Multimodal layer modelling reveals in vivo pathology in amyotrophic lateral sclerosis

Abstract Amyotrophic lateral sclerosis (ALS) is a rapidly progressing neurodegenerative disease characterized by the loss of motor control. Current understanding of ALS pathology is largely based on post-mortem investigations at advanced disease stages. A systematic in vivo description of the microstructural changes that characterize early stage ALS, and their subsequent development, is so far lacking. Recent advances in ultra-high field (7 T) MRI data modelling allow us to investigate cortical layers in vivo. Given the layer-specific and topographic signature of ALS pathology, we combined submillimetre structural 7 T MRI data (qT1, QSM), functional localizers of body parts (upper limb, lower limb, face) and layer modelling to systematically describe pathology in the primary motor cortex (M1), in 12 living ALS patients with reference to 12 matched controls. Longitudinal sampling was performed for a subset of patients. We calculated multimodal pathology maps for each layer (superficial layer, layer 5a, layer 5b, layer 6) of M1 to identify hot spots of demyelination, iron and calcium accumulation in different cortical fields. We show preserved mean cortical thickness and layer architecture of M1, despite significantly increased iron in layer 6 and significantly increased calcium in layer 5a and superficial layer, in patients compared to controls. The behaviourally first-affected cortical field shows significantly increased iron in L6 compared to other fields, while calcium accumulation is atopographic and significantly increased in the low myelin borders between cortical fields compared to the fields themselves. A subset of patients with longitudinal data shows that the low myelin borders are particularly disrupted and that calcium hot spots, but to a lesser extent iron hot spots, precede demyelination. Finally, we highlight that a very slow progressing patient (Patient P4) shows a distinct pathology profile compared to the other patients. Our data show that layer-specific markers of in vivo pathology can be identified in ALS patients with a single 7 T MRI measurement after first diagnosis, and that such data provide critical insights into the individual disease state. Our data highlight the non-topographic architecture of ALS disease spread and the role of calcium, rather than iron accumulation, in predicting future demyelination. We also highlight a potentially important role of low myelin borders, that are known to connect to multiple areas within the M1 architecture, in disease spread. Finally, the distinct pathology profile of a very-slow progressing patient (Patient P4) highlights a distinction between disease duration and progression. Our findings demonstrate the importance of in vivo histology imaging for the diagnosis and prognosis of neurodegenerative diseases such as ALS.


Introduction
Amyotrophic lateral sclerosis (ALS) is a rapidly progressing neurodegenerative disease resulting in the loss of motor control, with a median survival time of 3 years. 1,2It affects both upper (UMN) and lower (LMN) motor neurons, 1 and symptoms initially present focally in one body part.The disease spreads topographically, often first to the contralateral limb, 3 eventually resulting in bulbar symptoms, which are associated with poorer outcomes. 4Current knowledge of ALS pathology is largely based on post-mortem evidence, [5][6][7] therefore reflecting advanced disease stages.A detailed understanding of early stage ALS pathology is needed to understand disease mechanisms and to facilitate earlier diagnosis and prognosis.Such advances may identify novel therapeutic targets to slow disease progression.
Post-mortem studies have revealed abnormalities in the primary motor cortex (M1) in ALS, such as the depopulation of Betz cells in cortical layer 5(b), 5 iron accumulation in deep cortex 6 and increased intracellular calcium. 7As standard 1.5 T or 3 T MRI cannot detect these layer-specific features, our understanding of in vivo ALS pathology is limited and disease mechanisms are still debated, such as the focal hit versus multifocal pathology hypotheses. 8Recent advances in ultra-high field MRI at 7 T and above enable the automated assessment of anatomically-relevant cortical layers in vivo, [9][10][11] allowing us to achieve in vivo histology.The present study aims to apply this approach to provide a systematic description of in vivo M1 pathology in early stage ALS.
Previous MRI studies have shown iron accumulation in M1 of ALS patients, [12][13][14] which has been approximately localized to deep M1, where it reflects neuroinflammation and microglia activation, 6,15 in addition to UMN impairment. 16Moreover, iron accumulation strongly affects the topographic area corresponding to the symptom onset site (i.e.8][19] We previously showed that the topographic areas of M1 are microstructurally distinct in healthy adults and should therefore be considered cortical fields, 10 which emphasizes the need for precise localization.Moreover, we showed that the lower-limb (LL), upper-limb (UL) and face (F) fields show distinct levels of age-related iron accumulation, 10 highlighting the importance of using age-matched controls.In addition to iron, calcium is also dysregulated in ALS, where increased calcium is associated with glutamate excitotoxicity and motor neuron death in animal models. 20,21This excitotoxicity may reflect the increased excitability of human M1 in ALS. 22It is currently unclear how calcium dysregulation relates to iron accumulation and demyelination in ALS.
The low myelin borders that separate cortical fields in M1 23,24 show interruptions in myelination at the depths where the Betz cells are located, 10,23 which are known to be affected in ALS 5 and are associated with different functional networks compared to cortical fields. 25As maladaptive proteins often propagate along highly myelinated neurons in neurodegeneration, 26 these low myelin borders may act as natural boundaries to limit disease spread.Alternatively, the degeneration of these borders may contribute to topographic disease spread between cortical fields. 24An investigation into the stability of these low myelin borders in ALS is therefore of interest.
The present study aimed to provide a systematic description of in vivo pathology in ALS patients.We used topographic layer imaging 24 to characterize pathology with respect to both cortical layers and fields. 10Submillimetre quantitative T1 (qT1) and quantitative susceptibility mapping (QSM) data, along with functional localizers, were collected from 12 ALS patients and 12 age-, gender-, education-and handedness-matched controls.We used increased positive QSM (pQSM) values as a validated marker of iron, 6,15 increased qT1 values as a validated marker of (de)myelination 27 and decreased (more negative) negative QSM (nQSM) as a marker of calcium. 28We extracted microstructural profiles from each cortical layer (superficial, layer 5a, layer 5b, layer 6) according to a previously published approach 9,10 and calculated multimodal in vivo pathology maps.We then tested whether (i) the mean cortical thickness is decreased and/or the layer architecture of M1 is degenerated in patients compared to controls; (ii) pathological iron accumulation is layer-specific; (iii) pathological iron and calcium accumulation are topographic (i.e.largely restricted to the first-affected cortical field) or atopographic (i.e.not restricted to the first-affected cortical field); (iv) pathological iron and/or calcium accumulation hot spots precede demyelination; (v) low myelin borders between cortical fields are disrupted with disease progression, and whether they show high or low substance accumulation; and (vi) a very slow progressing patient shows a distinct multimodal pathology profile compared to other patients.).Note that these group averages exclude one patient (Patient P4) with a very long disease duration (166 months between diagnosis and MRI).Participants underwent 7 T MRI and behavioural assessments.Three patients (Patients P1, P2 and P4) were measured again within 1 year (T2), while one patient (Patient P4) was also measured 8 months after T2 (T3).Patients were recruited from the University Clinic Magdeburg, Hannover Medical School and Rostock University Medical Center (Table 1).An experienced neurologist (S.V.) performed clinical assessments (Table 2).Of the 12 patients, seven had UL-onset (three right-lateralized, two left-lateralized, one bilateral), two had LL-onset (both left-lateralized) and three had bulbar-onset (B-onset; considered bilateral).Healthy controls were recruited from the DZNE database in Magdeburg and exclusion criteria included sensorimotor deficits, neurologic disorders and 7 T MRI contraindications.All participants gave written informed consent and were paid.The study was approved by the local Ethics Committee of the Medical Faculty of the University of Magdeburg.

T MRI data acquisition
Data were collected using a 7 T MRI scanner (MAGNETOM, Siemens Healthcare) equipped with a 32-Channel Nova Medical Head Coil, located in Magdeburg, Germany.We acquired whole-brain, 0.7 mm isotropic resolution MP2RAGE images 36 (sagittal slices, repetition time = 4800 ms, echo time = 2.01 ms, field of view read = 224 mm, GRAPPA 2, flip angle = 5°/3°, inversion time TI1/TI2 = 900/2750 ms, bandwidth = 250 Hz/Px).We also acquired wholebrain, 0.5 mm isotropic resolution susceptibility-weighted (SWI)  stage 32 indicates the stage of disease progression based on the ALSFRS-R score, 33 where stage 2A reflects the involvement of one body part and that a clinical diagnosis has taken place, while stage 2B and stage 3 reflect the subsequent involvement of second and third body parts, respectively.The CNS-LS 34 indicates the frequency of pseudobulbar episodes, with higher scores indicating greater impairment.The CNS-BFS 35

Structural preprocessing
CBS Tools (v3.0.8) 37 in MIPAV (v7.3.0) 38was used to process the structural data.Skull stripping and dura estimation were used to remove extra-cranial tissue, which was manually refined using ITK-SNAP (v3.8.0).The TOADS algorithm 39 was used to segment the brain into different tissue types, before the CRUISE module was used to estimate tissue boundaries, 40 resulting in level set images.The distance field module was used to calculate mean cortical thickness (i.e.across cortical depth).The volumetric layering module 41,42 was used to divide the cortex into 21 depths, according to the equivolume approach.M1 masks were manually delineated using ITK-SNAP, as described previously. 10

Quantitative susceptibility mapping
After data quality checks, the SWI data of four patients (and the four matched controls) were excluded due to severe motion and truncation artefacts, leaving 16 participants for the SWI analyses.QSMbox (v2.0) was used to reconstruct QSM from the SWI images. 43n line with previous studies, 12,44 QSM values were not normalized.We divided the QSM data into pQSM and nQSM values (as previously described 43 ).

Defining cortical layers
Using the simple curvature function in ParaView (v5.8.0), we calculated the mean curvature of the cortex.We then performed a vertex-wise linear regression analysis to predict qT1 values from mean curvature for each cortical layer in M1, on a subject-specific basis, according to an existing approach. 45The raw residuals of the regression models are referred to as 'decurved' qT1, following our previously published approach. 10We then applied a datadriven approach to identify cortical layers in M1 from groupaveraged decurved qT1 profiles separately for patients and controls 9,10 (see Fig. 1B for details and Supplementary Fig. 1 for right hemisphere).

Functional data processing
The functional data were motion-corrected at acquisition using the Siemens 'MoCo' correction.Preprocessing was performed in Statistical Parametric Mapping 12 (SPM12), including smoothing [2 mm full-width at half-maximum (FWHM)] and slice-timing correction.Co-registration was performed using the automated registration tool in ITK-SNAP, with further manual refinement based on anatomical landmarks where necessary.A first-level analysis created t-statistic maps (t-maps) for each body part (e.g.left hand) based on contrast estimates (e.g.[1 0 0 0 0]).To create the subjectspecific functional localizers of the cortical fields (e.g.left hand), we took the peak cluster of each t-map and removed overlapping voxels between localizers, as previously described. 10The functional localizers were then mapped onto the same subject-specific cortical surfaces as used for the structural data.We extracted layer-wise qT1 and signed QSM values from the cortical fields, resulting in microstructural profiles for each cortical layer (n = 4), each cortical field (n = 3) and each hemisphere (n = 2).

Myelin border analysis
We identified myelin borders between the UL and F representations based on the highest qT1 value (i.e.lowest myelin) located between the peak t-values of the UL and F localizers, using a previous approach. 23,48For each cortical layer, we extracted qT1 and signed QSM values from the myelin borders and calculated the average qT1 and signed QSM values in the UL and F body part representations (UL + F/2), in order to compare values in the myelin border to the cortical fields.

Layer-specific multimodal in vivo pathology mapping
Using topographic layer imaging, we created multimodal in vivo pathology maps on inflated cortical surfaces for each individual patient, in reference to the respective matched control (Fig. 2 and Supplementary Fig. 2).Pathology maps were calculated to identify disease hot spots indicated by demyelination (increased qT1, +1SD to +4 SDs), iron accumulation (increased pQSM, +1SD to +4 SDs) and calcium accumulation (more negative nQSM, −1 SD to −4 SDs) in each patient, relative to the mean value of the matched control.The reference values were layer-specific to control for microstructural differences between layers in M1. 10 In addition, pathology maps are shown with individually localized cortical fields in M1, to account for microstructural differences between cortical fields. 10

Behavioural tests of motor function
All participants underwent body part-specific behavioural tests of motor function (as previously described 10 ).To quantify LL function, the 6 Minute Walking Test (6MWT) was used to measure walking distance. 49To quantify UL function, a dynamometer 50 was used to measure hand strength.In addition, the Purdue 51 (Lafayette Instrument, model 32020A), Grooved 52 (Lafayette Instrument, model 32025) and O'Connor 53 (Lafayette Instrument, model 32021) pegboards were used to measure hand dexterity.To quantify bulbar (face, F) function, we used an automated tool that extracts features (e.g.errors) of lateral tongue movements from short video clips. 54

Statistical analyses
Statistical analyses were performed using IBM SPSS Statistics (v26, IBM, USA).Given the characteristic impaired motor function in ALS patients, we used one-tailed paired-samples t-tests to test for group differences in motor behaviour.Based on evidence of demyelination, iron and calcium accumulation in ALS, [5][6][7] we used uncorrected onetailed paired-samples t-tests to test for differences in cortical microstructure (i.e.qT1, QSM) between groups.Uncorrected P-values were used to avoid false-negatives due to the small sample size.We also report effect sizes with 95% confidence intervals.Note that pairedsamples t-tests were used to account for dependencies in the data (i.e.matched patient-control pairs).We used two-tailed one-sample t-tests to test for differences in cortical microstructure between a single, very slowly progressing patient (Patient P4) and all other patients.
We used linear regression models to test whether the layer-specific qT1, pQSM or nQSM predict whether a given cortical field is behaviourally affected.Individualized pathology maps were calculated by thresholding surfaces to show demyelination (increased qT1, +1 SD The first and second columns show data for all controls and all patients, with the group mean plotted in bold black and red, respectively.The third column shows the mean group data of the controls and patients, with red lines representing the patients.Note that qT1 and pQSM are validated in vivo markers of myelin 27 and iron 6,15 content, respectively, while nQSM is largely considered to reflect calcium 28 content.Also note that the profiles here represent the data of the entire M1, while in other figures and statistics the data used were averaged across cortical fields.(B) According to a previously published approach, 9 we identified four compartments ('layers') based on the 'decurved qT1' profile: Ls = superficial layer including layers 2-3; L5a = layer 5a; L5b = layer 5b; L6 = layer 6.Note that Ls does not include layer 1 as it is inaccessible with MRI 9 or layer 4, as it is absent in M1.We show the layer definitions for healthy controls (n = 12) and ALS patients (n = 12) in the present study (centre), as well as for older adults (n = 18) in our previous study 10 (left).L5a and L5b were distinguished based on the presence of two small qT1 dips at the plateau of 'decurved qT1' values (indicating L5), while L6 was identified based on a sharp decrease in values before a further plateau indicating the presence of white matter.We show our layer approximations over schematic depictions (right) of M1 myelin 46 and cell histological staining. 47(C) Despite differences in the shape of the profile in the affected fields compared to the average across fields in patients, layers can be similarly identified in the affected fields.

Impaired motor function in ALS patients
We quantified body part-specific motor function in 12 ALS patients and 12 matched controls (LL: 6MWT; UL: hand strength, pegboards; F: tongue kinematics).For three patients, follow-up behaviour measurements were performed.The results show that ALS patients are significantly impaired in motor function compared to matched controls (Supplementary Table 1).

No significant M1 differences in patients when averaged across depths
We first tested whether the averaged qT1 and QSM values [averaged across cortical fields (LL, UL, F) and layers (L6, L5b, L5a, Ls) of M1] were significantly different in ALS patients compared to controls, to align with previous analyses. 12,16Using this approach, we show no significant differences in pQSM values between patients and controls in left M1 [patients: mean = 0.0172, SD = 0.0026; controls: mean = 0.0171, SD = 0.0028; t( 7 Finally, there are no significant differences in mean cortical thickness between patients and controls, although there is a trend towards significantly reduced cortical thickness in the patients compared to controls in the F field (Supplementary Table 2).

Layer-Specific M1 pathology in ALS patients M1 layer architecture in ALS patients
As in our previous work, 10 we used decurved qT1 values to identify four layers in M1 (Ls = superficial layer; L5a = layer 5a; L5b = layer 5b; L6 = layer 6) (Fig. 1B).Our data show that those four anatomically-relevant layers can be identified similarly for ALS patients and controls, where layers showed the same peaks (e.g. level set 11 for L5b) and relative thicknesses (e.g. level set 9-13 for L5b).In addition to the thickness, the principal layer architecture of M1 therefore also appears to be preserved in patients with early stage ALS (Fig. 1B), also in the affected cortical fields (Fig. 1C), although there are differences in the profile shape.

Iron accumulates in first-affected cortical field
Using topographic layer imaging, we created in vivo pathology maps for each individual ALS patient in reference to their respective matched control (Fig. 2 and Supplementary Fig. 2 for patients with qT1 data only).These maps reveal that demyelination is minimal compared to iron and calcium accumulation.As expected, [17][18][19] iron accumulation is higher in the first-affected cortical field compared to the other fields (excluding the contralateral field given evidence of highly symmetric pQSM increases 19 ) (filled red arrows in Fig. 2).This effect is specific to  In vivo ALS pathology BRAIN 2024: 147; 1087-1099 | 1093 95% CI (−0.570.65)].However, based on visual inspection, iron accumulation also occurs in unaffected body parts (non-filled red arrows in Fig. 2).An exception is Patient P12, who does not show iron accumulation, but shows demyelination in the F field, which may correspond to the bulbar-onset type.Linear regression analyses show that pQSM pathology (nor qT1 pathology or nQSM pathology) does not significantly predict whether a given cortical field is behaviourally affected (Supplementary Table 3).

Calcium accumulation is non-topographic and encompasses low myelin borders
The pathology maps (Fig. 2) provide a novel overview of calcium dysregulation in early-stage ALS, which suggests that calcium accumulation is atopographic.We show significantly decreased nQSM in the low-myelin borders (between the UL and F fields) compared to the cortical fields themselves (averaged UL and F fields), specifically in Ls [borders: mean = −0.0116,SD = 0.002; cortical fields: mean = 0.0159, SD = 0.027; t( 6

Multiple hot spots of pathology in M1
We also addressed whether in vivo pathology maps favour the hypothesis of one large pathology hot spot or multiple small hot spots of pathology.The maps reveal that pathology occurs in multiple, small areas, often beyond the first-affected cortical field (Fig. 2).For example, Patient P1 shows iron accumulation in an area between the left LL and UL fields compared to the control, while this patient shows demyelination in the (first-affected) left UL field.

Calcium accumulation precedes demyelination
Out of the 12 patients, three (Patients P1, P2 and P4) were also measured longitudinally (Fig. 4).To provide an overview over disease progression, we quantified the increase in the percentage of demyelinated vertices (+2 SD compared to matched control mean) in the first-affected cortical field between timepoints (Supplementary Table 4).In addition, we show that the topography of pathological calcium (vertices with QSM values > than −2 SD from the matched control mean) at T1 predicts myelination at T2 and T3, most strongly in Ls (Supplementary Table 5).We also show that the topography of pathological iron (vertices with QSM  values > +2 SD from the matched control M1 mean) at T1 predicts myelination at T2 (but not T3), specifically in L6 (Supplementary Table 6).However, the latter effect did not survive correction for multiple comparisons.

Disrupted low myelin borders with disease progression
We also addressed whether the low-myelin borders that separate the F and UL cortical fields in M1 are affected or preserved in ALS, to identify their role in disease progression.The longitudinal pathology maps show demyelination of the F-UL border at T2 compared to T1 (Patients P1, P2 and P4 in Fig. 4), and at T3 compared to T2 (Patient P4 in Fig. 4).Based on visual inspection and the quantified demyelination (Supplementary Table 4), this effect is largely restricted to Ls and L5a, where the low myelin borders are typically most defined in M1 of healthy adults. 10,23

Multimodal MRI markers of slow disease progression
To identify cortical markers related to slow disease progression, we compared the individualized in vivo pathology maps between the very slow progressing patient (Patient P4) and all other patients.We show that Patient P4 shows a significantly higher percentage of abnormal pQSM vertices (i.e.increased iron accumulation) in L5b

Discussion
ALS is a rapidly progressing neurodegenerative disease characterized by the loss of motor control. 1 The present study aimed to systematically describe the in vivo pathology of M1 in early stage ALS patients.Given the layer-specific and topographic signature of pathology in ALS, 5 we combined submillimetre structural 7 T MRI data (qT1, QSM) and automated layer modelling with cortical fieldspecific analyses to calculate precise in vivo histology profiles of the patients.Our data reveal critical insights into the in vivo pathology of ALS that can be summarized as (i) preserved mean cortical thick-   3 and Figs 2 and 3).With disease progression, we highlight increasing demyelination (Supplementary Table 4), particularly in Ls and L5a and in the low-myelin borders between adjacent cortical fields (see Fig. 4, based on visual inspection), corresponding to calcium accumulation at earlier time points (Supplementary Table 5).L6 = layer 6; L5b = layer 5b; L5a = layer 5a; Ls = superficial layer.
Across cortical depth in M1, we show no significant differences in microstructure or mean cortical thickness between patients and controls.7][58] However, the patients in the present study show a relatively slower disease progression rate compared to a previous 7 T study. 16Moreover, our cortical thickness estimates of healthy controls are comparable to previous automated 7 T estimates. 16,59We show a trend towards reduced mean cortical thickness in the F field of patients compared to controls.This field-specific trend may reflect a higher vulnerability of this field to neurodegeneration, as it has been shown that myelination is reduced in the F field compared to the other cortical fields of M1. 10 Our findings therefore confirm that layer-specific microstructural pathology in ALS is detectable in the absence of significant cortical atrophy.
Layer-specific analyses reveal that iron accumulation in ALS is specific to L6 of M1.Although this effect was moderate-to-large, it is critical to replicate this finding in a larger cohort of patients given the high heterogeneity of the disorder.Nevertheless, this finding extends previous studies showing deep iron accumulation in ALS, which is considered to reflect neuroinflammation and activated microglia. 6,15This suggests that iron accumulation in L6 may provide an earlier marker of pathology than the well established degeneration of L5(b). 5Our data also show that iron accumulates most strongly in the first-affected cortical field, as expected, [17][18][19] and that this is specific to L6.However, in our data, the degree of iron accumulation (% of pQSM pathology) does not predict whether a given cortical field is behaviourally affected.This may suggest that iron accumulation largely serves as a specific marker for the first-affected cortical field, which may support early diagnosis, and that this marker is a consequence rather than the cause of the disease.
Surface mapping reveals multiple, small hot spots of pathology rather than one large hot spot.For example, one patient (Patient P1) shows L6 iron accumulation in an area between the LL and UL fields, which may reflect pathology in the torso region where impairment is difficult to detect.Another patient (Patient P2) shows iron accumulation in Ls-L5a of the first-affected (UL) field, deviating from the expected location in L6, but in all layers of the contralateral F field.The latter may reflect the quantitatively impaired tongue function in this patient, highlighting challenges in detecting bulbar symptoms clinically. 54This demonstrates potential mismatches between brain and behaviour in ALS, 60 emphasizing the utility of MRI for individualized medicine.
We also investigated calcium dysregulation, demonstrating that Ls (including L2-L3) and L5a show calcium accumulation in ALS.In addition, we show that calcium accumulation is higher in the areas between cortical fields, where the low myelin borders are located, compared to the cortical fields themselves.This highlights the atopographic profile of calcium accumulation in ALS, in contrast to the more topographic iron accumulation.Previous evidence has shown disrupted calcium homeostasis in ALS, 7 where increased intracellular calcium is associated with glutamate excitotoxicity in animal models. 20,21Increased calcium in humans may reflect similar mechanisms, including the increased activity of M1 in ALS 61 or calcifications in blood vessels related to poor brain health. 62Interestingly, patients with pronounced calcium accumulation (Patients P8 and P12) largely show more focal behavioural impairment in our data.This may suggest that calcium accumulation provides an early disease marker, which may be related to compensation, as has been suggested for increased activity in M1. 61 Based on a small sample of patients with longitudinal data, we provide evidence towards widespread demyelination of M1 with disease progression.We also show that later demyelination is higher at previous calcium accumulation hot spots, whereas the relation to previous iron accumulation hot spots is weaker.Moreover, the low myelin borders between cortical fields that are disrupted with disease progression show earlier increased calcium accumulation compared to cortical fields.Overall, this points towards a partly atopographic character of disease spread in ALS, and a potential role of the low myelin borders, which are connected to multiple body parts, 23,25 in disease spread.These findings also suggest that metabolism disruption and inflammation may precede cell loss in ALS. 63An alternative perspective to this interpretation is that connections between low myelin borders and subcortical structures 25 are more affected by pathology, compared to connections between cortical fields and subcortical structures.From this perspective, the affected border areas would not reflect markers of disease spread but would be the consequence of affected subcortical systems.Further investigation is needed to characterize the role of these atopographic disease mechanisms in ALS.
Finally, we also investigated whether the in vivo pathology profile of a very slow progressing patient (Patient P4) is distinct from the other patients.We demonstrate that Patient P4 shows increased iron accumulation in L5b and L5a, and increased calcium accumulation in L5a, compared to other patients.Moreover, Patient P4 shows decreased demyelination in all layers compared to the other patients.These results suggest that Patient P4, despite the long disease duration, may be in the earlier stage of cortical pathology (Fig. 5) that is characterized by increased substance accumulation.One may interpret this finding to argue that there are distinct pathology profiles for disease duration (i.e.longer survival associated with substance accumulation) and disease progression (i.e.increased disease severity associated with demyelination).More specifically, one may argue that the 'turning point' between substance accumulation and cell loss is delayed in the slow-progressing patient, rather than a delay in substance accumulation as such.These assumptions must be verified in a larger patient sample with more slow-and more fast-progressing patients to avoid overinterpretation based on single patient profiles.
With respect to the applied methodology, although the method we used to define cortical layers in M1 is based on an in vivo-ex vivo validation model, 9,10 we cannot guarantee that our layer definitions correspond to the exact biological layers.This limitation should be addressed with more extensive in vivo-ex vivo validation studies.Finally, although quantitative and validated markers of cortical microstructure were used here, we cannot be certain of the exact tissue properties measured.
In summary, we provide novel insights into the in vivo pathology of ALS patients using topographic layer imaging.We show that layer-specific changes in tissue microstructure precede gross cortical atrophy.We highlight the topographic nature of iron accumulation as a marker for diagnosis, while atopographic calcium accumulation may precede demyelination and cell loss.The role of the low myelin borders between cortical fields, which show particularly high calcium accumulation, in disease progression needs further clarification.Finally, we highlight the distinct pathology profile of a very slow progressing patient, where increased substance accumulation and reduced demyelination may indicate a lack of pathology progression despite a long disease duration.Our efficient 7 T MRI scanning protocol, and use of open-source analysis tools, make our approach accessible to large cohort investigations.

Figure 1
Figure 1 Microstructure profiles of the left primary motor cortex (M1) in ALS patients and matched controls.(A) 'Raw qT1' (i.e.not decurved; n = 12), decurved qT1 (n = 12), positive QSM (pQSM; n = 8) and negative QSM (nQSM; n = 8) data extracted across all cortical depths (n = 21) of left M1.The first and second columns show data for all controls and all patients, with the group mean plotted in bold black and red, respectively.The third column shows the mean group data of the controls and patients, with red lines representing the patients.Note that qT1 and pQSM are validated in vivo markers of myelin27 and iron 6,15 content, respectively, while nQSM is largely considered to reflect calcium28 content.Also note that the profiles here represent the data of the entire M1, while in other figures and statistics the data used were averaged across cortical fields.(B) According to a previously published approach,9 we identified four compartments ('layers') based on the 'decurved qT1' profile: Ls = superficial layer including layers 2-3; L5a = layer 5a; L5b = layer 5b; L6 = layer 6.Note that Ls does not include layer 1 as it is inaccessible with MRI 9 or layer 4, as it is absent in M1.We show the layer definitions for healthy controls (n = 12) and ALS patients (n = 12) in the present study (centre), as well as for older adults (n = 18) in our previous study 10 (left).L5a and L5b were distinguished based on the presence of two small qT1 dips at the plateau of 'decurved qT1' values (indicating L5), while L6 was identified based on a sharp decrease in values before a further plateau indicating the presence of white matter.We show our layer approximations over schematic depictions (right) of M1 myelin46 and cell histological staining.47(C) Despite differences in the shape of the profile in the affected fields compared to the average across fields in patients, layers can be similarly identified in the affected fields.
to +4 SDs), iron accumulation (increased pQSM, +1 SD to +4 SDs) and calcium accumulation (more negative nQSM, −1 SD to −4 SDs) in each patient relative to the mean M1 value of the matched control.Pathology estimations based on 1-4 SD are often used in clinical diagnostic imaging tools as they provide a detailed overview of how much the respective pathology differs from a control brain (e.g.AIRAmed Software, https://www.airamed.de/de/startseite).

Figure 2
Figure 2 Multimodal in vivo pathology maps in ALS patients.Pathology maps were generated for each patient by thresholding the displayed value ranges at each layer to show increased qT1 and pQSM (+1 to +4 SD), and reduced nQSM (−1 to −4 SD), with respect to the mean M1 value of the matched control.Subject-specific cortical fields representing the lower limb, upper limb and face areas are outlined in yellow, red and blue, respectively.Pathology maps are shown for Patients P1, P2, P8, P12, P6, P7, P4 and P5 (left) and their corresponding matched controls C1, C2, C8, C12, C6, C7, C4 and C5 (right).Pathology maps: note that filled red and blue arrows indicate iron accumulation and demyelination in the first-affected cortical field, respectively.Unfilled red arrows indicate iron accumulation in cortical fields other than the first-affected field.Body maps: note that red-outlined circles on the body maps indicate the onset site (i.e.first-affected body part) of the patient.Filled red and green circles indicate impaired or better motor function in the circled body part compared to the matched control, respectively.The stage indicates the King's College (KC) stage 32 of disease progression based on ALSFRS-R score 33 : stage 2A reflects the involvement of one body part and that a clinical diagnosis has taken place, while stage 2B and stage 3 reflect the subsequent involvement of second and third body parts, respectively.L6 = layer 6; L5b = layer 5b; L5a = layer 5a; Ls = superficial layer; nQSM = negative QSM; pQSM = positive QSM; qT1 = quantitative T1.

[t( 6 )Figure 4
Figure 4 Longitudinal multimodal in vivo maps in ALS patients.Individualized in vivo pathology maps were generated by thresholding the displayed value ranges at each layer to show increased qT1 and pQSM (+1 to +4 SD), and reduced nQSM (−1 to −4 SD), with respect to the mean M1 value of the matched control at baseline.Row 1 shows the pathology maps for Patient P1 at T1 (time point 1) and T2 (time point 2), row 2 shows the pathology maps for Patient P2 at T1 and T2 and row 3 shows the pathology maps for Patient P4 at T1, T2 and T3 (time point 3).Pathology maps: red and blue arrows on the pathology maps indicate iron accumulation and demyelination in the cortical field corresponding to symptom onset site, respectively.Body maps: red-outlined circles on the body maps indicate the onset site of the patient, while filled red and green circles indicate impaired or better motor function in the circled body part compared to the matched control, respectively.QSM data were excluded at T2 for Patients P1 and P2 due to severe artefacts.Clinical information: ALS Functional Rating Scale-Revised (ALSFRS-R) 31 indicates disease severity, where lower values indicate greater impairment, with subscores for fine, gross and bulbar motor function.The Penn Upper Motor Neuron Scale (PUMNS) 29 score indicates clinical signs of upper motor neuron involvement, with higher scores indicating greater impairment.The King's College (KC) stage 32 indicates disease progression based on ALSFRS-R score 33 : stage 2A reflects the involvement of one body part and that a clinical diagnosis has taken place, while stage 2B and stage 3 reflect the subsequent involvement of second and third body parts, respectively.L6 = layer 6; L5b = layer 5b; L5a = layer 5a; Ls = superficial layer; nQSM = negative QSM; pQSM = positive QSM; qT1 = quantitative T1; LL = lower limb; UL = upper limb; F = face.
ness and preserved layer architecture of M1; (ii) layer-specific iron (layer 6) and calcium (layer 5a, superficial layer) accumulation; (iii) topographic iron accumulation, but atopographic calcium accumulation; (iv) disrupted low myelin borders with increased calcium; (v) later demyelination may occur more in the earlier hot spots of calcium accumulation and does therefore not show a strictly topographic profile; (vi) demyelination characterizes disease progression (particularly in layer 5a and superficial layer); and (vii) the very slow-progressing patient has a distinct pathology profile in M1 compared to the other patients.Overall, our study provides novel insights into the in vivo M1 pathology in ALS, offering new perspectives into disease mechanisms and diagnosis (Fig. 5).

Figure 5
Figure 5 Model of in vivo M1 pathology progression in ALS.Schematic depiction of the layer-specific pathology features shown in ALS patients compared to matched healthy controls in the present study.In the early pathology stage, we demonstrate topographic (i.e.most in first-affected cortical field-example upper limb shown here) iron accumulation in L6 and non-topographic (i.e. more widespread) calcium accumulation in Ls and L5a (see Table3and Figs2 and 3).With disease progression, we highlight increasing demyelination (Supplementary Table4), particularly in Ls and L5a and in the low-myelin borders between adjacent cortical fields (see Fig.4, based on visual inspection), corresponding to calcium accumulation at earlier time points (Supplementary Table5).L6 = layer 6; L5b = layer 5b; L5a = layer 5a; Ls = superficial layer.