Mitochondrial genes in single cell nuclear RNAseq data

Question

[dalhoomist]

I have done several single cell nuclear RNAseq experiments on human brain tissue. I followed this protocol outlined in Krishnaswami et al https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4941947/. The percentage of viable (intact) cells out of the total nuclei suspension is less than 5% - and these cells can be excluded in the analysis. The nuclear specific transcripts are enriched (for example MALAT-1). However, some mitochondrial specific genes like MT-ND1, MT-ND2, MT-ND3, MT-CO1, MT-CO2, and MT-CO3 are still detected at high levels (And often differentially expressed between clusters). Given that the RNA sequenced is nuclear RNA, how can I explain the presence of mitochondria-encoded RNA in my data?

[GenoMax]

Since the explanation in this case is likely going to be related to experimental protocol/technique, you may want to post this at SeqAnswers and biology SE.

Informatically you could ignore the counts on mito genes if they are not of interest. But their presence seems to indicate that your nuclear preps were likely not clean or perhaps mitochondrial RNA's get transferred to nuclei? It also looks like mitochondrial genes are found in nuclei so that may also be another explanation?

[raths.f]

Hi, I am facing the same question in a single nuclei dataset. The same genes that you are mentioning and they are also differential between my treatment groups. Did you ever get an idea on what is going on here?

[jared.andrews07]

Variable mitochondrial genes are pretty typical, and as genomax said, this could be indicative of contamination (either through intact cells or ambient RNA). What % of reads are in mitochondrial genes? If it's only a few percent, you can likely ignore them.

Answer 1

[ldaveyl]

I've read in some papers that use single nucleus datasets that mitochondrial genes are ignored for downstream analysis.

https://www.nature.com/articles/s41586-019-1404-z#Sec8

"Mitochondrial RNA genes were filtered out as well to exclude transcripts originating from outside the nucleus and avoid biases introduced by the isolation of nuclei and ultracentrifugation."

Answer 2

[Prakki Rama]

Quoting from 10X Genomics documentation page:

Even in 10x assays with single nuclei as input (for example, Multiome assay), an elevated level of mitochondrial genes would mean that mitochondrial RNA could remain “stuck” to nuclear membranes or may get partitioned into GEMs and be detected by the gene expression biochemistry.