebo period. Notably, two correlations after the placebo period survived FDR correction, while none did after the Phe period. This finding, combined with our previous baseline analysis showing FDR‐corrected correlations between DTI scalars and cognitive performance (Muri et al., 2023), suggests that the brain‐behaviour relationship in PKU may be more stable or better detectable at lower Phe levels. Interestingly, this observation aligns with our previous study using a similar patient sample, which demonstrated correlations between cognitive performance and neural activation only after placebo and not after Phe exposure. ²² This may be because lower Phe levels likely result in less metabolic interference. In contrast, higher Phe levels potentially disrupt the brain‐behaviour relationship through direct and indirect effects on WM integrity, which ultimately impacts cognitive functions. Thus, high Phe exposure may attenuate or mask the typical associations between brain structure and function and cognitive performance in adults with PKU.

After applying FDR correction, the only statistically significant correlation between Phe levels and DTI scalars was observed between MD in the left posterior limb of the internal capsule and blood Phe levels after the placebo period. Nevertheless, the pattern of the results suggests that, at a certain threshold, blood and brain Phe levels affect the WM integrity regardless of concentration. It also indicates that blood and brain Phe levels alone may not fully explain the effects of elevated Phe on the brain. Since high Phe levels can trigger a cascade of metabolic disruptions, additional markers of the altered metabolism in PKU may be necessary for a comprehensive evaluation on the impact of PKU on the brain. For instance, high Phe levels may disrupt glucose metabolism ⁴⁴ as well as neurotransmitter and protein synthesis, ⁴⁵ , ⁴⁶ which may all contribute to the WM pathology.

5.4. Integration of findings

The present findings, corroborated by our previous publication on GM alterations, ¹⁸ indicate that transient Phe elevations may precipitate intramyelinic oedema, detectable via structural MRI modalities such as T1‐weighted and diffusion‐weighted imaging. However, the observed reductions in cortical thickness and DTI scalars do not necessarily signify permanent structural brain changes. Instead, these alterations appear reversible as shown in the present and previous study. ¹⁸ Interestingly, our study on the effect of a 4‐week high Phe intervention on neural activation and cerebral blood flow revealed no significant difference between the Phe and placebo period. ²² Despite the non‐significant results, the general pattern of lower cerebral blood flow and neural activation following the Phe period could indicate that if Phe exposure is prolonged, the effects of elevated Phe levels may become more noticeable on functional MRI. Indeed, lower cerebral blood flow and slight alterations in neural activation in patients with PKU compared to controls were found previously. ⁴⁷ , ⁴⁸

It is important to note that the analysis of the primary outcome of the PICO study (accuracy in working memory performance) demonstrated that a 4‐week high Phe intervention was noninferior to placebo. ²⁰ Additionally, no significant differences emerged between the Phe and placebo periods for other cognitive and mood‐related outcomes, except for sustained attention. When asked if patients could guess the intervention allocation (first Phe or placebo), only 43% of patients guessed correctly. ⁴⁹ However, those who guessed correctly also had significantly higher Phe levels after the intervention and a greater Phe level increase compared to those who guessed wrong.

The findings of the PICO study collectively suggest that short‐term elevations in Phe levels in adults with PKU can induce detectable, yet reversible, changes in brain structure without significantly affecting most aspects of cognitive function, brain activation, or cerebral perfusion. This disconnection between structural changes and functional outcomes highlights the complex relationship between brain structure, function, and metabolism in PKU. The subtlety of these effects is further underscored by patients' inability to consistently identify periods of Phe elevation, although those more sensitive to Phe fluctuations may experience more noticeable effects. This variability in individual responses suggests a need for personalised approaches in PKU management. Furthermore, these results provide reassurance that short‐term lapses in dietary control may not have immediate, severe consequences on brain function in adults with PKU. However, the observed structural changes and trends in functional measures underscore the need for continued vigilance in PKU management. While the brain shows resilience to short‐term Phe elevations, the potential cumulative effects of repeated or prolonged exposures remain a concern and warrant further investigation.

It is also crucial to recognise the challenge of distinguishing between the effects of high Phe exposure in childhood compared to adulthood. The brain changes and functional outcomes observed in adult patients with PKU likely reflect a complex interplay of historical Phe exposure during critical developmental periods and current metabolic status. This complicates the interpretation of adult PKU studies and underscores the need for longitudinal research that can disentangle the long‐term effects of childhood Phe exposure from the more immediate impacts of adult metabolic control.

5.5. Limitations

The results of this study should be interpreted in light of certain limitations. DTI scalars alone, such as those derived from the tensor model, may not fully elucidate the nature of the underlying biological changes as they may not comprehensively reflect the intricate microstructural changes induced by elevated Phe levels. Furthermore, the TBSS analysis employed, although widely used, may be susceptible to partial volume effects and the inability to resolve crossing fibres within a voxel accurately. Future studies could benefit from complementary advanced diffusion imaging techniques, such as high‐angular resolution diffusion imaging or diffusion kurtosis imaging, which may provide more accurate representations of WM microstructure and potentially enhance the characterisation of the observed alterations.

5.6. Conclusion

We showed reversible decreases in DTI scalars after a 4‐week high Phe period compared to placebo in adults with PKU. Cognitive performance and metabolic parameters were only related to DTI scalars after the placebo period. Further investigation is necessary to determine the clinical implication of these changes.

AUTHOR CONTRIBUTIONS

Raphaela Muri: Conception and design, funding acquisition, data collection, MRI/MRS analysis and evaluation, statistical data analysis, interpretation of data, drafting of the manuscript, reviewing and editing the manuscript. Murray Bruce Reed: MRI/MRS analysis and evaluation, interpretation of data, reviewing and editing the manuscript. Stephanie Maissen‐Abgottspon: Funding acquisition, data collection, interpretation of data, reviewing and editing the manuscript. Roland Kreis: Data collection, MRI/MRS analysis and evaluation, interpretation of data, reviewing and editing the manuscript. Michel Hochuli: Interpretation of data, reviewing and editing the manuscript. Rupert Lanzenberger: Interpretation of data, reviewing and editing the manuscript. Roman Trepp: Conception and design, funding acquisition, interpretation of data, reviewing and editing the manuscript. Regula Everts: Conception and design, funding acquisition, interpretation of data, reviewing and editing the manuscript, supervision, Guarantor.

FUNDING INFORMATION

The study was funded by a project grant (192706) and a doc.CH grant awarded to RM (184453) of the Swiss National Science foundation (SNSF), the Vontobel Foundation (Switzerland), the Bangerter Rhyner Foundation (Switzerland), a young investigator grant from the Inselspital Bern (CTU grant, Switzerland), the Nutricia Metabolics Research Fund (Netherlands), the Fondation Rolf Gaillard pour la recherche en endocrinologie, diabétologie et métabolisme (Switzerland), and a grant from the Swiss Foundation for Nutrition Research awarded to SA. The funders had no involvement in the study design, collection, analysis, and interpretation of the data.

CONFLICT OF INTEREST STATEMENT

Raphaela Muri, Murray Bruce Reed, Stephanie Maissen‐Abgottspon, Roland Kreis, Michel Hochuli, Rupert Lanzenberger, Roman Trepp, and Regula Everts declare that they have no conflict of interest.

ETHICS STATEMENT

All authors were compliant and followed the ethical guidelines, according to the requirements of the JIMD.

PATIENT CONSENT STATEMENT

All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2013. Informed consent was obtained from all patients prior to study participation.

Supporting information

Data S1. Supporting Information.

Muri R, Reed MB, Maissen‐Abgottspon S, et al. Reversible white matter changes following a 4‐week high phenylalanine exposure in adults with phenylketonuria. J Inherit Metab Dis. 2025;48(1):e12823. doi: 10.1002/jimd.12823

DATA AVAILABILITY STATEMENT

Neuroimaging data, cognitive data, metabolic data, and some demographical variables (age, sex) are available upon reasonable request after signing a confidentiality statement and a data sharing agreement.