An objective of this study was to clarify conflicting reports regarding the utility of MRS as an HD biomarker modality. Our focus was mI, tNAA, and their differences between control, pre-HD, and early HD individuals. In early HD, we correlated metabolites with performance on standard and novel quantitative motor tasks sensitive to early deficits in HD. A strength of this study is the normalization of metabolites to unsuppressed water signal, a technique previously reported in HD spectroscopy studies.5,7,8
By avoiding normalization to other metabolites (e.g., tCr), which are also disturbed in the HD brain,8
these measurements more likely reflect true biochemical change and permit direct comparison between studies. Furthermore, unlike most HD MRS studies, which utilize 1.5 T magnets,7,9,10,16–18
3 T MRS provides improved signal-to-noise ratio (SNR). This study also has larger, more homogeneous subject groups than earlier studies and the longitudinal design of TRACK-HD allows future evaluations of the same subjects.
The role of NAA in neuronal function is poorly understood. Synthesized within mature neurons, tNAA concentration within gray matter reflects neuronal number and viability, while in white matter it is a marker of axonal density.19–22
Total NAA was 15% lower in early HD than controls. It also correlated with a glossomotography measure in pre-HD and across pre-HD and early HD. Total NAA correlated with DBS across pre-HD and early HD. In pre-HD we did not identify the recognized association of tNAA with DBS.10
A narrow DBS range in pre-HD, a likely indirect consequence of enforcing a strict UHDRS motor score upper limit, may have contributed to this lack of correlation. In early HD also, tNAA did not correlate with DBS. A possible explanation may be that loss of putaminal tissue in early HD is paralleled by tNAA reduction, theoretically causing absolute tNAA concentration to plateau with resulting lack of correlation with DBS. Our findings are consistent with reported lower tNAA in akinetic HD (−66%) compared to controls, and correlation of tNAA with motor dysfunction.7
Interpretation of findings from other studies is limited by reporting tNAA values normalized to other metabolites, obscuring individual metabolite changes. Lower striatal and thalamic tNAA/tCr has been reported in early HD.6,11
Mean tNAA/tCr was comparable in our groups (mean [SD], controls: 1.01 [0.10] vs pre-HD: 1.04 [0.09] vs early HD: 1.05 [0.17], p
> 0.05), suggesting that early HD absolute tNAA levels, in previous reports, may have been lower than our observations. NAA reductions have been reported on MR spectra of transgenic HD mice,23
and in putamen of HD-affected individuals.24
Lack of consistent striatal NAA and tNAA alteration in premanifest and early HD has been reported.8,10
Premanifest cohort heterogeneity, in time to predicted disease onset (range −7 to +29 years vs +7 to +18 years in our study), may have obscured metabolite variation in one study.8
Heterogeneity in our pre-HD subgroup was minimized by ensuring adherence to a maximum UHDRS motor score, a practice not specifically reported in other studies.8,10
mI is an osmolyte and astrocyte marker that is elevated in Alzheimer disease (AD).25,26
Manifest HD in adults and juveniles has been associated with high striatal mI.5,27
We demonstrate mI differences in early HD that correlate with UHDRS measured motor dysfunction. We found no relationship between mI and quantitative motor battery performance in pre-HD or early HD. mI/tNAA correlated less strongly with motor measures in early HD, and across pre-HD and early HD, than the correlations observed separately for mI and tNAA. Similar metabolite changes occur in asymptomatic individuals carrying gene defects associated with familial AD.28
Interestingly, in “premanifest AD,” tNAA was reduced, and tNAA/mI discriminated presymptomatic spectra from controls.28
While tNAA correlated with DBS across premanifest and early HD, it correlated more strongly with glossomotor dysfunction in these individuals. Together with mI, tNAA appears to have a potential role as a marker of HD-related motor decline. Putaminal metabolite changes are pertinent since this is a site of early atrophy, identifiable years prior to clinical diagnosis.29
Furthermore, putaminal atrophy correlates with psychomotor and motor deficits in premanifest HD30
and early HD.31
Our correlations suggest a possible role for MRS in evaluating the ability of an intervention to slow disease progression. It should be borne in mind that these cross-sectional assessments were performed at a single timepoint. Future longitudinal comparison is necessary to appreciate metabolite sensitivity to disease progression.
NAA and tCr were correlated against motor performance. NAA was less robustly measured than tNAA (11 measurement exclusions vs no exclusions) and is difficult to differentiate from NAAG, the other tNAA constituent. The lower NAA in pre-HD compared to controls was not accompanied by lower tNAA. NAA analyses should be treated cautiously, even though potential causes of NAAG artifact—spectral linewidth broadening (mean [SD], pre-HD: 8.00 [2.37] Hz vs controls: 7.55 [1.66] Hz, p > 0.05) and low SNRs (pre-HD: 9.36 [2.12] vs controls: 10.17 [2.18], p > 0.05)—were comparable in pre-HD and controls. NAA was 8% lower in pre-HD than controls. NAA correlated with quantitative motor performance almost consistently and more strongly than any other metabolite in early HD and across pre-HD and early HD.
In all groups tCr values were comparable to tNAA, while normally tNAA is about 30%–40% higher (depending on brain region).32
Our findings are supported by very similar, low tNAA/tCr ratios identified within basal ganglia structures of controls and premanifest and manifest HD individuals at 1.5 T and 7 T MRS.5,33
This possibly reflects increased basal energy metabolism within these structures.5
Since all spectra were analyzed using the same protocol with consistent scaling applied across metabolites, scaling difficulties would not have affected comparisons of estimated tNAA and tCr. Despite lower tCr in early HD, and reports of potential biomarker roles for tCr in HD,7
we were unable to correlate tCr with early HD motor performance. In pre-HD, and across pre-HD and early HD, tCr correlated with a number of quantitative motor measures. While this may represent altered energy metabolism,7
our tCr observations are likely to have been impacted by partial volume effects (discussed below).
Spectral quality was worst in early HD, with lower SNRs (mean [SD], early HD: 6.70 [2.00], controls: 10.17 [2.18], p
< 0.001) and broader spectral linewidths (early HD: 10.51 [2.67] Hz, controls: 7.55 [1.66] Hz, p
< 0.001) than controls. Movement artifact probably contributed by attenuating SNR.17
However, based on findings from a 4-T MRS study, we would expect the SNR reductions and linewidth increases seen in our early HD group to generate up to 4%–6% variation in tNAA and mI estimations.34
Small SNR or linewidth effects alone would therefore not account for larger mI and tNAA differences in early HD. Furthermore, at 1.5 T, marked SNR reduction and linewidth broadening either did not affect estimated tNAA or generated only 4% increases in tNAA and mI estimations.35,36
While the earliest HD neuropathologic changes occur in caudate and putamen,37
the putamen was selected to avoid averaging CSF into the brain parenchymal compartment. CSF partial volume effects may generate spurious metabolite measurements5
and is a risk with caudate voxel placement. Due to scanning time restrictions and the finding of early metabolic changes within the left striatum in premanifest HD,7
the left putamen was chosen as the structure of interest.
The use of larger voxels to improve SNR generates partial volume white matter effects through inclusion of nonputaminal tissue within the voxel. A 3-T MRS study of brain metabolites reported lower mI in white matter than gray matter.32
This suggests that partial volume white matter effects caused by loss of putaminal volume in transition from premanifest to early HD would diminish the mI difference observed. Hence concentrations presented may underestimate true mI differences. Studies indicate a cortical gray matter/white matter tNAA ratio between 0.8 and 1.2, in adulthood.38,39
We cannot guarantee that white matter components are not, partly, responsible for group tNAA differences, but our findings are consistent with smaller voxel studies that minimized partial volume effects.6,7
Higher tCho, lower Glu and tCr are found in white matter,32
making it impossible to know whether our observations reflect authentic pathophysiologic change. Consequently, tCr analyses presented here should be interpreted with care. Estimations of percentage putamen volumes per voxel were performed solely to ascertain the importance of partial volume effects in interpreting results (see appendices e-1 and e-2).
Subjects participating in TRACK-HD were demographically matched across international sites rather than at individual sites, explaining demographic discrepancies between our groups. Fewer early HD women may have impacted metabolite concentrations for this group () but regional brain MRS did not identify relevant gender differences in tNAA or mI.38
Mean age in pre-HD was lower than in early HD and control groups (which were comparable). Consequently, age differences would not explain metabolite differences in early HD.
Use of neurotropic medications was highest in early HD (). Antidepressant use in a total of over 30 depressed individuals was associated with either unchanged or increased cortical tNAA/tCr ratios at 1.5 T and 3 T.40
All neuroleptic medicated patients with early HD were taking atypical agents. Atypical antipsychotic medication usage in over 50 patients with schizophrenia did not impact water-normalized thalamic or caudate tNAA levels (1.5 T).e2,e3
At 1.5 T, an atypical neuroleptic caused 27% elevation in thalamic mI/H2
While one-third of patients with early HD could have experienced medication-induced mI elevations, this would not explain the 50% higher mean mI concentration. Although not all studies were water-normalized, the evidence suggests that neurotropic usage would not explain our findings.
MRS has advantages over PET in monitoring disease progression. MRS data collection is less expensive and time-consuming, and does not require IV tracer isotope injection.8
Additionally, MRS may be less prone to interoperator variability than some VBM biomarker applications.e7
These advantages support ongoing assessment of MRS as an HD biomarker modality.
The ability of MRS to track pathologic change at the biochemical level makes it potentially responsive to acute therapeutic interventions. In a creatine supplementation study in early HD, MRS identified cortical glutamate reductions within 10 weeks.18
Thus MRS can potentially identify early reversal of pathologic processes, a finding that would be unlikely with structural imaging. The longitudinal assessment of our cohort, and replication of this study in a second large premanifest and early HD cohort, ideally in the setting of a therapeutic trial, will be necessary to fully validate these findings.