ADRC_FC
Julie Wisch
September 26, 2018
INTRODUCTION OUTLINE
*Alzheimer’s disease is characterized by the accumulation of amyloid plaques in the brain (Selkoe & Hardy, 2016; Brier et al, 2016).
*Genetic differences (apoe4) have been associated with increased risk of Alzheimer’s disease (Tiraboschi et al, 2004); however, a substantial body of literature exists that suggests apoe4’s role in Alzheimer’s disease may be independent of amyloid accumulation (Huynh et al, 2017).
*Differences in functional connectivity have been previously observed in individuals diagnosed with Alzheimer’s disease and mild cognitive impairment (e.g. Johnson et al, 2012; Chen et al, 2011; Lustig et al, 2003).
*Connections between functional connecitivity and apoe4 status have been considered, but results are mixed (Trachtenberg et al, 2012).
*The goal of this paper is to consider longitudinal changes in functional connectivity with respect to amyloid accumulation, apoe4 status, and cognitive impairment.
BACKGROUND
*AD is bad because it affects lots of people…
*Explain CDR
*Explain apoe4
*Explain PET imaging
*Explain BOLD imaging
METHODS
*CDR diagnosis
*Imaging and processing
*Define amyloid “abnormal”
*Explain how we assigned “years” to non-converters (cite Cathy)
ANALYSIS
*Chi-squared tests were performed to test for an association between CDR status, amyloid status, and apoe4 status.
*PCA was used to identify a cortical signature for the functional imaging results (cite Robert). This cortical signature was plotted over time as a function of both CDR status and apoe4 status to look for patterns in functional connectivity changes over time.
*We used multivariate analysis of variance (MANOVA) followed by analysis of variance (ANOVA) to look for differences in the functional connecitivity principal components between individuals who convert to CDR > 0 as compared to those who remain CDR = 0 throughout their participation in the study, as well as to compare individuals with abnormal amyloid levels and those with normal levels.
*We also used ANOVA followed by Tukey’s range test to look for differences in individual network trajectories over time between converters and non-converters as well as individuals with abnormal amyloid levels and those with normal levels.
RESULTS AND DISCUSSION
Amyloid has been previously linked to cognitive status as well as apoe4 status. Chi-squared tests showed significant relationships between conversion and amyloid level and apoe4 status and amyloid level, but not apoe4 status and conversion.
## [1] "CDR status (0 = never converts, 1 = converts) and Amyloid level (0 = normal, 1 = abnormal):"##
## Pearson's Chi-squared test with Yates' continuity correction
##
## data: CDRandSUVR
## X-squared = 23.559, df = 1, p-value = 1.211e-06## [1] "apoe4 status (0 = no apoe4 genes, 1 = at least 1 apoe4 gene) and Amyloid level (0 = normal, 1 = abnormal):"##
## Pearson's Chi-squared test with Yates' continuity correction
##
## data: apoe4andSUVR
## X-squared = 71.632, df = 1, p-value < 2.2e-16## [1] "apoe4 statatus and CDR status at the end of the study"##
## Pearson's Chi-squared test with Yates' continuity correction
##
## data: apoe4andCDR
## X-squared = 0.00022647, df = 1, p-value = 0.988We performed a PCA to reduce dimensionality of the network functional connectivities, and used these components to look for differences in connectivity for converters vs. non-converters (p = 6.65e-04), amyloid positive vs. amyloid negative (p = .034), and apoe4+ vs. apoe4- individuals (p = 0.78).
## [1] "MANOVA and Logit Regression for cdr status:"## Df Pillai approx F num Df den Df Pr(>F)
## pcaOutput2$groups_cdr 1 0.049765 3.463 8 529 0.0006646 ***
## Residuals 536
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1##
## Call:
## glm(formula = pcaOutput2$groups_cdr ~ pcaOutput2$PC1 + pcaOutput2$PC2 +
## pcaOutput2$PC3 + pcaOutput2$PC4 + pcaOutput2$PC5 + pcaOutput2$PC6 +
## pcaOutput2$PC7 + pcaOutput2$PC8, family = "binomial")
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.3042 -0.5970 -0.4873 -0.3491 2.5979
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.906952 0.138471 -13.771 < 2e-16 ***
## pcaOutput2$PC1 -0.089468 0.035586 -2.514 0.01193 *
## pcaOutput2$PC2 0.085052 0.045666 1.862 0.06254 .
## pcaOutput2$PC3 0.157541 0.055071 2.861 0.00423 **
## pcaOutput2$PC4 0.054175 0.060330 0.898 0.36920
## pcaOutput2$PC5 -0.190930 0.066953 -2.852 0.00435 **
## pcaOutput2$PC6 0.079593 0.068772 1.157 0.24713
## pcaOutput2$PC7 -0.048751 0.070378 -0.693 0.48850
## pcaOutput2$PC8 -0.002381 0.081235 -0.029 0.97662
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 452.40 on 537 degrees of freedom
## Residual deviance: 423.62 on 529 degrees of freedom
## AIC: 441.62
##
## Number of Fisher Scoring iterations: 5## [1] "MANOVA and Logit Regression for amyloid status:"## Df Pillai approx F num Df den Df Pr(>F)
## pcaOutput2$groups_SUVR 1 0.030816 2.1025 8 529 0.03395 *
## Residuals 536
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1##
## Call:
## glm(formula = pcaOutput2$groups_SUVR ~ pcaOutput2$PC1 + pcaOutput2$PC2 +
## pcaOutput2$PC3 + pcaOutput2$PC4 + pcaOutput2$PC5 + pcaOutput2$PC6 +
## pcaOutput2$PC7 + pcaOutput2$PC8, family = "binomial")
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.0821 -0.7634 -0.6543 -0.4073 2.2407
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.250182 0.107113 -11.672 <2e-16 ***
## pcaOutput2$PC1 -0.063366 0.028281 -2.241 0.0251 *
## pcaOutput2$PC2 -0.058850 0.035798 -1.644 0.1002
## pcaOutput2$PC3 -0.001716 0.045040 -0.038 0.9696
## pcaOutput2$PC4 0.093500 0.049117 1.904 0.0570 .
## pcaOutput2$PC5 -0.051433 0.053079 -0.969 0.3326
## pcaOutput2$PC6 0.022333 0.055308 0.404 0.6864
## pcaOutput2$PC7 -0.134635 0.058785 -2.290 0.0220 *
## pcaOutput2$PC8 -0.061014 0.065387 -0.933 0.3508
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 583.29 on 537 degrees of freedom
## Residual deviance: 565.65 on 529 degrees of freedom
## AIC: 583.65
##
## Number of Fisher Scoring iterations: 4## [1] "MANOVA and Logit Regression for APOE status:"## Df Pillai approx F num Df den Df Pr(>F)
## pcaOutput2$groups_apoe 1 0.008985 0.59952 8 529 0.7786
## Residuals 536##
## Call:
## glm(formula = pcaOutput2$groups_apoe ~ pcaOutput2$PC1 + pcaOutput2$PC2 +
## pcaOutput2$PC3 + pcaOutput2$PC4 + pcaOutput2$PC5 + pcaOutput2$PC6 +
## pcaOutput2$PC7 + pcaOutput2$PC8, family = "binomial")
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.1224 -0.8809 -0.8170 1.4176 1.8112
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.788472 0.093553 -8.428 <2e-16 ***
## pcaOutput2$PC1 -0.011073 0.024468 -0.453 0.651
## pcaOutput2$PC2 0.006485 0.030912 0.210 0.834
## pcaOutput2$PC3 -0.046765 0.040680 -1.150 0.250
## pcaOutput2$PC4 -0.046972 0.043635 -1.076 0.282
## pcaOutput2$PC5 0.007767 0.046458 0.167 0.867
## pcaOutput2$PC6 0.019406 0.048533 0.400 0.689
## pcaOutput2$PC7 -0.070548 0.052064 -1.355 0.175
## pcaOutput2$PC8 -0.009030 0.057609 -0.157 0.875
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 669.66 on 537 degrees of freedom
## Residual deviance: 664.84 on 529 degrees of freedom
## AIC: 682.84
##
## Number of Fisher Scoring iterations: 4Components 1, 3, and 5 contributed significantly to differentiating between CDR converters and non converters, and Components 1 and 7 contributed significantly to differentiating between amyloid abnormal and amyloid normal individuals. There was no difference in BOLD imaging by APOE status. The networks that comprise these components are visualized in Fig. 2 (this seems like it would actually go in the supplemental materials). Sensory motor networks are most heavily emphasized in the components that are used to classify between converters and non-converters, as well as between those with increased amyloid uptake.
Further analysis by ANOVA and Tukey’s HSD of individual network trajectories over time demonstrated that significant differences exist between converters and non-converters in six networks: SMlat x SMlat, SMlat x VAN, CO x CO, MEM x MEM, FP x SAL, SUBCort x SUBCort. Differences do not exist by APOE status in these networks, except for in the SUBCort x SUBCort network.
## Region pValue_cdr pValue_APOE
## 14 zSM_lat_x_SM_lat 0.012752490 0.3707278
## 23 zSM_lat_x_VAN 0.006240083 0.6549784
## 26 zCO_x_CO 0.017665955 0.4622179
## 72 zFP_x_SAL 0.032585090 0.4971116## Region pValue_cdr pValue_APOE
## 11 zSM_x_VAN 0.15037705 0.01129553
## 67 zVIS_x_SUBCort 0.51040225 0.02874788
## 71 zFP_x_FP 0.17876435 0.04106525
## 82 zSUBCort_x_SUBCort 0.06412174 0.03055952
## 85 zSUBCort_x_CEREB 0.27607401 0.01097768
## 86 zVAN_x_VAN 0.71851455 0.02983418
For the subset of individuals who had PIB-PET imaging, we compared network trajectories as a function of amyloid accumulation status. Differences in network trajectory were found for the following regions: DMN x FP, AUD x DAN, DAN x DAN, and DMN x DAN. Amyloid accumulation appears to affect network connectivity in the default mode network and default attention network, two networks frequently associated with changes in the progression in AD (citation); however, none of these regions differentiated between converters and non-converters.
## NAMES[12:102] CDRpValue SUVRpValue APOEpValue
## 45 zAUD_x_DAN 0.2184562 0.017633701 0.03471848
## 50 zDMN_x_FP 0.4430663 0.003610025 0.49690645
## 54 zDMN_x_DAN 0.9972030 0.043803435 0.94395730
## 89 zDAN_x_DAN 0.4679378 0.025504998 0.54361230In order to validate the random “conversion date” assignment for non-converters, we used linear regression to compare the relationship of each significant network connection and time. A Fisher’s r to z comparison indicated that there was a significantly weaker temporal correlation for non-converters than converters for each network, indicating that non-converter network activity was not correlated with time, supporting our position that random assignment of converstion date to non-converters would have no impact on results.
## Call: paired.r(xy = m.corr[1, 2], xz = m.corr[2, 2], n = 663, n2 = 120)
## [1] "test of difference between two independent correlations"
## z = 2.51 With probability = 0.01## Call: paired.r(xy = m.corr[1, 2], xz = m.corr[2, 2], n = 663, n2 = 120)
## [1] "test of difference between two independent correlations"
## z = 2.76 With probability = 0.01## Call: paired.r(xy = m.corr[1, 2], xz = m.corr[2, 2], n = 663, n2 = 120)
## [1] "test of difference between two independent correlations"
## z = 2.44 With probability = 0.01## Call: paired.r(xy = m.corr[1, 2], xz = m.corr[2, 2], n = 663, n2 = 120)
## [1] "test of difference between two independent correlations"
## z = 2.06 With probability = 0.04CONCLUSIONS
*Differences in functional connectivity can be seen in indiviudals who convert to cdr > 0 as opposed to those who do not.
*These differences appear to be independent of APOE status.
*These changes occur in diffuse networks, but appear to disproportionately affect sensory-motor network connections and intranetwork connections.
REFERENCES Brier, M. R., Gordon, B., Friedrichsen, K., McCarthy, J., Stern, A., Christensen, J., … & Cairns, N. J. (2016). Tau and Ab Imaging, CSF Measures, and Cognition in Alzheimer’s Disease. Science Translational Medicine, 8, 338. http://stm.sciencemag.org/content/scitransmed/8/338/338ra66.full.pdf
Chen, G., Ward, B. D., Xie, C., Li, W., Wu, Z., Jones, J. L., … & Li, S. J. (2011). Classification of Alzheimer disease, mild cognitive impairment, and normal cognitive status with large-scale network analysis based on resting-state functional MR imaging. Radiology, 259(1), 213-221. https://pubs.rsna.org/doi/pdf/10.1148/radiol.10100734
Huynh, T. P. V., Davis, A. A., Ulrich, J. D., & Holtzman, D. M. (2017). Apolipoprotein E and Alzheimer Disease: The influence of apoe4 on amyloid-beta and other amyloidogenic proteins. Journal of Lipid Research, jlr-R075481. http://www.jlr.org/content/early/2017/02/27/jlr.R075481.full.pdf
Johnson, K. A., Fox, N. C., Sperling, R. A., & Klunk, W. E. (2012). Brain imaging in Alzheimer disease. Cold Spring Harbor perspectives in medicine, a006213. http://perspectivesinmedicine.cshlp.org/content/early/2012/01/31/cshperspect.a006213.full.pdf
Lustig, C., Snyder, A. Z., Bhakta, M., O’Brien, K. C., McAvoy, M., Raichle, M. E., … & Buckner, R. L. (2003). Functional deactivations: change with age and dementia of the Alzheimer type. Proceedings of the National Academy of Sciences, 100(24), 14504-14509. http://www.pnas.org/content/pnas/100/24/14504.full.pdf
Selkoe, D. J., & Hardy, J. (2016). The amyloid hypothesis of Alzheimer’s disease at 25 years. EMBO molecular medicine, 8(6), 595-608. http://embomolmed.embopress.org/content/8/6/595
Tiraboschi, P., Hansen, L. A., Masliah, E., Alford, M., Thal, L. J., & Corey-Bloom, J. (2004). Impact of apoe4 genotype on neuropathologic and neurochemical markers of Alzheimer disease. Neurology, 62(11), 1977-1983. http://n.neurology.org/content/neurology/62/11/1977.full.pdf
Trachtenberg, A. J., Filippini, N., & Mackay, C. E. (2012). The effects of apoe4-??4 on the BOLD response. Neurobiology of aging, 33(2), 323-334. https://www.sciencedirect.com/science/article/pii/S0197458010001284