Network-wide abnormalities explain memory variability in hippocampal amnesia

(A) Composite Memory scores for autoimmune LE patients and healthy controls; ‘max = 18’: the maximum score sum attainable for the autobiographical memory for childhood and early adulthood in the AMI (Kopelman et al., 1989), which does not use age-scaled / standardized scores; p-corr: p values are adjusted for multiple testing using the Holm-Bonferroni sequential method (n=3); U: Mann-Whitney U; Wt: Welch’s test; z: standardized age-scaled scores; *: Shapiro-Wilk test: p < 0.05. (B) Relationship of delay since symptom onset with structural and functional abnormalities. Given our hypothesis that broader network abnormalities unfold as a consequence of acute HPC atrophy (i.e. at a time-point after the acute focus of HPC damage), we also assessed the relationship of the delay between symptom onset and research participation with the extent of HPC and extra-HPC structural and functional abnormalities in a series of bivariate correlations. Of the structural/functional abnormalities (n=13) identified in our patient group, only inter-HPC rsFC decreased across patients as a function of the delay between symptom onset and research scan (rho=-0.58, p-corr=0.004). No other brain abnormalities showed this relationship, even at uncorrected levels (all rhos, |rho| ≤ 0.255; all ps, p-unc ≥ 0.122); bold: p-corr < 0.05; HPC: hippocampus; L, R: left, right (hemisphere); MPFC: medial prefrontal cortex; p-corr: p values are adjusted for multiple testing (n=13) using the Holm-Bonferroni sequential method; PCC: posterior cingulate cortex; PrCu: precuneus; rho: Spearmann’s rank correlation coefficient; rsALFF: resting-state amplitude of low frequency fluctuations; rsFC: resting-state functional connectivity; OPTIMA: Oxford Project To Investigate Memory and Aging; MAP: Memory and Amnesia Project; TIV: total intracranial volume; z-res: volumes are residualized against age, sex, scan source (MAP, OPTIMA), and TIV; functional abnormalities are residualized against age and sex. (C) Correlation of memory scores with structural / functional abnormalities across patients [analysis-level correction for multiple testing (n=39)]; bold: p-corr < 0.05; GM: gray matter; HPC: hippocampus; L: left hemisphere; M: medial; MAP: Memory and Amnesia Project; MPFC: medial prefrontal cortex; OPTIMA: Oxford Project To Investigate Memory and Aging; PCC: posterior cingulate cortex; p-corr: p values of bivariate correlations are corrected for multiple testing using the Holm-Bonferroni sequential method of correction for the total number of correlations (n = 39); PMC: posteromedial cortex; PrCu: precuneus; rsALFF: amplitude of low frequency fluctuations; rsFC: resting-state functional connectivity; R: right hemisphere; TIV: total intracranial volume; VBM: voxel-based morphometry; z-res: Mean rsALFF and rsFC values are residualized for age and sex across participants; volumes are residualized against age, sex, scan source (OPTIMA, MAP) and TIV across participants. (D) Correlation of memory scores with structural / functional abnormalities across patients [score-level correction for multiple testing (n=13), separately for each of the different memory scores (n=6) examined]; bold: p-corr < 0.05; HPC: hippocampus; MAP: Memory and Amnesia Project; MPFC: medial prefrontal cortex; OPTIMA: Oxford Project To Investigate Memory and Aging; PCC: posterior cingulate cortex; p-corr: p values of bivariate correlations are corrected for multiple testing using the Holm-Bonferroni sequential method of correction for the number of different variables (n = 13) per memory score examined; PrCu: precuneus; rsALFF: amplitude of low frequency fluctuations; rsFC: resting-state functional connectivity; VBM: voxel-based morphometry; z-res: Mean rsALFF and rsFC values are residualized for age and sex across participants; volumes (derived from manual / automated segmentation or from the mean GM volume expressed by VBM clusters) were residualized against age, sex, TIV, and scan source (MAP, OPTIMA) across participants. (E) Relationship of memory scores with volumes of manually delineated HPC portions. *: verbal / visual recognition, visual recall composite scores: No significant correlations were observed between those scores and the volumes of the manually delineated left/right anterior/posterior HPC portions at uncorrected levels (all ps, p-unc > 0.05); **: verbal recall composite scores: A weak correlation was observed with the left anterior HPC at uncorrected levels (r = 0.351, p-unc = 0.033; rest of ps, p-unc ≥ 0.095). Since patients’ rsALFF in the PCC did not correlate significantly with the volume of the manually delineated left anterior HPC (r = 0.333, p = 0.050), we entered these two factors as independent variables in a multiple step-wise linear regression, with verbal recall scores as the dependent variable. The regression terminated in a single step, with rsALFF in the PCC as the only predictor of patients’ performance (R²=0.34; β(z)=0.58; F=16.40, p < 0.0005); ***: visual forgetting scores: There was no clear evidence for a selective relationship of visual forgetting with anterior vs. posterior portions of the HPC (all rhos, 0.471≥ rho ≥ 0.332; all ps, 0.055 ≥ p-unc ≥ 0.005); ****: Correlations of remote autobiographical memory with volumes of manually delineated portions of the HPC (anterior / posterior right / left HPC) were observed at uncorrected levels (0.429 ≥ r ≥ 0.334; 0.066 ≥ p-unc ≥ 0.016). All four HPC portions correlated volumetrically with the left thalamus (right anterior HPC: r = 0.327, p-unc = 0.045; left anterior HPC: r = 0.427, p-unc = 0.007; right posterior HPC: r = 0.362, p-unc = 0.025; left posterior HPC: r = 0.360, p-unc = 0.026). A series of four partial correlational analyses demonstrated that left thalamic volume correlated with remote autobiographical memory scores over and above the volume of each of those four manually delineated HPC portions (control variable: right anterior HPC: r = 0.489, p = 0.006; left anterior HPC: r = 0.482, p = 0.007; right posterior HPC: r = 0.505; p = 0.004; left posterior HPC: r = 0.509, p = 0.004); aHPC: anterior hippocampus; HPC: hippocampus; L: left hemisphere; MAP: Memory and Amnesia Project; OPTIMA: Oxford Project To Investigate Memory and Aging; pHPC: posterior hippocampus; p-unc: p values of bivariate correlations are presented at uncorrected levels for display purposes; R: right hemisphere; rho: Spearmann’s rank correlation coefficient; r: Pearson’s correlation coefficient; TIV: total intracranial volume; z: average of age-scaled standardized scores on neuropsychological tests of episodic memory; z-res: volumes for each HPC portion are residualized against age, sex, TIV, and scan source (MAP, OPTIMA) across participants. (F) ‘Impaired’ vs. ‘Unimpaired’ patients on visual forgetting: Structural/Functional abnormalities. Given that 17 of our patients reached ceiling scores in visual forgetting, we also dichotomized our patient group into two subgroups, those that attained ceiling scores (z=0.33), and those with lower scores (z<0.33). We therefore compared the two patient subgroups across the 13 structural and functional abnormalities identified above at whole-group level. Consistent with our correlational approach, the two patient subgroups differed only with respect to the volume of the right HPC, as expressed in manual volumetry (t = -3.32, p-corr = 0.027) and in the right HPC VBM cluster (t = -4.05, p-corr = 0.004; rest of ps, p-corr ≥ 0.143); We then iterated these comparisons with a series of one-way ANCOVAs, including patients’ HADS scores for depression as a covariate of no interest. The results retained their significance (manually delineated right HPC volume: F = 9.96; p-corr = 0.048; right HPC VBM cluster: F = 14.94, p-corr = 0.007; rest of ps, p-corr ≥ 0.232); p-corr: significance values are corrected for multiple testing using the Holm-Bonferroni sequential method (Holm, 1979); t/Wt: comparison between the two subgroups across the 13 structural/functional abnormalities identified at group level for patients; F: these comparisons were iterated in the form of a series of univariate ANCOVAs, including patients’ scores for depression (HADS) as between-subjects covariates; z-res: volumes are residualized against age, sex, scan source (MAP, OPTIMA), and TIV; functional abnormalities are residualized against age and sex; t: Student t-test; Wt: Welch t-test; SD: standard deviation; bold: p-corr < 0.05; HADS: Hospital Anxiety and Depression Scale (Zigmond and Snaith, 1983); ANCOVA: analysis of covariance; OPTIMA: Oxford Project To Investigate Memory and Aging; MAP: Memory and Amnesia Project. (G) Comparison of treated vs. untreated patients on memory scores and structural/functional brain abnormalities. We investigated whether patients that had been treated with immunosuppressive therapy (n=31) showed less pronounced memory impairment and brain abnormalities as compared with those that had not received treatment (n=7), in order to ensure that the structure/function-behavior relationships disclosed across all 38 patients were not driven by the subgroup of patients that had not received such treatment. Patients who had not received such treatment scored lower than the rest of the patients on anterograde retrieval, but not on visual forgetting, or remote autobiographical memory. Nevertheless, the two patient subgroups did not differ with respect to the extent of structural or functional brain abnormalities that we observed above for the entire group of patients (all ps, p-corr ≥ 0.832), even at uncorrected levels (all ps, p-unc > 0.06); p-corr: Holm-Bonferroni (Holm, 1979) correction applied separately for the number of memory scores (n=3) and the number of structural/functional abnormalities (n=13); bold: p-corr < 0.05; z: average of age-scaled standardized scores on neuropsychological tests of episodic memory; z-res: volumes are residualized against age, sex, scan source (MAP, OPTIMA), and TIV; functional abnormalities are residualized against age and sex. (H) Correlation of anterograde retrieval scores with structural / functional abnormalities across patients that had received immunosuppressive therapy. Given the more pronounced impairment on anterograde retrieval of patients that had not received immunosuppressive therapy, we sought to determine whether the relationship of this composite score with reduced rsALFF in the PCC held when the analysis was confined to the 31 patients that had received immunosuppressive therapy. Indeed, the relationship between reduced rsALFF in the PCC and impaired anterograde retrieval retained its significance across this patient subgroup; L, R: left, right (hemisphere); HPC: hippocampus; MPFC: medial prefrontal cortex; p-corr: p values are adjusted for multiple testing using the Holm-Bonferroni sequential method; bold: p-corr < 0.05; rsALFF: resting-state amplitude of low frequency fluctuations; rsFC: resting-state functional connectivity; PrCu: precuneus; PCC: posterior cingulate cortex; z-res: volumes are residualized against age, sex, scan source (MAP, OPTIMA), and TIV; functional abnormalities are residualized against age and sex. (I) Participants - data availability. Numbers of healthy controls and patients that underwent structural, functional MRI, and neuropsychological assessment. *: While the healthy controls whose structural MRI datasets were added from the OPTIMA project had not been assessed with our laboratory’s neuropsychological battery, they had been assessed with tests measuring overall cognitive impairment [Mini-Mental State Examination – MMSE; (Folstein et al., 1975)]. Expectedly, these scores indicated that none of those healthy controls had any apparent cognitive impairment [mean = 29.74, SD = 0.56, min = 28, well above widely accepted cut-offs, for example (Aevarsson and Skoog, 2000; Di Carlo et al., 2002)]. OPTIMA: Oxford Project To Investigate Memory and Aging; MAP: Memory and Amnesia Project; (rsf)MRI: (resting-state functional) Magnetic Resonance Imaging; n: number of participants.