2 July 2019�Biggs Journal Club, UT Health, San Antonio

Dr. Habil Zare

PI of the Oncinfo Lab

​

2

Design overview

​

48 *

80K single-nucleus transcriptomes

from the prefrontal cortex

(~1700 nuclei per brain)

3

Design details

​

48 *

80K single-nucleus transcriptomes

from the prefrontal cortex (BA10)

(~1700 nuclei per brain)

ROSMAP

24 AD-pathology +

24 no-pathology’

4

Clustering thousands of cells

80K cells

20 clusters

8 cell types

Marker genes

5

6 major cell types

​

6

Differential gene expression

AD-pathology vs. no pathology

​

1. Excitatory neurons have the largest number of deregulated genes (565/1031).

​

• Most deregulated genes in neurons are downregulated.

7

Validation using q-PCR

• 5/6 genes were confirmed.
• Genes with a more homogeneous response, such as LINGO1 (?), might have more potential for intervention [assuming gene expression deregulations in a cell type are not reactive].

8

Comparison with bulk RNA-Seq

• scRNA-Seq identifies many more deregulated genes than bulk RNA-Seq, which is dominated by expression signals of neurons and oligodendrocytes.
• The direction of deregulation agrees with bulk RNA-Seq.

9

Comparison with bulk RNA-Seq

Bulk data are dominated by excitatory neurons and oligodendrocytes

�=> Changes in microglia are underrepresented

Rank permutation test

10

DEGs with opposite directionality in differencell types

APOE is overexpressed in microglia and downregulated in astrocytes.

11

Progression of AD pathology based on clinico-pathological variables

48 individuals were clustered based on 7 clinico-pathological variables

12

Upregulation across cell types

• Downregulated genes are mostly cell type specific.
• Only early- vs. late-pathology upregulated genes are consistent across cell types.�=> Variation in gene expression are largely cell-type specific at an early stage.

13

Upregulation across cell types

11 genes are upregulated in at least 5 cell types. They “involve in the proteostasis network—the molecular machinery that operates to maintain protein integrity.”

14

Correlation with traits

- Spearman rank correlation values were visualized using a self-organizing map (SOM).

- Genes with similar correlation patterns are grouped close together.

15

Correlation with traits

Correlation patterns differ across cell types.

16

Post-mortem interval and age at death have distinct correlation patterns.

17

Genes with similar correlation pattern

across cell types

Example: M7�- Positive pathological correlation in microglia

- Enriched in immune and inflammatory pathways, as well as in pathways associated with the clearance of β-amyloid

18

Expression and mutation

• M6 and M7 overlapped with genes that are AD risk factors, including APOE, TREM2, MEF2C, PICALM, etc.
• M3 overlapped with genes that are associated with cognition and the expression of which is negatively correlated with pathology in neuron

19

Expression and mutation

• M6 and M7 overlapped with genes that are AD risk factors, including APOE, TREM2, MEF2C, PICALM, etc.
• M3 overlapped with genes that are associated with cognition and the expression of which is negatively correlated with pathology in neuron

Can these modules be used to increase power to detect more AD genes?

20

Excitatory neurons subpopulations

• 13 excitatory neurons subpopulations were identified.
• The ratio of cells in each subpopulation varies across individuals with different amyloid levels.

Excitatory neurons 6

​

Excitatory neurons 2

​

Excitatory neurons 4

​

21

Subpopulation marker genes

22

• 40 subpopulations were identified.�
• Subpopulation ratios depends on pathology

​

​

23

Sex-specific differential response to AD pathology

​

24

Expression of marker genes

across genders

Excitatory neurons subpopulation #4 have higher levels of expression in females.

=> Transcriptional response to AD pathology may be sex-specific.

25

Gene expression correlation across genders

The distribution of correlation values computed between gene-expression profiles (n = 17,926 genes) in oligodendrocytes averaged over individuals.

Generally positive correlation in males, but not as much in females.

26

Unlike males, WMH is associated with cognition in females.

=> Reduced transcriptional response, particularly in oligodendrocytes, in females.

White-matter hyperintensities variation across genders

27

Gene expression levels in neurons generally anticorrelate with pathology in females, but not as much in males.

Gene expression correlation across genders

28

Supplementary

​

29

Summary

​

• The strongest disease-associated changes appeared early in pathological progression and were highly cell-type specific, whereas genes upregulated at late stages were common across cell types and primarily involved in the global stress response.
• Myelination-related processes were recurrently perturbed in multiple cell types
• Transcriptional responses were substantially different between sexes.