The Combination of Metabolic Posterior Cingulate Cortical Abnormalities and Structural Asymmetries Improves the Differential Diagnosis Between Primary Progressive Aphasia and Alzheimer’s Disease

INTRODUCTION

Primary progressive aphasia (PPA) is a neurodegenerative disorder with heterogeneous neuropathological patterns, characterized by an isolated and gradual dissolution of language functions [1, 2]. Impairment of language abilities could also be an early manifestation of Alzheimer’s disease (AD), therefore a differential diagnosis based on clinical grounds alone is difficult, especially in the early stage of the disease [3]. A careful definition of the language phenotype, together with a multimodal neuroimaging approach, has been demonstrated to be the best diagnostic predictor of the underlying pathology [4, 5]. Significant progress has been made in the understanding of the relationship between regions of peak atrophy, brain metabolite alterations, and clinical findings in PPA patients; however, quantitative biomarkers able to differentiate the clinico-anatomical course of this disease from other neurodegenerative disorders remain relatively sparse [6].

It has been observed, initially from neuroimaging studies and later from autopsy-confirmed cases, that PPA patients showed greater atrophy severity, neuronal loss, and disease-specific proteinopathy in the left hemisphere, usually the language-dominant hemisphere [6, 7]. The asymmetry of neurodegeneration persists at the time of postmortem study, even many years after the onset of the first symptoms [8]. Despite recent evidence have established the left-lateralized atrophy of PPA patients [9], no studies provide quantitative parameters of cortical asymmetry that may be helpful to differentiate in vivo PPA patients from patients with other neurodegenerative disorders, such as AD.

Proton magnetic resonance spectroscopy (¹H-MRS) is a non-invasive MR technique able to detect metabolic neurodegenerative changes by the quantification of N acetyl-aspartate (NAA), neuro-axonal marker, and myo-Inositol (mI), glial marker, [10] and these metabolic abnormalities were shown to precede the evidence of atrophy on structural imaging [11]. In the last years, while a large body of literature focused on metabolite alterations of posterior cingulate cortex (PCC) in AD, mild cognitive impairment, and healthy older adults [12–14], only few studies used this technique in PPA patients [15, 16].

The aim of this study was to investigate the ability of the combination of quantitative measures of brain volume asymmetries and ¹H-MRS PCC to differentiate PPA from AD patients.

METHODS

This study included 19 PPA patients, 23 AD patients, and 18 healthy controls. All patients were referred to the Functional MR Unit, S. Orsola-Malpighi Hospital, Bologna (IT), to perform brain MR investigation as part of the diagnostic workup. Diagnosis was performed by neurologists experienced in neurodegenerative disorders, according to international criteria for PPA [1, 2] and AD [17]. Following the clinical classification of Gorno-Tempini and colleagues [9], the PPA group was composed of 12 patients with the non-fluent/agrammatic variant, 4 patients with the semantic variant, and 3 patients with the logopenic variant. To explore the general cognitive status all patients were administered the Mini-Mental State Examination (MMSE) [18]. Furthermore, the PPA group also underwent two language tests, specifically Verbal Associative Fluency Test [19] and Category Words Fluency Test [20]. All subjects gave consent to personal data processing for research purposes and the protocol was approved by the local Ethical Committee.

Demographic and clinical features of patients and healthy controls are summarized in Table 1.

All participants underwent a brain MR protocol using a 1.5 Tesla GE scanner, including high-resolution 3D T1-weighted FSPGR sequence and ¹H-MRS of PCC. Differences in brain volumetry and cortical thickness were evaluated on 3D T1 FSPGR images (TR = 12.5 ms, TE = 5.1 ms, TI = 600 ms, 1 mm³ isotropic voxels) using the freely available software FreeSurfer v5.3.0 (http://surfer.nmr.mgh.harvard.edu/). Asymmetry indexes (Left–Right)/(Left + Right) were also calculated for each brain area [21]. Negative asymmetry indexes indicate leftward asymmetry (left thinner than right) and positive asymmetry indices indicate rightward asymmetry.

Single voxel ¹H spectra were obtained from the PCC covering both hemispheres using the three planes of high resolution 3D T1 FSPGR sequence to optimize the localization (Fig. 1B). Suppressed-water proton MR spectra were acquired using the PRESS localization sequence (PROBE, TR = 4000 ms, TE = 35 ms, 128 averages, volume of interest = 8 cm³) [22].

OPEN IN VIEWER

Fig. 1

Posterior cingulate cortex (PCC) structural asymmetry and metabolic abnormalities: A) Distribution of asymmetry index of the PCC in PPA and AD patients (negative value indicate leftward asymmetry, left thinner than right, and positive value indicate rightward asymmetry); B1) Representative PCC MRS voxel localization projected onto sagittal plane of subject’s own T1-w image; B2) Distribution of NAA/mI ratio in PPA and AD patients indicating lower values in the AD group.

Peak areas were calculated using version 6.3 (http://www.lcmodel.com/) of the fitting program LCModel [23]. Spectra quality was visually examined by an expert spectroscopist. Signal to noise ratio (SNR) and full width at half maximum (FWHM) were also inspected (overall average and standard deviations being respectively of 15.2±3.5 and 3.3±0.7 Hz). The exclusion criteria for metabolite evaluation was LCModel estimated fitting error > 20%, this being a reliable indicator of poor-quality spectra [24]. No data were excluded due to low quality. Relative metabolite contents were considered using creatine (Cr) or myo-Inositol (mI) as a reference: N-acetyl-aspartate (NAA)/Cr, choline-containing compounds (Cho)/Cr, mI/Cr and NAA/mI ratios were evaluated.

The normality of the distribution of all parameters was tested using Shapiro-Wilk test and the results for each value showed a p-value >0.05. Therefore, T-tests and ANOVAs, followed by a Bonferroni post-hoc test for multiple comparisons, were used. In addition, to explore the association of these parameters with language deficits, Pearson’s correlation was run. In order to determine the level of accuracy of the most significant parameters, single and combined receiver operating characteristics (ROC) curves were also performed. Specifically, the multiparametric combined ROC curve was built by initially performing a binary logistic regression analysis constructing a model with two parameters that showed the highest discriminative power in single ROC analyses, the PCC volume asymmetry and the NAA/mI ratio; then the obtained predicted probability was used as test variable in the ROC analysis. Significance values were corrected by false discovery rate (FDR) and statistical significance was set at p < 0.05; all analyses were performed using IBM SPSS v.25.

RESULTS

All participants did not significantly differ in any of the demographic features. As expected, the global cognitive level, measured with the MMSE, was lower in the AD group compared to the PPA (Table 1). Compared with healthy controls, PPA and AD patients showed a widespread pattern of cerebral cortical volume reduction, respectively predominantly in fronto-temporal and temporo-parietal regions. Instead, comparing cortical volumes and cortical thickness between the two neurodegenerative diseases no statistically significant differences appeared: PPA and AD patients did not differ in the right nor in the left hemisphere.

A comparison of the asymmetry indexes showed significant differences in posterior cingulate volume (p < 0.001), inferior temporal volumes (p = 0.001), precuneus thickness (p = 0.002), inferior temporal (p = 0.003), and middle temporal regions thickness (p = .003), indicating a greater left-lateralized atrophy in the PPA group compared to AD (Fig. 1A, Supplementary Figure 1). Significant correlations were also found in the PPA group between the asymmetry of middle temporal regions and semantic fluency abilities (r = –0.495; p = 0.043).

ROC curve analyses of these indexes showed a level of accuracy ranging between 74.6%and 81.4%, and the PCC volume showed the highest level of accuracy with high sensitivity and high specificity, in discriminating between the two neurodegenerative groups.

As for PCC metabolic evaluation, Cho/Cr levels were significantly higher in AD patients compared to healthy controls (p = 0.006); however, only mI/Cr and NAA/mI ratios were significantly different between PPA and AD patients (p = 0.004 and p = 0.005, Fig. 1B, Table 2).

As for the discrimination between PPA and AD patients, ROC curve analyses showed that mI/Cr ratio has an accuracy of 77.2%(sensitivity 72.2%and specificity 69.6%), while NAA/mI ratio has an accuracy of 76.8%(sensitivity 72.2%and specificity 78.3%).

We carried out a further analysis to determine whether an incremental accuracy in the differential diagnosis could be achieved by combining the most significant asymmetry index with the proton MR spectroscopy parameters. The overall accuracy in distinguishing PPA from AD patients, obtained with this combined ROC curve, showed a strong increase from 81.4%(PCC volume asymmetry) and 76.8%(NAA/mI ratio) to 94%with a sensitivity of 91%and a specificity of 94%(Supplementary Figure 1).

DISCUSSION

The present study highlights that an asymmetrical volume reduction of temporo-parietal and limbic areas, in particular posterior cingulate cortex, together with a milder reduction of the NAA/mI ratio in bilateral PCC are highly sensitive indices to distinguish PPA from AD patients.

The majority of ¹H-MRS studies in neurodegenerative disorders focused on AD, mild cognitive impairment [12], Parkinson’s disease [25], and amyotrophic lateral sclerosis [26] patients compared with healthy older controls. Proton MRS has been shown to provide biomarkers for tracking early pre-dementia in AD [27]. Longitudinal increase in mI/Cr and decrease in NAA/mI are associated with cognitive decline and underlying amyloid pathology [28] and levels of mI are elevated even in asymptomatic stages of AD [29]. It was also shown that ¹H-MRS of PCC in combination with cerebral cortical volumetry can predict the progression from mild cognitive impairment to AD, aiding differential diagnosis at the early pre-dementia stages [12]. Moreover, a combination of automated structural MRI and MRS measures was also successfully applied to discriminate between AD and healthy controls [30].

To the best of our knowledge, in only one study proton MR spectroscopy data were evaluated in PPA versus AD patients [15]. Catani and colleagues studied a region corresponding to the paratrigonal white matter of both hemispheres with the aim of including the majority of the fibers of the superior longitudinal fasciculus, as component of the language network. In this study PPA patients in the early stage of disease, but not AD, showed, compared to healthy controls, reduction of NAA/Cr in the left side indicating neuro-axonal loss or dysfunction, and an increase of the mI/Cr, indicating glial activation, in both hemispheres [15].

The brain volume asymmetry that we found in PPA patients is consistent with prior longitudinal reports showing that neuronal loss and peak atrophy primarily occur in the language-dominant (mostly left) hemisphere [31]. Moreover, we also found an association between the asymmetry of middle temporal regions and semantic fluency abilities in the PPA group, highlighting the predominant role of the left hemisphere in language abilities. These data confirmed also the longitudinal results of Rogalski and colleagues that showed a decline in all language scores together with left greater than right hemisphere asymmetry in all PPA subtypes over a 2-year interval [31]. Compared with healthy controls, PPA and AD patients showed a widespread pattern of volume reduction; however, no significant differences appear in the left nor in the right hemisphere when comparing PPA with AD patients. This may be due to the heterogeneity of patients’ severity and different disease progression; for instance, left hemisphere degeneration could be more conspicuous in the initial stages of the disease in PPA patients compared to AD patients in the late stage of the disease.

Our findings specifically highlight that volume asymmetry indexes are instead a more accurate indicator able to clearly differentiate these two neurodegenerative disorders. Previous post-mortem study investigated hemispheric asymmetry indexes in a PPA population, demonstrating quantitative regional changes prevalent in the left compared to the right hemisphere in terms of higher distribution of TDP-43 inclusions associated with lower neurons and greater activated microglia density respectively [32]. Moreover, Mesulam and colleagues in a postmortem study showed that the most common and distinctive features for all pathologies associated with PPA were asymmetrically distributed [8]. Despite previous evidence having already established the presence of asymmetric neurodegeneration in PPA patients, the innovative aspect of our finding is the ability of this highly sensitive indicator, combined with metabolic PCC differences, in distinguishing in vivo PPA patients from patients with other neurodegenerative disorders, such as AD.

In conclusion, our study reveals, for the first time, that the combination of brain metabolic changes and brain cortical volume loss asymmetries, strongly increases the level of accuracy in the differential diagnosis between these two complex neurodegenerative disorders. However, a few limitations should be discussed. First, as the main limitation of the study, we should highlight the small sample size and the consequent difficulty to explore the potential differences between the three PPA variants. In addition, all patients included in this study were right-handed, therefore was not possible to explore the influence of handedness. Finally, no significant sex differences appeared, probably due to the non-homogeneous male/female ratio in each group and to the small size.

Additional studies with larger and well-matched samples are needed to confirm these encouraging results, determining whether they are generalizable to clinical practice.

DISCLOSURE STATEMENT

Authors’ disclosures available online (https://www.j-alz.com/manuscript-disclosures/21-0211r2).