Cerebral microstructural changes in children suffering from hemolytic uremic syndrome

Typical hemolytic uremic syndrome (HUS) is characterized by the triad of acute renal failure, thrombocytopenia and hemolytic anemia and is mostly caused by Shiga toxin-producing Escherichia coli (STEC) in Europe. Neurological signs are reported in 17 – 52% of cases [1], with Central Nervous System (CNS) involvement contributing significantly to chronic morbidity [2], and potentially to fatal outcome [3]. Magnetic resonance imaging (MRI) may reveal abnormalities affecting the basal ganglia and / or thalami, as well as the deep white matter [4]. Clinical courses of HUS patients with neurological involvement are highly variable, and few prognostic factors have been described to date [3]. Early MRI may offer such options, as postulated by Donnerstag et al. [4].

While renal failure is caused by thrombotic microangiopathy leading to vessel occlusions by platelet–fibrin thrombi [5, 6], the cause of neurological complications is still poorly understood. Several pathophysiological mechanism have been discussed in neurological involvement, such as endothelial injury, thrombus formation, hemorrhage, and posterior reversible encephalopathy syndrome [7, 8]. As Weissenborn et al. [9] and Donnerstag et al. [4] hypothesized, both metabolic and toxic effects may be responsible for CNS involvement. Additionally, Weissenborn et al. [9] indicated microstructural rather than macrostructural cerebral changes, reporting increased T2 relaxation times in basal ganglia in all adult STEC-positive HUS (STEC-HUS) patients showing inconspicuous conventional magnetic resonance imaging (cMRI) on qualitative analysis.

We aimed to define prognostic factors at early stages of the disease, further investigating possible microstructural cerebral alterations in pathological as well as normal appearing brain regions. We retrospectively determined apparent diffusion coefficient (ADC) values in different supra- and infratentorial locations early in the acute phase of neurological involvement in children with STEC-HUS and compared them with age-related ADC reference values from published pediatric controls [10].

This retrospective study was approved by the local institutional review board. Data and neurological symptoms from 12 children suffering from STEC-HUS between 2009 and 2021 and had an MRI at 1.5 T were retrospectively analyzed (5 females, 7 males, median age 2.7 years [range: 0.8–14.6 years]). No patient was excluded. Some patients were previously included in the cohort described by Donnerstag et al. [4] and Pape et al. [11]. Cerebral MRI was performed within 24 h of presentation of neurological symptoms in 6/12 STEC-HUS patients, and between 24 and 48 h in 5/12 STEC-HUS patients to assess severity of structural cerebral alterations and rule out infarction during acute illness. For one STEC-HUS patient (ID 9), MRI was measured 4 days after onset of neurological symptoms. Additionally, four children had an early follow-up MRI within one week and three children received long-term follow-ups between 6 and 18 months. Early follow-up MRI were included only in qualitative analysis. Patients’ clinical courses were variable and we therefore divided our cohort into 2 subgroups depending on outcome. Subgroup A (Patient ID 1, 3, 5, 7 and 8) showed poor neurological outcome with severe neurological deficits including two fatal outcomes, whereas children in subgroup B (Patient ID 2, 4, 6, 9, 10, 11 and 12) clearly improved and were without neurological deficit at discharge. For age-specific comparison with our control subjects [10], children were divided into age-dependent subgroups (0–0.25, 0.25–0.5, 0.5–1, 1–2, 2–3, 3–5, 5–7.5, 7.5–10, 10–12.5 and 12.5–17.2 years); there were no STEC-HUS children in the two youngest age groups and the 7.5–10 years subgroup. The control group comprised 112 children (0–17.2 years) from our pool of pediatric MRI examinations at 1.5 T which were retrospectively selected showing no signal abnormalities [10] Brain MRI was performed for various clinical indications including headache, dizziness and vomiting, sensorineural hearing loss, first epileptic seizures, and newly diagnosed leukemia before therapy. Between 0 and 2 years of age, the age structure of girls and boys was similar without gender-dependent differences in ADC values in any region. In the older age group the age distribution was skewed, so that a gender comparison could not be performed.

Routine MRI was performed on 1.5 T MR systems (Avanto and Aera MR systems, Siemens AG or Genesis Signa, GE Medical Systems) and reviewed qualitatively by one experienced neuroradiologist (E.B., more than 15 years of experience). All examinations included axial diffusion-weighted, single-shot, spin-echo echo-planar sequences with 4 mm or 5 mm slices and an imaging matrix = 192 × 192 or 128 × 128) and all but one patient received a SWI or T2* sequence. In three (x, y, z) orthogonal axes an effective b value of 1000 s/mm² was used and additional measurement without diffusion weighting (b = 0 s/mm²) was acquired. ADC maps were generated from the scanner. Diffusion-weighted imaging (DWI) and ADC data were transferred to an external workstation for further analysis.

Using manually delineated circular or rectangular regions of interest (ROI) with minimum area 20 mm² (20–283 mm²) depending on region and patient anatomy, ADC values were measured in 35 infra- and supratentorial locations as published by Bültmann et al. [10].

ROI position was carefully chosen for anatomically comparable positions on B0 images by the experienced neuroradiologist (E.B.) and transferred to ADC images using the program imageJ (based on Java, public domain https://​imagej.​nih.​gov/​ij/​). For each location, mean ADC value, standard deviation, as well as minimal and maximal ADC values, were calculated (×10–3 mm²/ sec). Two measurements were excluded due to partial volume effects. One patient in subgroup A showed a few small punctate lesion on T2* and one patient developed multiple punctate SWI lesions in the putamen and thalamus on follow-up. These changes were not relevant to the ADC measurements. Control ADC data was also measured at 1.5 T and previously published [10].

Statistical analyses were performed with SAS 9.4 (SAS Institute Inc., Cary, NC, USA). Initially, the control and STEC-HUS groups were compared descriptively relative to age (median and quartiles) and gender (absolute and relative frequencies). The STEC-HUS cohort was then described individually and in subgroups depending on clinical outcome (median age and number of involved regions with comparison by exact Wilcoxon two-sample test). To compare age-dependent ADC values for STEC-HUS and control children, values for the former were evaluated individually as below, within or above the reference range (taken from [10]), and corresponding absolute and relative frequencies were calculated.

Although CNS involvement frequently complicates typical STEC-HUS, microangiopathy and small vessel involvement have rarely been clearly demonstrated in patients [12‐14]. Our group [4, 9] has previously discussed possible metabolic and / or toxic contribution and imaging abnormalities on cMRI scans in STEC-HUS patients. In addition to macrostructural changes, Weissenborn et al. [9] described microstructural alterations. For all adult STEC-HUS patients with normal cMRI scans using quantitative MRI parameters, significantly elongated T2 relaxation times in basal ganglia were reported compared to healthy controls but ADC values were unchanged. In contrast, Donnerstag et al. [4] measured ADC values in infra- and supratentorial lesions on diffusion-weighted images in STEC-HUS children. They found increased as well as decreased intralesional ADC values compared to age-related values obtained from Morriss et al. [15]. Thus, the question arose as to whether we could detect ADC value alterations in STEC-HUS children as an indicator of microstructural changes independent of findings on cMRI scans.

In the present study, we present data for a cohort of 12 children with typical STEC-HUS and clinical CNS involvement divided into two subgroups depending on neurological outcome. Qualitatively, the majority of our STEC-HUS patients (67%) showed T2 hyperintense lesions in basal ganglia and/ or thalami. The presence of T2 hyperintense lesions together with corresponding reduced diffusivity in these brain regions was only observed in patients with poor outcome. Quantitative assessment of ADC values in multiple standardized brain regions showed ADC abnormalities involving brain structures that appeared normal on qualitative evaluation. These findings were most pronounced in patients with devastating neurological outcome.

In accordance with Donnerstag et al. [4], but in contrast to Gitiaux et al. [16], we also measured decreased ADC values, as a sign of limited diffusion in cytotoxic edema, as well as increased ADC values, as part of a vasogenic edema. However, Gitiaux et al. [16] performed intra-individual comparisons of ADC values between involved and normal appearing areas, while we compared values for STEC-HUS children with those from a pediatric control group of children with age-appropriate development.

Using ADC values, we could identify more extensive changes in comparison to qualitative lesions detected by cMRI. ADC alterations were measured in normal appearing brain regions on T2w-, T1w- and FLAIR images indicating a more widespread involvement of the brain. Although Weissenborn et al. [9] did not report such changes, it should be noted that their study involved adults, while we investigated children. As described in several publications [17‐19], development of the human brain extends through to adulthood, with myelination an essential phase of postnatal brain maturation [20]. In accordance, ADC and T2 values decrease until late childhood [9, 21, 22]. Some publications have already considered whether increasing T2 hypointensity during brain maturation may reflect chemical maturation of the myelin sheath, involving decreasing water content, increasing hydrophobicity [23] and tightening of the myelin spiral around the axon [24]. These ongoing maturational processes could underly a higher vulnerability in the pediatric brain and contribute to differences between child and adult STEC-HUS patients.

When considering neurological outcome, basal ganglia and thalamic lesions on diffusion-weighted images appeared to be a prognostic factor for poor outcome in our children, in particular in association with corresponding T2 hyperintensities in the same location. Gitiaux et al. [16] recorded the most severe T2 hyperintensities and the highest ADC alterations in basal ganglia of two patients with a fatal outcome. Weissenborn et al. [9] also measured quantitative basal ganglia involvement in terms of T2 relaxation times in patients suffering from STEC-HUS neurological complications. Taken together, these studies along with our present data indicate a significant role for basal ganglia involvement in STEC-HUS. In a rat model, Goldstein et al. [25] demonstrated direct effects of Shiga toxin 2 in neurons and glial cells after intracerebroventricular microinfusion and confirmed Shiga toxin 2 presence in different striatum cell compartments, supporting our hypothesis. On the contrary, Nathanson et al. [26] reported that neurological outcome did not depend on localized cerebral involvement. However, the latter neither described how CNS damage on MRI was defined or included quantitative imaging. MRI was mostly obtained at a later timepoint after disease onset, which may influence signal changes. Early MRI was performed in 22/29 patients later than 2 days after onset of neurological symptoms, in contrast to 1/12 in our cohort. A recent retrospective study of pediatric patients with STEC-HUS including 22 with neurological involvement showed good neurological outcome in spite of DWI abnormalities involving the deep white matter and grey matter. Similarly to our observation, the authors also described DWI abnormalities with mixed increased and decreased ADC values [27], and good outcome in most patients in spite of DWI abnormalities, including 2/3 with involvement of basal ganglia. However, the study does not include information on timing of investigation and most importantly quantitative ADC values. Therefore, the results cannot be directly compared to our findings and do not contradict our observation that basal ganglia and thalamic lesions on diffusion-weighted images appeared to be a prognostic factor for poor outcome.

Our present study further indicates that poorer prognosis is associated with involvement of a greater number of regions showing quantitative ADC value alterations reflecting a more widespread, extensive brain involvement. No relevant differences were seen in the ratio of ADC values below versus above the normal range (Subgroup A 2.4: 1; Subgroup B 2.9:1). Gitiaux et al. [16] described milder and less widespread hyperintensities on T2/FLAIR images compared to DWI abnormalities. We suggest quantitative assessment of diffusion-weighted images may predict neurological outcome in acute disease and would therefore be required as soon as practicable after presentation of neurological symptoms.

While age did not significantly differ between STEC-HUS subgroups, children with a poor prognosis tended to be younger, in accordance with previous literature [28]. Age at STEC-HUS onset may be relevant, possibly linked to a greater sensitivity of the more immature brain.

Our study has practical limitations. While the number of STEC-HUS children is low, as typical HUS is rare and neurological involvement is only observed in a third of cases [3], achieving a large series would be highly challenging. Consequently, with a small cohort and age dependency of ADC values, we were confined to descriptive analysis. We also compared patient ADC values with published control reference values [10], which, though not verified from additional studies, are comparable with data from the literature [29‐31].

For the first time using early MRI and quantitative ADC analysis, we were able to demonstrate changes in otherwise normal appearing brain regions in STEC-HUS children with neurological involvement. We were also able to define potential predictive factors that require ratification in further imaging studies including different quantitative parameters. Predictors of poor prognosis are necessary to guide treatments such as the C5-complement inhibitor Eculizumab [11].

In conclusion, our results suggest that DWI and quantitative ADC value measurements are able to detect subtle microstructural changes in STEC-HUS children, which are in part not visible on cMRI. This may add to the understanding of pathophysiological mechanisms involved in neurological involvement. Involvement of deep gray matter structures on DWI and T2w images may provide key prognostic factors, though this needs to be confirmed in larger case series.