Background R2′ is an MRI measure of microscopic magnetic field inhomogeneity, and is increased by the paramagnetic effect of iron and the diamagnetic effect of myelin. R2′ may detect features of multiple sclerosis (MS) not evident with conventional MRI.
Methods Multiecho T2 and T2* weighted sequences were obtained from 21 healthy controls (nine men, 12 women; mean age 36 years) and 28 MS patients (seven men, 21 women; 18 relapsing remitting, 10 secondary progressive; mean age 42 years). T2 and T2* relaxation time maps were created from the multiecho sequences, and R2′ maps were created using the formula R2′ = R2*−R2 = 1/T2*−1/T2. R2′ was measured in MS white matter lesions and in regions of interest in normal appearing white matter (NAWM) and grey matter in all subjects.
Results R2′ was reduced in NAWM in MS compared with controls (9.5/s vs 10.1/s, p=0.05). R2′ was additionally reduced in lesions, both T1 isointense (8.5/s vs 9.5/s, p=0.02) and T1 hypointense (7.7/s vs 9.5/s, p=0.003) compared with NAWM. R2′ tended to be higher in the basal ganglia of MS patients compared with controls, and was significantly higher in the caudate nucleus in secondary progressive MS (12.9/s vs 10.9/s, p=0.03). Increased T2 lesion volume predicted an increase in R2′ in the caudate (β=0.412, p=0.02).
Conclusions Reduction in R2′ in NAWM and lesions is consistent with a decreases in myelin, tissue iron and/or deoxyhaemoglobin. Increased caudate R2′ in patients with secondary progressive MS is consistent with increased iron deposition, as corroborated by other techniques.
Statistics from Altmetric.com
Multiple sclerosis (MS) is a chronic disease characterised by multifocal demyelinating white matter lesions disseminated in time and space within the CNS.1 These lesions are readily visualised with conventional MRI sequences2 and reduction in the frequency of appearance of new lesions by disease modifying therapies is associated with a reduction in relapses in the early phases of relapsing remitting MS.3 However, such treatments are less successful in preventing the accumulation of fixed disability, indicating that there are pathophysiological mechanisms in MS not visualised by conventional MRI.4 A likely explanation for this clinicoradiological paradox is that conventional MRI sequences do not directly identify neuroaxonal loss, which is the pathological substrate of irreversible disability in MS, and occurs in normal appearing white matter (NAWM)5 ,6 and grey matter,7 in addition to white matter lesions.8
Current understandings of the mechanisms of neuroaxonal loss in MS are limited. Experimental and animal studies suggest that disturbance in neuronal metabolism may be a common pathway to progressive neuroaxonal loss.9 Suggested mechanisms include mitochondrial dysfunction, hypermetabolism and hypoperfusion.10 Accumulation of iron could also lead to tissue damage by catalysing the production of free radicals, increasing oxidative stress.11 Focal accumulation of iron has been seen in histopathological studies in MS adjacent to veins,12 and at the peripheries of lesions,13–15 however this is not seen in all lesions, nor in all patients.16 MRI changes suggestive of increased iron are also seen in the basal ganglia using T2 weighted images17 and R2* mapping.18
Measurement of microscopic magnetic field inhomogeneity may help to understand the role of iron deposition and metabolic disturbance in neuroaxonal injury. Static intra-voxel magnetic field inhomogeneity increases the R2* (1/T2*) relaxation rate. The increase in R2* can be substantially reversed using a radiofrequency refocusing pulse, which allows calculation of the underlying R2 (1/T2) relaxation rate. The R2′ rate, calculated by subtracting R2 from R2*, describes the reversible part of the signal decay caused by static intra-voxel field inhomogeneities. R2′ is a quantitative measure that is independent from tissue relaxation rate and proton density, and changes in R2′ can be compared within MS lesions, NAWM and grey matter, where tissue relaxation rates vary.19
Tissue iron and deoxyhaemoglobin are paramagnetic: when placed in an external magnetic field their magnetic moment aligns along the field, causing a slight increase in magnetic field strength and in intra-voxel field inhomogeneity; thus, increase in tissue iron, or deoxyhaemoglobin (from increased oxidative metabolism), will increase R2′.20 Myelin is diamagnetic: when placed in an external magnetic field its intrinsic magnetic moment aligns against the field, causing reduction in magnetic field strength but also an increase in magnetic field inhomogeneity.21 Thus myelin also increases R2′.22
Reduction in R2′ may therefore be due to reduction in paramagnetic effect from reduction in tissue iron and deoxyhaemoglobin and/or reduction in diamagnetic effect from decreased myelin content.
In this study, we investigated for potential evidence of these processes in vivo by measuring R2′ in NAWM, grey matter and lesions in patients with MS.
Materials and methods
The study was approved by the University College London and University College London Hospital's joint research ethics committee. Written informed consent was obtained from all participants. Patients were recruited from MS clinics within the hospital, and healthy controls were recruited from a registry. The inclusion criteria were diagnosis of clinically definite MS of either the relapsing remitting or secondary progressive type for the patients,23 ,24 no neurological disease in controls and age between 18 and 60 years. Exclusion criteria were presence of other neurological conditions confirmed by history and neurological examination, contraindication to MR imaging, pregnancy or breastfeeding, and cardiac, respiratory or vascular disease.
Patients and controls had a full medical and neurological history and examination. Patients with MS had quantification of disability by means of the Expanded Disability Status Scale score, 25 ft walk, 9 hole peg test of both hands and Paced Auditory Serial Addition Test B 3 s (PASAT-B 3s) from which the MS functional composite was calculated. Clinical assessment was performed immediately before the MRI scan.
Twenty-one healthy controls, 18 patients with relapsing remitting MS and 10 patients with secondary progressive MS were scanned. Of the patients with MS, 15 were taking disease modifying treatments, with eight taking β interferon 1a preparations, five taking β interferon 1b and two taking glatiramer acetate. Thirteen patients were taking no disease modifying treatment. Scans were performed between November 2010 and September 2011. Characteristics of the subjects scanned are shown in table 1. All of the healthy controls had a normal MRI scan without any lesions.
MRI imaging acquisition
Imaging was performed in all patients using a 3.0 T MRI scanner (Achieva; Philips Healthcare, Best, The Netherlands) using a 32 channel receive coil. Image orientation parallel to the subcallosal line was achieved by acquiring a multiplanar T1 weighted localisation sequence at the beginning of the study.
All images were acquired with identical positioning and brain coverage, using a field of view of 240×180×152 mm3, 76 contiguous 2 mm thick slices and voxel size of 1×1×2 mm3. For T2 determination a fast spin echo sequence was used to acquire images at seven different echo times with the following parameters: TE1 16 ms, ∆TE 16 ms, TR 5230 ms, NEX 1, acquisition time 6 min 48 s. For T2* determination, a gradient echo sequence was used, acquiring images at 10 different echo times with the following parameters: TE1 7.1 ms, ∆TE 6.3 ms, TR 5667 ms, NEX 1, flip angle 90°, acquisition time 8 min 47 s, SENSE=2. A T1 weighted spin echo sequence was also acquired with parameters: TE 10 ms, TR 625 ms, NEX 1, flip angle 90°, acquisition time 9 min 30 s. Weekly quality assurance tests on phantoms were performed throughout the period of the study to ensure the stability of measurements.
Image post processing was performed on a Sun Ultra 20 workstation running the UNIX operating system. The VTK CISG toolkit (Kitware Inc, NewYork, USA, http://www.vtk.org/) was used to register images obtained at later echo times to the image obtained at the shortest echo time in the T2 and T2* weighted sequences using sinc interpolation. This was performed because without this step we found that small patient movement during sequence acquisition could lead to artefacts in the R2′ map, particularly at the boundary between areas of differing T2. T2 and T2* relaxation times were calculated from the registered multiecho sequences by solving the equation using in house software, where S=MR signal at time TE and S0= MR signal at TE=0. Since the signal to noise ratio of all images at all echo times was significantly >2, noise was modelled as adhering to a Gaussian distribution. The VTK CISG toolkit was used to register the first echo of the T2* weighted sequence to the first echo of the T2 weighted sequence, and this transformation was applied to the T2* relaxation time map using sinc interpolation. The transformed T2* relaxation time map was used with the T2 relaxation time map to create R2′ maps using the image analysis toolkit of JIM 6.0 (Xinapse systems, Northants, UK, http://www.xinapse.com) and the formula R2′=R2*−R2=1/T2*−1/T2.
Lesion and region of interest identification
T2 lesions were identified using the registered fourth echo of the multiecho spin echo sequence, corresponding to a T2 weighted image with TE of 64 ms, and T1 hypointense regions were identified using the T1 weighted sequence. All visible lesions were outlined using the semiautomatic lesion segmenting tool in JIM 6.0 and these regions of interest (ROI) were combined into masks. These masks were applied to R2′ maps, and all of the voxels covered were averaged to calculate a single mean R2′ from each patient for all lesions, and for T1 hypointense and T1 isointense lesions, separately. These values were used to compare R2′ between lesions and NAWM, and between T1 hypointense and T1 isointense lesions. In addition, a single mean R2′ was calculated for every lesion in each patient and these individual lesion values were used to create a histogram to demonstrate the spread of R2′ values in lesions.
Using the T2 weighted image and coregistered T1 image, the caudate, globus pallidus and putamen nucleus were outlined manually. Six ROI of 30 mm3 in size were placed in the thalamus. Thirty ROI of 76 mm3 were placed in the NAWM: 10 in the frontal lobe, six in the occipital lobe, six in the parietal lobe, four in the temporal lobe and four in the corpus callosum. Two hundred and sixteen ROI of 8 mm3 were placed in the normal appearing grey matter: 102 along the normal appearing grey matter of the frontal cortex, 18 along the insula cortex, 24 along the occipital cortex, 36 along the parietal cortex and 36 along the temporal lobe cortex. These ROI were combined together to create masks of frontal, occipital, parietal and temporal NAWM, total NAWM, thalamus, caudate, globus pallidus, putamen and cortical grey matter. These masks were applied to R2′ maps and all the voxels covered were averaged to calculate a single mean R2′ for each mask per subject.
Statistical analysis was performed using a Windows PC running SPSS V.19. General linear model for repeated measures was used to assess the significance of the difference in R2′ between regions in controls, and between T1 hypointense lesions, T1 isointense lesions and NAWM in patients with MS, with Bonferroni correction for multiple comparisons. Multivariate linear regression was used to assess for association between regional R2′ values with age in controls and to assess for association between R2′ values in caudate and NAWM, and T2 lesion volume with covariate of age. General linear model was used to assess for the significance of difference between controls and patients with MS with the covariate of age, and between subgroups of relapsing remitting and secondary progressive MS and controls with Bonferroni correction for multiple comparisons. A corrected two sided p value of <0.05 was considered to indicate a statistically significant difference or association. Results are presented as mean R2′ ± SD between subjects.
R2′ values in controls
In controls, a trend towards higher R2′ was seen in the white matter compared with cortical grey matter (10.1±1.0 vs 9.2±1.0/s, p=0.083). R2′ was higher in the caudate (11.1±1.2 vs 9.1±1.1/s, p<0.001), thalamus (11.3±2.3 vs 9.2±1.1/s, p=0.03), putamen (13.8±1.7 vs 9.2±1.1/s, p<0.001) and globus pallidus (18.7±3.5 vs 9.2±1.0/s, p<0.001) compared with the cortical grey matter (table 2) (figure 1). Increasing age predicted a higher R2′ in the caudate, (β=0.52, p=0.03), thalamus (β=0.54, p=0.02), putamen (β=0.66, p=0.04) and globus pallidus (β=0.51, p=0.04) (figure 2). No significant association was seen between age and R2′ in white matter (β =−0.18, p=0.49) or cortical grey matter (β=0.39, p=0.12). No significant differences were seen between women and men in any regions.
R2′ was lower in NAWM in patients with MS
R2′ was significantly lower in NAWM of MS patients compared with controls when all regions were combined (9.5±1.0 vs 10.1±1.0/s, p=0.05) (figure 3A). Table 2 shows NAWM R2′ values by individual region. While a trend towards reduction in R2′ was seen in all white matter regions, it was only significant following Bonferroni correction for multiple comparisons in the frontal region (8.4±1.1 vs 9.2±1.1/s, p=0.032).
R2′ was lower in lesions in patients with MS
There was considerable heterogeneity in mean R2 values in individual lesions. Most lesions had lower or similar R2′ to the NAWM and 29% of lesions had R2′ values 2 SDs or lower than the NAWM mean (figure 4). Lesions with lower R2′ than surrounding NAWM can be seen in figure 5A. A smaller number of lesions had higher R2′, and 16% of lesions had R2′ values more than 2 SDs higher than the NAWM mean. A lesion with increased R2′ can be seen in figure 5B.
When a mean value of R2′was obtained for the entire lesion load of each patient, R2′ was significantly lower in lesions compared with NAWM (8.2±1.5 vs 9.5± 1.1/s, p<0.001). This was evident in T2 lesions that were T1 isointense (8.5±1.5 vs 9.5±1.1/s, p=0.02) and T1 hypointense (7.7±2.4 vs 9.5±1.1/s, p=0.003). A non-significant trend towards reduction in R2′ in T1 hypointense lesions compared with T1 isointense lesions was seen (7.8±2.4 vs 8.3±1.5/s, p=0.38).
R2′ tended to be higher in deep grey matter in patients with MS
R2′ showed a non-significant trend to be higher in deep grey matter structures of MS patients compared with controls (table 2). A significant increase in R2′ was seen in patients with secondary progressive MS in the caudate (12.9±1.5 vs 10.9±1.2/s, p=0.03), with a strong trend towards increase in the globus pallidus (22.5±4.1 vs 18.5±2.3/s, p=0.06). Increased R2′ could also be seen visually within the caudate nucleus and globus pallidus in some patients (figure 3B). There were no significant differences in cortical grey matter R2′ between MS and controls.
In patients with MS, using a multivariate regression model with additional predictor of age, increase in T2 lesion volume predicted an increase in R2′ in the caudate (β=0.412, p=0.02) (figure 6). There were no significant differences in R2′ values in the caudate, globus pallidus, putamen and thalamus in patients with relapsing remitting MS compared with controls.
No significant differences were seen in R2′ between patients with relapsing remitting and secondary progressive MS in ROI in NAWM, lesions or cortical grey matter. No significant association was seen between R2′ in the caudate, or NAWM and MS functional composite or its component scores.
We found significantly decreased R2′ in MS lesions and NAWM and a trend for increased R2′ in basal ganglia in MS patients compared with healthy controls. These findings will be discussed in turn.
Decreased R2′ in MS lesions
A decrease in R2′ could be mediated by a decrease in tissue iron, deoxyhaemoglobin or myelin. Phase imaging, along with relaxation time measurement, offers a way to explore and distinguish the paramagnetic effects of iron from the diamagnetic effects of myelin, since although the presence of either will increase intra-voxel inhomogeneity, their phase effect differ: a paramagnetic effect being negative and a diamagnetic effect positive.15 A recent study, with both in vivo MRI and post mortem MRI histopathology data, showed that a majority of MS lesions exhibited neutral phase with decreased R2* that corresponded with loss of both tissue iron and myelin.15 The same study found that a significant minority of lesions had a negative phase with variable R2*, which was consistent with decreased myelin and variable iron deposition. Another study that used susceptibility weighted imaging to study MS lesions has reported an increase in lesion susceptibility compared with NAWM.25 The authors proposed that this indicated an increase in lesions iron content: however, the observed changed could also be explained by a reduction in diamagnetic myelin.21
Because demyelination is a hallmark of MS lesions, loss of myelin will have contributed to the lower R2′ that we saw in lesions. In addition to loss of myelin, oligodendrocyte loss is a typical finding in longstanding MS lesions,1 and since the majority of cellular iron in the white matter is located in oligodendrocytes,26 it is plausible that a decrease in tissue iron also contributed to the lower lesion R2′, which was often quite marked.
A decrease in deoxyhaemoglobin due to reduced oxidative metabolism in the lesion or increased perfusion in excess of demand might also contribute to the lower lesion R2′. This would be consistent with studies indicating reduction in function of mitochondrial complexes I and IV in MS lesions,27 and with the occasional spectroscopic finding of increased lactate in lesions which may indicate compensatory anaerobic metabolism to maintain energy production.10 However, tissue iron has a larger paramagnetic effect than deoxyhaemoglobin in normal white matter,21 suggesting that the decrease in lesion R2′ is more likely due to loss of tissue iron and myelin.
While the mean lesion R2′ was lower than NAWM, there was considerable variability of individual lesion R2′ and a substantial number of lesions were in a range similar to or higher than that of normal white matter (figure 4). A normal lesion R2′ may identify lesions with relatively preserved oligodendrocyte content and thereby possibly having greater potential for remyelination. An increased lesion R2′ probably indicates an increase in tissue iron within the lesion. Iron deposits have been observed in a minority of lesions at post mortem and iron may be seen in activated microglia and macrophages at the lesion edge.14 We analysed the averaged R2′ of the whole lesion and a future study of voxel level variations in R2′ within lesions may help to detect the effects of more localised iron deposition.
Decreased R2′ in MS NAWM
The reduced R2′ observed in myelinated NAWM is perhaps more likely due to a decrease in paramagnetic effect—that is, reduced tissue iron or deoxyhaemoglobin content—although a contribution from a subtle decrease in myelin content is possible. The majority of cellular iron in white matter is located in oligodendrocytes,26 but although they may exhibit oxidative stress in MS NAWM,28 their numbers do not appear to be significantly reduced.29 Alternatively, there could be a reduction in oxidative metabolism in NAWM, causing a reduction in paramagnetic deoxyhaemoglobin within the white matter vasculature. Other study findings support such a mechanism: susceptibility weighted imaging of cerebral veins suggests reduced venous concentration of deoxyhaemoglobin and therefore reduction in oxidative metabolism,30 and histopathological studies show reduction in the activity of mitochondrial complexes I, III and IV in the NAWM.9 Neuronal mitochondrial DNA deletions and clonal expansion secondary to oxidative damage have been suggested as a possible cause.9 ,10 The smaller R2′ decrease in MS NAWM compared with lesions would also be consistent with the smaller R2′ effect of a decrease in deoxyhaemoglobin, whereas the larger R2′ effects in lesions reflect loss of tissue iron and myelin.
Increased R2′ in deep grey matter
We found higher R2′ values within the deep grey matter of controls compared with the cortex which was probably due to the presence of a greater amount of tissue iron within these nuclei, as is seen histologically in normal basal ganglia.31 Our observed regional variation of R2′ within the basal ganglia, with highest values seen in the globus pallidus, followed by the putamen and then the caudate, is in concordance with published values of iron concentration,31 and our finding of a correlation of R2′ in the caudate, globus pallidus and thalamus with age concurs with histological studies showing an increase in basal ganglia iron with age per se.11
Higher R2′ in the caudate in secondary progressive MS compared with controls might be explained by increased tissue iron deposition. Abnormalities reported in other MRI metrics sensitive to iron, including T2 hypointensities,17 R2* deviations18 or changes in magnetic field correlation imaging,32 and susceptibility weighted imaging25 are also consistent with increased iron deposition in the basal ganglia in MS.
Increased iron deposition within the basal ganglia is not specific to MS but appears in other diseases, including following severe ischaemic–anoxic insult, Parkinson's disease, Huntington's disease, acaeruloplasminaemia, neuroferritinopathy and pantothenate kinase associated neurodegeneration.11 Neuronal cell bodies within the basal ganglia express transferrin receptors and transporter proteins for iron,33 and autoradiographic studies in rats indicate iron accumulates first in the basal ganglia and then spreads throughout the brain in a pattern and speed consistent with intra-axonal transport.34 Our finding of a correlation between caudate R2′ and T2 lesion volume in MS may suggest that distal disruption of iron transport at the site of axonal damage in white matter lesions leads to proximal iron accumulation. This is in line with a previously reported correlation between lesion number and increase in magnetic field correlation in the thalamus and putamen,32 and basal ganglia T2 hypointensity,17 and increased R2* with T2 lesion load.18 Further support for this hypothesis is seen in histopathological studies that show iron deposition adjacent to lesions, suggesting local disruption of axonal iron transport.16
Two study limitations warrant discussion. First, the clinical subgroups were not large enough to robustly investigate potential differences between them. However, the significant increase in caudate R2′ in secondary progressive MS only is consistent with studies using other MR techniques sensitive to iron, which have shown greater basal ganglia abnormality in subjects with higher disability. Second, although there were significant reductions in R2′ in both MS lesions and NAWM, the extent to which a decrease in tissue iron, deoxyhaemoglobin and myelin contribute separately to these finding could not be measured. In addition, as contrast was not given, we were unable to distinguish between acute and chronic lesions. Future studies with additional imaging measures may elucidate these separate contributions—for example, phase imaging to separate paramagnetic (iron and deoxyhaemoglobin) and diamagnetic (myelin) effects; magnetisation transfer and diffusion tensor imaging to evaluate myelination; and MR techniques sensitive to the effects of inspired oxygen or near infrared spectroscopy to detect deoxyhaemaglobin.35
We found decreased R2′ in NAWM and lesions in patients with MS, suggesting reduction in tissue iron, reduction in oxidative metabolism or reduction in myelin. Substantial iron deposition in NAWM and most lesions is unlikely. We also found increased R2′ in the caudate in patients with secondary progressive MS, suggesting iron accumulation.
Funding This work was undertaken at UCLH/UCL who received a proportion of funding from the Department of Health's NIHR Biomedical Research Centres funding scheme, and was supported by grant code 6DFB. The NMR Research Unit is also supported by grants from the UK MS Society, and Philips Healthcare.
Competing interests Dr David Paling has received funding for travel from the European Committee for Treatment and Research in Multiple Sclerosis. Dr Daniel J Tozer position is partially funded by the commercial companies Biogen Idec and Novartis. This funding is for work on clinical trials unconnected with this work. Dr Claudia Wheeler-Kingshott has received funding from the commercial company Biogen Idec for consultancy work, unconnected with this work. Dr Raju Kapoor receives funding for serving on the advisory boards for the commercial companies Biogen Iden, Novartis and Genetech. Dr Kapoor has received funding for travel to medical conferences from the commercial companies Biogen Idec, Merk Serono and TEVA. Dr Kapoor has received research support from National MS Society, The MS society of Great Britain and Northern Ireland and the commercial company Novatis for work unconnected with this work. Professor David Miller has received research grants (held by University College London) from Biogen Idec Inc, GlaxoSmithKline, Schering AG, and Novartis to perform MRI analysis in multiple sclerosis trials. Professor Miller has also received honoraria and travel expenses for advisory committee work or as an invited speaker from Biogen Idec Inc, GlaxoKlineSmith, Bayer Schering, Novartis and the US National Institutes of Health. Professor Xavier Golay receives funding from the commercial company Philips for advisory work unconnected with this work. Professor Xavier Golay holds two patents for arterial spin labelling imaging unconnected with this work. Professor Xavier Golay serves on the journal editorial boards of MAGMA and NMR in biomedicine.
Ethics approval The study was reviewed and approved by the local research ethics committee NRES London–Bentham.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.