Introduction

Several human ocular tissues are collected post-mortem and used for either transplants or biomedical research. Specifically, human eyes collected from deceased donors are often used for corneal transplants1. Donor eyes are also used for studies of the causes and progression of retinal diseases, such as age-related macular degeneration (AMD), which affects almost 11 million people in the US alone2. Gene expression changes are used as disease biomarkers and for the identification of molecular mechanisms controlling disease progression. It is critically important to distinguish such disease-related changes from those that result from differences in post-mortem interval (PMI), which is also referred to as death to preservation time. In the case of studies utilizing RNA from post-mortem samples PMI represents the time window from death to stabilization of RNA, either by cold storage or ideally by chemical RNA stabilization reagents such as RNAlater followed by immediate use or cold storage. Data obtained through personal communication with the Eye Back Association of America (EBAA) details the death-to-cooling and death-to-preservation times from 3,567 eye donors intended for corneal transplants. As death-to-cooling and PMI times of tissue collection vary (Fig. 1)3,4, along with the PMI of tissues used in research, a clear understanding of gene expression changes in the key cell types in the post-mortem eye is much needed and cannot be met by studies that use GTEx samples because GTEx does not collect retina or RPE and no other comprehensive post-mortem retina or RPE database exists5. It is also important to note that PMI times are reported in the literature utilizing post-mortem tissues, while death-to-cooling times are not reported in the literature, allowing for an unaccounted variable in these samples.

Figure 1
figure 1

Ocular tissue donor collection stage times. Death to Cooled (Red), Cooled to Preserved (Blue), and Death to Preserved (Purple) collection stage times from 2019 Eye Bank Association of America data. (A) Violin plot displaying distribution of collection stage times. (B) Stacked bar graph displaying variance of collection stage times. n = 3567.

Full size image

Gene expression is a tightly regulated process controlled by tissue-specific transcription factors, cofactors directing the binding of transcription factors, chromatin accessibility, and by various noncoding RNAs. This process results in the precise combination of transcribed mRNAs and protein production to maintain the physiological function of the cell and adaptation to external stimuli6,7,8, as steady state expression levels of individual genes are a sum of both the rate of de novo transcription and RNA degradation9.

Previous studies have been conducted on the degradation of RNA to determine the half-life of individual RNA species in many organisms and cell lines10,11,3) and RPE/choroid (Fig. 4) for all samples examined compared to the 0 h time point. Genes determined to be DE preserved (relatively upregulated) or DE degraded (relatively downregulated) are highlighted. In both mouse retina and RPE/choroid, we observed changes in RNA expression as early as 15 min post-mortem, with 1,021 genes down-regulated and 1,149 genes upregulated in the mouse retina and 342 genes down-regulated and 277 genes upregulated in the mouse RPE/choroid. As the PMI increased, the number of DE genes also increased, as indicated in Table 1. The largest number of DE genes was observed at 24 h post-mortem in the RPE/choroid. The number of DE genes in retina samples from mice increased over time, with a sharp drop at 24 h. This reduction of DE genes at 24 h is explained by the sharp increase in the BCV, resulting in less total DE genes despite an overall increase in log fold change (LFC) compared to 12 h.

Figure 3
figure 3

Post-mortem analysis of gene expression in mouse retina. Differentially expressed genes in mouse retina post-mortem compared to the zero-hour time point.

Full size image
Figure 4
figure 4

Post-mortem analysis of gene expression in mouse RPE/choroid. Differentially expressed genes in mouse RPE/choroid post-mortem compared to the zero-hour time point.

Full size image
Table 1 Number of mouse DE genes preserved or degraded relative to 0 h.
Full size table

In our baboon studies, we had access to fewer time points including 0 h, 3 h, 6 h and 24 h post-mortem. In these studies, we observed little change in mRNA levels in the baboon retina at 3 h post-mortem, but substantial changes at 6 h and 24 h. Specifically, we observed 2,280 genes showing reduced and 2,689 genes showing increased expression at the 6 h time point, and 4,820 genes showing reduced and 4,243 genes showing increased expression at the 24 h time point. By contrast, we observed mRNA changes in the baboon RPE/choroid as early as 3 h post-mortem, with 4,258 genes showing reduced and 4,457 genes showing increased expression, with a similar number of preserved and degraded genes at the 24 h time point (Fig. 5 and Table 2).

Figure 5
figure 5

Post-mortem analysis of gene expression in baboon retina. Differentially expressed genes in baboon retina and RPE/choroid post-mortem compared to the zero-hour time point.

Full size image
Table 2 Number of baboon DE genes preserved or degraded relative to 0 h.
Full size table

The number of DE genes between all individual time points for mouse and baboon in both retina and RPE/choroid were measured in order to determine the effect of using samples of varying time-points as replicates, as can be the case with rare samples of human diseased tissue. As expected, mice with neighboring PMI time points had the lowest number of DE genes, with multiple neighboring time points having no DE genes, and a consistent and steady increase in the number of DE genes as the PMI differences increased. In baboons, the trend in the number of DE genes was similar to what we observed in the mouse retina; however, the analysis in RPE/choroid showed less consistent trends. Additionally, any genes of interest could be checked against this data set to determine the potential masking or exaggerating effect of PMI differences between replicates or experimental conditions (Supplemental Fig. 4).

A subset of mouse retina and RPE candidate genes were selected (see Methods) to determine the ability to predict PMI of samples using a qRT-PCR assay. Four additional mice for each of the 0, 6 hr, 12 hr, 18 hr, and 24 hr PMI time points had retina and RPE samples collected and subjected to qRT-PCR for the 17 retina and 18 RPE candidate genes. (Supplemental Fig. 5, Supplemental Tables 3–5). Multiple linear regression models were generated for both retina and RPE to estimate the PMI. These data (Supplemental Table 6) suggest that RNA metabolism is tissue-specific and has a unique genetic signature correlated to the PMI of a sample.

Analysis of preserved and degraded genes by species, tissue, and RNA class

Timeline plots of differentially expressed genes were generated for both mouse and baboon retina and RPE/choroid tissue, with preserved and degraded genes presented separately (Fig. 6). Similarly, timeline plots were created displaying only lncRNAs, pseudogenes, and mRNAs (Supplemental Figs 6–8). Additionally, the number of expressed genes, preserved genes, and degraded genes for each biotype was calculated (Supplemental Tables 1, 2). The results show an enrichment of lncRNAs preserved in the mouse RPE/choroid (55.4% of expressed genes) and a reduction of lncRNAs being degraded (17.9% of expressed genes). By contrast, lncRNAs in the mouse retina contributed only 2.6% of degraded RNAs and 1.8% of preserved RNAs of total expressed genes respectively. Additionally, increased percentages of pseudogenes in RPE/choroid are preserved relative to degraded for pseudogenes, providing evidence of tissue specific RNA regulation for these classes of RNA (Fig. 7).

Figure 6
figure 6

Z-Scores of preserved and degraded differentially expressed genes.

Full size image
Figure 7
figure 7

DE gene analysis of 24 PMI samples. Number of genes expressed (E) (CPM ≥1), preserved (P) (q ≤0.05) and degraded (D) (q ≤0.05) in mouse and baboon retina and RPE/choroid, presented by gene biotype.

Full size image

Since it has been shown that lncRNAs have regulatory roles33,34,35, further analysis of lncRNAs in baboon versus mouse retina and RPE/choroid demonstrates both tissue- and species-specific differences in their PMI metabolism. While lncRNAs in the mouse RPE/choroid showed a global shift towards preservation, lncRNAs in baboon RPE/choroid showed a global shift towards degradation. In the mouse retina the overall number of lncRNA DE genes was reduced and showed a minor global shift towards preservation, while lncRNA in the baboon retina showed a global shift towards preservation. These data provide evidence for both tissue- and species-specific post-mortem regulation of lncRNA metabolism at a global level.

To further measure the effect of species- and tissue-type on post-mortem RNA metabolism, GO term enrichment was performed on preserved and degraded genes. GO term overlap between mouse and baboon was determined for both retina and RPE/choroid for the intervals of 0 h to 6 h PMI and 0 h to 24 h PMI (Supplemental Tables 7,8). Additionally, the overlap of preserved and degraded genes between the retina and RPE/choroid was determined for both mouse and baboon for both time comparisons (Tables 3, 4). All statically unaffected, preserved, and degraded genes were converted to their human homologs for the cross-species and cross-tissue comparisons to be conducted (see Methods). While retina and RPE/choroid are believed to have similar biological roles in both the mouse and baboon species, this analysis demonstrates species-specific RNA metabolism at the gene level, which contributes to the levels of functional proteins, evidenced by only a mild overlap of DE genes between species. Surprisingly, when performing intra-species analysis on the overlap of degraded and preserved genes between retina and RPE/choroid there is a 2.3 fold increase in the number of DE genes relative to an identical intra-tissue analysis, providing evidence that species-specific RNA metabolism plays a stronger role than tissue-specific RNA metabolism.

Table 3 Number of DE gene overlaps by species.
Full size table
Table 4 Number of DE gene overlaps by tissue.
Full size table

Gene Ontology (GO) term analysis of preserved and degraded genes

These RNA-seq data sets of PMI require that preserved and degraded genes are independently analyzed for GO term enrichment, as opposed to traditional RNA-seq experiments which typically requires all DE to be analyzed together. In order to determine the set of genes preserved across both species and tissue types, all genes were first converted to their human homologs and genes common to all gene sets for either preserved or degraded genes were subjected to GO term enrichment using DAVID 6.8 (Supplemental Table 9). The disadvantage in converting genes to human homologs is the loss of genes without clear homologs and the ambiguity when a gene converts to multiple homologs. For these analyses genes without homologs were removed from the gene list, and genes with multiple homologs had the one with the lowest ensembl gene id number retained. 89% and 93% of mouse and baboon genes, respectively, mapped to human homologs, and were used for downstream inter- and intra-species analyses where noted. While the loss of partial information is undesirable, it allowed for cross-species GO term analysis and gene overlap analysis.

For degraded genes, terms relating to transcriptional regulation were enriched, i.e. transcription-DNA-templated, post-transcriptional regulation of gene expression, nucleus, transferase activity, DNA binding, and protein domain specific binding. In contrast, preserved genes reflected constitutively active cellular processes, i.e. intracellular signal transduction, cellular response to hormone stimulus, response to cAMP, response to mechanical stimulus, extracellular matrix, and membrane. These data demonstrate that RNA regulating different biological processes has distinct stability profiles.

Discussion

Next generation RNA sequencing (RNA-Seq) technology has enabled the characterization of gene expression in numerous human tissues and cell types, which has in turn led to better classifications of cell types within complex tissues and their functions36,37. It has also enabled the identification of gene expression differences between healthy and diseased tissues38. However, one major challenge with gene expression studies of human tissues is that many investigations use tissues that were collected post-mortem23,24,25,26,27,28,29,30,31,39 and thus, how these findings reflect RNA expression levels of living tissues remains to be characterized40. A recent study using tissues obtained through GTEx (Genotype-Tissue Expression) analyzed the RNA expression of 41 human tissues post-mortem and provides access to tissue samples from donors with PMIs ranging from immediately after death up to 24 h post-mortem41. However, the authors concluded that human tissue sample collection PMI time and corresponding degradation level are dependent on the collection site. These differences in collection times are likely contributing to differences in gene expression among tissues analyzed, making it challenging to extrapolate these findings to healthy living tissues. Thus a clear understanding of how gene expression changes post-mortem in controlled experimental settings would aid in the design and interpretation of future experiments utilizing human post-mortem samples.

One primary purpose for studying diseased tissue by RNA-Seq analysis is to directly compare gene expression in altered phenotype samples to healthy samples while minimizing as many variables as possible. As human samples for some diseased tissues are difficult to obtain, controlling for the PMI is often not possible and samples from different PMIs are treated as biological replicates for the same condition in the literature26,42. Thus, there is an urgent need to determine the degree of RNA degradation post-mortem in a controlled setting. Since it is not feasible to do these studies in humans, we utilized two mammalian species–mouse and baboon–that are widely used in biomedical research of the eye.

Although they are neighboring tissues, retina and RPE/choroid have different biological roles in which they function to maintain healthy and viable ocular organs. In mice, while retina and RPE/choroid demonstrate ≈90% overlap of expressed genes, 14,484 of the 16,077 expressed genes showed differential expression between the two tissue types. In addition to the biological and genetic differences, retina and RPE/choroid have different tissue compositions. RPE, exempting potential choroidal contamination, is thought to be composed of a single cell type, whereas the retina is composed of five major neuronal cell types and up to 100 subtypes43,44, where each cell type possesses specific combinations of RNA levels and RNA degradation machinery41,45. The directed RNA degradation machinery of a single cell type, e.g. enriched RPE, with choroid contamination, would result in a uniform pattern, whereas multiple RNA degradation machinery combinations, i.e. retina cell types, would result in multiple effects of different trajectories and velocities, resulting in an averaging and masking of weaker signals. This could in part explain the reduction of DE genes in mouse retina compared to RPE/choroid.

Prior to examining post-mortem tissue samples, we characterized tissues harvested immediately following death in mouse and baboon retina and RPE/choroid tissues. This baseline in gene expression allowed comparisons between tissue types and species and enabled us to determine relative degradation levels of all RNAs present in the retina and RPE/choroid as compared to this 0 h time point. This body of research examined the effect of the PMI on the overall quality of the RNA through RIN analysis, its variability through BCV analysis, and the changes of expression in individual RNAs through RNA-Seq. Thus, to our knowledge, our current study represents the first comprehensive multi-species analysis of RNA metabolism in post-mortem mammalian eye utilizing next-generation sequencing technology.

When conducting biological experiments, variability between replicates can mask changes that might otherwise provide scientific insight. In this study we wanted to determine if the PMI effected the BCV, and what that effect might be. In well-controlled experiments, BCV is typically 0.4 for human data and 0.1 for data on genetically identical model organisms46. Our data matches the expected BCV at 0 h and shows that the PMI’s effect on BCV is minimal, except in the mouse retina at 24 h. One limitation in characterizing the BCV is the limited sample size of 4, whereas a larger sample size would better characterize the BCV for these species and tissues at the PMIs tested. We also examined the PMI’s influence on RIN values and surprisingly saw little to no effect on mouse retina and RPE/choroid up to the 24 h time point. Baboon data showed steady decreases in both tissue types tested with increased reduction in the retina. Interestingly, the baboon RIN data closely resembles what has been shown in the literature relating to degradation of RNA as the PMI increases5,47, while the mouse RIN data remained stable. The species-specific difference in retina RIN values can in part explain the increased number of differentially expressed genes seen in baboon retina at the 24 h PMI. While the RIN value primarily measures rRNA degradation levels, it is also used as a universally accepted indicator of mRNA degradation levels. In contrast to this, the RIN values for our mouse data would indicate the mRNAs are relatively undegraded and intact throughout the PMIs collected, while the sequenced data shows clear progression in the number of DE preserved and degraded genes, reflecting the altered mRNA abundance resulting from mRNA degradation. Additionally, our timeline analysis of DE genes shows clear, progressive preservation or degradation of individual genes. Understanding the reason for mouse RIN values remaining stable through the 24 h PMI while mRNA levels became altered due to gene specific global degradation would provide insight into overall RNA metabolism and the disadvantages and considerations required when using the mouse model for select experiments. Furthermore, this data provides evidence of RIN values incorrectly indicating mRNA quality in post-mortem conditions using mouse retina and RPE/choroid samples. The multiple regression models generated using the qRT-PCR data were better suited to quantify the level of RNA degradation in these samples. This body of research demonstrates tissue- and species-specific differences in RNA metabolism and details a tissue- and species-specific method superior to RIN values for determining RNA degradation in mouse retina and RPE samples.

While RIN values for mouse samples are not a good indicator for post-mortem mRNA degradation, the number of DE genes progresses as the PMI becomes larger, providing evidence for systemic mRNA change post-mortem even in the absence of reduced RIN values. Studies utilizing human retina vary in their death to preservation time, with studies utilizing samples with <6 h death to preservation23,24,25,26,27, studies utilizing samples with <24 h death to preservation28,29, and studies where the death to preservation time is not given30,31. Our mouse data shows that retina and RPE/choroid PMI samples from 3 to 6 hours can be treated as biological replicates while reducing the inter-sample variability. Our data suggest that samples with PMIs greater than 6 hours are still viable options for research on rare tissues, but consideration should be given to match PMIs both in biological replicates and to the control, and the increased BCV may reduce the statistical power of DE testing. Our dataset also provides quantification of changes in every gene between any of the PMIs collected, aiding the design and interpretation of experiments in which rare disease samples of varying PMIs must be used.

Mammalian genes maintain a high degree of homology and sequence conservation between species48, while enhancer and regulatory regions have rapidly adapted in a species-specific manner to create divergent function and morphology49. As lncRNAs are known to be involved in multiple aspects of genetic regulation, we further characterized the preservation and degradation of lncRNAs in post-mortem samples to gain insight into lncRNA metabolism50. One major species-specific difference observed in this data set is the number of expressed lncRNAs in mouse retina and RPE/choroid, 843 and 717 respectively, in contrast to baboon retina and RPE/choroid, 101 and 99, respectively, which is explained by lncRNAs currently being poorly annotated in the baboon species, with only 707 lncRNAs compared to 8897 lncRNAs in the mouse genome currently annotated.

Though intra-species analysis based on the number of lncRNAs is not possible due to differences in annotation levels, inter-species analysis of mouse lncRNA and pseudogenes indicates an enrichment in preserved transcripts in the RPE, but not the retina, relative to total expressed transcripts. As highly stable RNAs are typically enriched in housekee** functions, and less stable RNAs are enriched in transcription and signaling factors, chromatin-modifying enzymes and genes with cell-cycle-specific functions51, our data show that mouse lncRNAs and pseudogenes have a similar ratio of preserved to degraded genes in a tissue-specific manner (Fig. 7)51, and supports the finding that pseudogenes can act in a regulatory manner mechanically equivalent to lncRNAs52.

While some mammalian genes are thought to maintain a high degree of homology48 the preservation and degradation of individual transcripts vary more by tissue species than by species tissue (Tables 3, 4). Similarly, GO term enrichment analysis also varies more by tissue species than by species tissue (Supplemental Tables 7, 8), indicating that transcript level species differences are not randomly selected but systematically programmed, with species level programming outweighing tissue level specification. While the mechanistic explanation for these species and tissue differences are beyond the scope of this manuscript, we believe our data to be the first to contrast species and tissue differences of RNA metabolism in a controlled and directed experimental setting. Better understanding the driving factors of species level global differences of RNA metabolism coupled with transcript level motifs will help elucidate steady state system dynamics and perturbations from this steady state which can lead to disease phenotypes.

Materials and Methods

Mice and baboon

All experiments were approved by the Institutional Animal Care and Use Committees at Case Western Reserve University (IACUC protocol No. 2014-0071) and University of California, Irvine (IACUC protocol No. AUP18–124), and were conducted in accordance with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Visual Research. Male C57BL/6J mice were obtained from the Jackson Laboratory and handled at Case Western Reserve University. Baboon tissue samples were obtained from Southwest National Primate Research Center and processed at Case Western Reserve University. The baboons were 11-year-old females and had no known family relationship. Mice were delivered at 21 days of age and sacrificed at 30 days of age. Four mice were allocated to each PMI. Mice were cervically dislocated and allowed to sit at room temperature (RT) until the determined harvesting times of 0 h, 15 min, 30 min, 45 min, 60 min, 3 h 6 h, 12 h, or 24 h. An additional 4 male C57BL/6J mice of 30 days of age were cervically dislocated and had their orbits extracted upon death and stored at RT in sterile PBS in 1.5 microcentrifuge tubes until the 24 h PMI. Extraction of tissues from Baboon was performed by staff at the National Primate Research Center, and intact eye globes were collected at 0 h and stored at RT in sterile PBS in a sealed 50 mm conical tube until the required PMI was reached.

Isolation of retina and RPE/choroid

Mouse and baboon retinas were extracted, and the RPE/choroid was isolated according to published protocols54, and read counts were generated using the featureCounts function of Rsubread55. Genes with 1≥ CPM in 4 or more replicates were considered expressed and used in all downstream analyses. Differential gene expression analysis was performed using edgeR32,56. Preserved and degraded DE gene timelines were created using a custom script and the R package ggplot2. Functional annotations were obtained using DAVID (Version 6.8)57,58.

To determine gene overlap between species homologous human genes were obtained by biomaRt using either baboon or mouse ensembl IDs as inputs. 89% and 93% of mouse and baboon genes, respectively, mapped to human homologs, and were used for downstream inter- and intra-species analyses where noted.

Genes which had a BCV of less than 0.075 for the 0, 3, 6, 12, and 24 hr time points had their Z score plotted from 0 to 24 hrs. The top genes that had visually consistent slopes, 9 positive and 9 negative, were selected for primer design and RT-qPCR (sup table X). The mean RT-qPCR cycle threshold (Ct) values for each biological replicate were determined and each gene had the replicate’s mean Ct value subtracted from its Ct value, providing an internal normalization for each biological replicate.

A custom R script was designed to perform multiple linear regression on every possible 3 gene combination of the 18. The models generated with a P value ≥ 0.05 were used to generate PMI estimates. The means of the PMI estimates were then determined and used as the model’s final PMI estimate.