Dendritic spine head diameter predicts episodic memory performance in older adults

Abstract

Episodic memory in older adults is varied and perceived to rely on numbers of synapses or dendritic spines. We analyzed 2157 neurons among 128 older individuals from the Religious Orders Study and Rush Memory and Aging Project. Analysis of 55,521 individual dendritic spines by least absolute shrinkage and selection operator regression and nested model cross-validation revealed that the dendritic spine head diameter in the temporal cortex, but not the premotor cortex, improved the prediction of episodic memory performance in models containing β amyloid plaque scores, neurofibrillary tangle pathology, and sex. These findings support the emerging hypothesis that, in the temporal cortex, synapse strength is more critical than quantity for memory in old age.

Fig. 1.. Overview of workflow.

Sections (250 μm) of the BA37 temporal cortex and BA6 premotor cortex from 128 older adults were Golgi stained and imaged with high-resolution bright-field microscopy. Z-stacks were imported to Neurolucida 360 for 3D digital reconstruction. Representative 60X bright-field images of Golgi-stained dendrites in BA37 and BA6 of cognitively normal older adults are shown on the right with the digital 3D reconstructions of those segments. Scale bars, 10 μm. Blue, thin spines; orange, stubby spines; green, mushroom spines; yellow, filopodia. Three cases were removed from downstream analyses due to missing data, resulting in a final sample size of 125. BA37 and BA6 datasets were used separately. Each dataset was randomly split into a discovery set containing 63 participants and a validation set containing 62 participants. LASSO regression was performed on the discovery set to identify spine traits associated with episodic memory. K-fold cross validation (K = 10) was used to select the penalty factor used for coefficient shrinkage during LASSO regression. Nested model leave-one-out cross-validation was then performed using the validation set. Multiple linear models containing different combinations of variables (age, pathology scores, sex, and spine density, head diameter, length, and volume) were generated. Pathology scores included were NP score and NFT score. The model with the smallest MSE best predicted episodic memory.

Fig. 2.. Dendritic spine head diameter improves the model prediction of episodic memory scores.

(A) LASSO coefficients after shrinkage for each BA37 dendritic spine feature included in the regression. Of the four spine features, head diameter had the strongest association with episodic memory. (B) Density plot comparing the observed episodic memory scores with those predicted by the full model consisting of pathology scores, sex, age, and BA37 dendritic spine density, length, head diameter, and volume. (C) Density plot comparing the observed episodic memory scores with those predicted by a model containing all features except age. Removing age improves model prediction. (D) Density plot comparing the observed episodic memory scores with those predicted by a model containing pathology scores (path), sex, and BA37 dendritic spine head diameter (hd). This model provided the best prediction of episodic memory scores. (E) Pearson correlation between the observed episodic memory scores and those predicted by the best model (pathology, sex, and BA37 head diameter). Cog, cognitive. (F) Density plot comparing the observed episodic memory scores with those predicted by the model containing pathology scores and sex. (G) LASSO coefficients after shrinkage for each BA6 dendritic spine feature included in the regression. Of the four spine features, BA6 spine length had the strongest association with episodic memory. (H) Density plot comparing episodic memory scores with the full model, consisting of pathology scores, sex, age, and BA6 dendritic spine density, length, head diameter, and volume. (I) Density plot comparing episodic memory scores with a model containing all features except age. Removing age improves model prediction. (J) Density plot comparing episodic memory scores with a model containing pathology scores, sex, and BA6 dendritic spine length (sl). The model fit (R²) is slightly lower than the model in (F), which contains only pathology scores and sex as predictors.

Fig. 3.. Dendritic spine morphology correlates with cognition and pathology in BA37 but not BA6.

(A) The heatmap depicts Spearman correlations between BA37 dendritic spine measurements and both cognitive test scores and pathology scores. An FDR of 10% (q < 0.1) was applied, and the unadjusted P value is shown for those correlations with P < 0.05 and q < 0.1. Correlations with P < 0.05 and q < 0.05 following FDR adjustment are indicated with double asterisks (**). (B) Representative Spearman correlation (corr) scatter plots for BA37. (C) Spearman correlations between BA6 dendritic spine measurements and both cognitive test scores and pathology scores are shown in the heatmap. No correlations were statistically significant (P < 0.05) following FDR adjustment (q < 0.1).