Impact of enzymatic digestion on single cell suspension yield from peripheral human lung tissue

Abstract

An increasing number of translational investigations of lung biology rely on analyzing single cell suspensions obtained from human lungs. To obtain these single cell suspensions, human lungs from biopsies or research-consented organ donors must be subjected to mechanical and enzymatic digestion prior to analysis with either flow cytometry or single cell RNA sequencing. A variety of enzymes have been used to perform tissue digestion, each with potential limitations. To better understand the limitations of each enzymatic digestion protocol and to establish a framework for comparing studies across protocols, we performed five commonly published protocols in parallel from identical samples obtained from 6 human lungs. Following mechanical (gentleMACS^™) and enzymatic digestion, we quantified cell count and viability using a Nexcelom Cellometer and determined cell phenotype using multiparameter spectral flow cytometry (Cytek^™ Aurora). We found that all protocols were superior in cellular yield and viability when compared to mechanical digestion alone. Protocols high in dispase cleaved immune markers CD4, CD8, CD69, and CD103 and contributed to an increased monocyte to macrophage yield. Similarly, dispase led to a differential epithelial cell yield, with increased TSPN8⁺ and ITGA6⁺ epithelial cells and reduced CD66e⁺ cells. When compared to collagenase D, collagenase P protocols yielded increased AT1 and AT2 cells and decreased endothelial cells. These results provide a framework for selecting an enzymatic digestion protocol best suited to the scientific question and allow for comparison of studies using different protocols.

Introduction

Multiparameter flow cytometry (FC) and single cell RNA sequencing (scRNA) are invaluable tools for phenotyping cellular populations from human tissue, including lungs obtained through biopsy, organ donation, or explantation^12,34. These tools have helped shape the understanding of organ-specific immune responses and provided definitive evidence that mucosal immune populations are phenotypically and functionally distinct from the circulating and lymphatic compartments. scRNA has helped define epithelial cell and fibroblast subsets associated with lung damage and regeneration, contributing to promising new therapeutics^4–8. Using curated libraries of receptor-ligand relationships, computational analyses of scRNA can predict cellular interactions. In-vitro functional analyses from single cells can provide insights into how local cells interact with each other to alter the inflammatory environment within mucosal tissues⁹. However, the interpretation of any single cell analyses is directly impacted by the reliability of the methods of cellular extraction from solid organs.

To obtain single cell suspensions from human lungs, tissues are typically subject to a combination of mechanical and enzymatic digestion in the attempt to disrupt cell-cell adhesions, break down the extracellular matrix, and liberate cells. The enzymes typically used for tissue digestion include various types of collagenase, deoxyribonuclease 1 (Dnase), and in some instances, dispase. Collagenase degrades collagen triple helices in lung tissue; there are distinct types of collagenase which vary by enzyme activity and purity. Dnase helps prevent cell aggregation by the hydrolytic cleavage of phosphodiester linkages, leading to the degradation of free-DNA released by dying cells¹⁰. Dispase is a neutral protease with mild proteolytic activity that cleaves an amino acid sequence that is found in high frequency in collagen IV and fibronectin¹¹. Importantly, dispase has been shown to cleave certain cell surface markers in T cells such as CD4, CD8, CD25, and CD69¹². Typically, these enzymes are used in various combinations to generate single cell suspensions, although there is no known standardization in published literature. A better understanding of how enzyme combinations impact the cellular yield from human lungs can provide a framework for tailoring a digestion protocol to the study question, and for predicting limitations in interpretation.

We compared five commonly utilized protocols for mechanical and enzymatic digestion and compared them with mechanical digestion alone^{1–3,13–19}. All protocols use Dnase but vary in the type of collagenase, the presence and concentration of dispase, and the duration of enzymatic digestion. We used multiparameter spectral FC (Cytek^™ Aurora) to analyze the cellular phenotypes generated from the single cell suspensions and to quantify the degree of epitope cleavage. Finally, one additional lung sample was digested using the two most similar protocols and sent for single cell RNA sequencing to analyze differences in cell subsets and gene expression.

Methods

Tissue Collection

Human lungs were obtained through the University of Pittsburgh Biospecimen Core via our local organ procurement organization, the Center for Organ Recovery and Education (CORE). Acquisition of lungs was approved by the University of Pittsburgh’s Committee for Oversight of Research and Clinical Training Involving Decedents (CORID: 1116). All lungs were obtained from donors without antecedent lung disease that were rejected for clinical transplantation, except the lung used for single cell RNA sequencing analysis which was obtained from a lung transplant recipient with COVID-19 related fibrosis. The procurement of explanted lungs from transplant recipients was approved by our institutional review board (STUDY21080113: Lung Explant Biorepository). All lung samples were weighed and cut into six 1.5g pieces for each protocol. Each segment of lung was subject to mechanical dissection (gentleMACS^™ Dissociator using gentleMACS^™ C tubes at a Lung 2 setting per manufacturers recommendations), and six different protocols (Table 2) that varied based on enzymatic digestion composition and duration. Following mechanical and enzymatic digestion, samples were strained through a 70um filter and centrifuged at 300 G for 5 minutes at 4°C. Samples were resuspended in 5mL ACK lysing buffer (Thermo Fisher) on ice for 5 minutes, washed in phosphate-buffered solution and centrifuged again at 300 G for 5 minutes at 4°C. Afterwards, cells were cryopreserved in fetal bovine serum (FBS) with 10% DMSO at a concentration of 5 x 10^6 cells per 1 mL media. Cells were slowly cooled at a rate of 1°C/minute (using a Nalgene^® Mr. Frosty) in a −80°C freezer and then transferred to vapor-phase liquid nitrogen tank for long-term storage.

Table 2.

Organ donor demographics

Sample ID	Ethnicity	Age	Sex	Cause of Death	Analysis
2021-026-C	White	20	M	Drug Intoxication	Flow Cytometry
2021-038-C	White	25	M	Drowning / drug intoxication	Flow Cytometry
2021-039-C	White	50	M	Asphyxiation s/p hanging	Flow Cytometry
2021-040-C	White	20	M	Intracranial hemorrhage	Flow Cytometry
2021-041-C	White	81	F	Intracranial hemorrhage	Flow Cytometry
2021-014-LT	White	24	M	COVID-19	scRNA-seq

Cell Counting and Viability

Prior to cryopreservation, 10uL of the single cell suspension from each protocol was combined in a 1:1 ratio with fluorescent Vitastain dye and plated on a cell counting chamber; cell number and viability were then automatically detected with a Cellometer2000 (Nexcelom Bioscience). The program “immune cells low RBC” was run and the calculated cell yield and viability were recorded.

Flow Cytometry

Cryopreserved cells were thawed with warmed media (10% FBS in RPMI) and re-expanded in 100 μl of fluorescence-activated cell sorter (FACS) buffer (PBS + 1% FBS) in the presence of 5ml of Fc receptor blocking solution (Human TruStain FcX^™) for 10 minutes at room temperature. Cell surface antibodies were then applied at room temperature for 30-60 minutes in a total FACS buffer volume of 200 μl. Cells were then fixed with eBioscience Intracellular Fixation and Permeabilization Buffer Set (Invitrogen) on ice for 60 minutes as per manufacturer guidelines. Intracellular staining was performed in the presence of permeabilization buffer at room temperature for 30-60 minutes. Flow Cytometry was performed using a Cytek^™ Aurora spectral flow cytometer (CyTek Biosciences). Generated flow cytometry standard (FCS) files were analyzed using FlowJo software (FlowJo LLC). All antibodies used for flow cytometry can be found on Suppl Table 1 & 2. Unmixed fcs files have been uploaded to the online flow repository (flowrepository.org: FR-FCM-Z6C8, FR-FCM-Z6CJ).

Single Cell RNA Sequencing

One human lung was divided into two equivalent segments and underwent mechanical and disparate enzymatic digestion (collagenase P versus collagenase D protocols). Cells were stained with cell hashing antibody from Biolegend following the CITE-seq protocol (https://cite-seq.com/protocols/). After the final wash, cells were passed through a 40uM cell strainer and enumerated by Cellometer2000 right before loading onto 10x Chromium controller for cell capture using the 5’ V2 kits. Libraries for gene expression and hash tag oligos were constructed following 10x Genomics user guide and sequenced on an Illumina Novaseq 6000 targeting 50,000 reads per cells. Sequencing data were processed with Cellranger5.0 before downstream analysis using Seurat.

Single Cell RNA-Seq data processing

Single cell RNA sequencing analysis was performed using Seurat 4.0 with R (version 4.1.1). Normalization and variance stabilization of count data was performed using scTransform²⁰. Seurat objects were then integrated using 3000 identified anchors based on the previously transformed normalization values. Spatial visualization of distinct clusters was created using Uniform Manifold Approximation and Projection (UMAP) for dimension reduction²¹. Differential gene expression was performed from this subset using non-parametric Wilcoxon rank sum test. The results were adjusted for multiple comparisons using Bonferroni correction. All code used for single cell analyses can be found in the following GitHub repository (https://github.com/markesnyder/Lung_Digest), including list of all R packages used for analyses.

Statistical Analysis

Statistical analyses were performed using GraphPad (Prism). For all analyses, a two-tailed p-value of < 0.05 was the threshold used to determine statistical significance. Paired t-test was used to test for difference in cell counts, cellular viability, and flow cytometry cellular phenotypes between digestion protocols. R statistical software (R Foundation for statistical computing, Vienna, Austria) was used to analyze single cell RNA-seq results. Manuscript figures were compiled using Adobe Illustrator CC 2017 (Ventura, CA).

Results

We identified 5 distinct published methods of mechanical and enzymatic digestion to isolate single cell suspensions from human lung tissue (Table 1). These protocols were named according to the most unique component of each protocol as follows: Control, dispase^lo, dispase^hi, Col_P (collagenase P), Col_D (collagenase D), and Cold_D+ (collagenase D + density centrifugation). These methods, as well as a control method without enzymatic digestion, were used on identically weighed biopsies of lungs from 5 research-consented organ donors (Table 2). Using a Cytek^™ Aurora spectral flow cytometer we were able to differentiate CD4⁺ T cells, CD8⁺ T cells, NK cells, NKT cells, B cells, alveolar macrophages, and interstitial macrophages/monocytes with one panel (Suppl fig 1) and Type 1 pneumocytes, type II pneumocytes, endothelial cells, and fibroblasts from a second panel (Suppl fig 2).

Table 1.

Mechanical and enzymatic digestion protocols

Name	Protocol	Enzyme	Digestion	Duration (Hours)	Citations**
Control	1	None	-Dissect with scissors and use gentleMACS for 30 seconds -Place on 37C shaker for 1.5 hours	1-3	-
Dispaselo	2	Liberase DL* Dnase	-Dissect with scissors, liberase, and Dnase 1 -Place in 37C water bath for 35 minutes -Swirl mixture and heat for 10 more minutes	1	1,2,3,4,5
Dispasehi	3	Dispase Collagenase D Dnase	-Dissect with scissors and dispase -GentleMACS for 30 seconds -Incubate at 37C on shaker for 30 minutes -gentleMACS for 30 seconds	1	6
Col_P	4	Collagenase P Dnase	-Dissect tissue with scissors and add digestion media -Mechanical digestion with heated octoMACS dissociator (35 minutes)	0.75	7
Col_D	5	Dnase Collagenase D Trypsin Inhibitor	-Inject digest into lung and let sit for 5 minutes -Dissect with scissors and use gentleMACS for 30 seconds -Place on 37C shaker for 1.5 hours	3	8,9
Col_D+	6	Dnase Collagenase D Trypsin Inhibitor	-Inject digest into lung and let sit for 5 minutes -Dissect with scissors and use gentleMACS for 30 seconds -Place on 37C shaker for 1.5 hours -Density centrifugation with Percoll PLUS	4	11,12

^*Liberase DL contains collagenase I, collagenase II, and low quantity of dispase

^**Studies using same or similar enzymatic digestion approach

Unsurprisingly, the cell viability and count were lowest in the control protocol (Figure 1A, 1B). The highest degree of cellular viability was found in the dispase^hi and Col_D+ protocols but with no statistically significant difference in viability with either the dispase^lo, Col_P, or Col_D protocols (Figure 1A). Overall cell counts were highly variable across samples within each protocol, but consistent trends were found across protocols for each sample, with a trend towards increased cellular yield in the dispase^lo, dispase^hi, and Col_P protocols (Figure 1B). Immune cell yield varied greatly across protocols (Figure 1C); the Col_D protocol had the highest yield of T cell, B cell, and macrophages subsets; the dispase^hi protocol yielded the highest percentage of monocytes; the Col_P protocol yielded the highest percentage of NK cells. There was no statistically significant difference in the proportion of global populations across protocols (epithelial cells, endothelial cells, fibroblasts, immune cells); however, substantial endothelial populations resulted only from the Col_P and Col_D protocols, fibroblasts were found predominantly in the dispase^hi and Col_D protocols, and epithelial cells were found to be significantly higher than lymphocytes in the dispase^hi protocol (Figure 1D).

Figure 1. Cellular yield by digestion protocol.

(A) Histogram of percent live cells obtained by each protocol (N = 5 lungs). (B) Histogram of cells per million live cells obtained by each protocol (N = 5 lungs). (C) Percent of live immune populations per protocol (N = 5 lungs). (D) Percent of live overall cellular populations (N = 5 lungs).

We next set out to determine the differential impact of enzymatic digestion on the cleavage of common immune cell surface markers. Like prior reports, we found that dispase cleaved both CD4 and CD8 in a dose-dependent manner (Figure 2A, 2B). Despite the cleavage of CD4 and CD8, there was no significant difference found in the markers for T cell subsets- TCM, TEM, TEMRA, and naïve T cells (Figure 2D, 2E). For CD69 and CD103 markers on flow cytometry, there was also significant cleavage in the dispase^hi protocol, as compared to the Col_D protocol (Figure 2F). The effect of this cleavage was more prominent in the CD4⁺CD69⁺CD103⁻ population compared to the CD4⁺CD69⁺CD103⁺ population (Figure 2G). The cleavage effect was similar between CD8⁺CD69⁺CD103⁻ and CD8⁺CD69⁺CD103⁺ populations.

Figure 2. Impact of enzymatic digestion on immune cell surface markers and proportions.

(A) Proportion of CD8+ T cell subsets based on protocol (N = 5 lungs). (B) Proportion of CD4+ T cell subsets based on protocol (N=5 lungs). (C) Representative flow cytometry plots showing dose effect of dispase on CD4 and CD8 epitope cleavage. (D, E) Cumulative data of (D) CD4+ and € CD8+ T cell expression based on dispase concentration (***p-value <0.001, ***p-value <0.0001). (F) Representative flow cytometry plots showing the effect of dispase on markers of tissue residency, CD69 and CD103. (G, H) Cumulative data showing the degree of cleavage of CD69 and CD103 based on protocol.

Although it cleaved many important immune cell markers, the dispase protocol consistently yielded the highest percentage of live epithelial cells. We compared the percent yield of the specialized lung epithelial cells to see the effect of the digestion methods. CD66e⁺ epithelial cells were lowest in the dispase protocol and highest in the liberase protocol (Figure 3A). There was no significant difference in ITGA6⁺ epithelial cells across methods (Figure 3B). The dispase^hi protocol yielded the highest TSPN8+ population compared to all other methods (Figure 3C). No significant difference was found for HLADR+ and T1a+ populations (Figure 3D, E) and there was high inter-sample variability.

Figure 3. Epithelial cell surface markers by protocol.

The percentage of live, CD45-CD31-EPCAM+ cells (%) for (A) CD66e - Carcinoembryonic antigen-related cell adhesion molecule 5, (B) ITGA6 – Integrin subunit alpha 6, (C) TSPN8 – tetraspanin 8, (D) Major histocompatibility complex II – HLADR, and (E) podoplanin – T1a.

Due to the similarities of the digestion components of Col_P and Col_D protocols, we performed single cell RNA sequencing to examine differential effects between these protocols from one donor who underwent lung transplant for COVID-related fibrosis. We identified 15 distinct clusters shared between the two protocols (Figure 4A). Directly comparing the proportion of cells in each cluster between the two protocols, we found that the ColP protocol yielded significantly more macrophages and epithelial cells (ciliated, type 1 pneumocytes, and type 2 pneumocytes) than the ColD protocol; the ColD protocol yielded significantly more immune cells, endothelial cells, basophils, and NK cells (Figure 4A, 4B). We next subset the T lymphocytes (based on CD3E expression, and lack of HBB, LYZ, CD68, CD19 expression) and macrophages populations (based on LYZ and CD68 expression) to identify any transcriptional differences between protocols. We found very few differentially expressed genes with a p-value <0.05 and log2 fold change > 0.8 or < −0.8 (5 for T cells and 1 for myeloid cells). From the T cell subset, the ColD protocol had higher expression of transcripts related to lymphocyte cytotoxicity (GNLY and FGFBP2), while the ColP protocol had higher expression of transcripts related to immune tolerance (DUSP4), negative regulation of cell-cell interactions (MYADM), regulator of cell cycle (RGCC), a transcriptional repressor (ZNF331), and a cytokine preferentially found in senescent T cells (GZMK) (Figure 4C). In the macrophage subset, the ColP protocol had higher expression of genes related to cell stress (BAG3, HSPA6), activation (IL1B), and apoptotic signaling (GOS2) (Figure 4D).

Figure 4. Impact of collagenase P vs collagenase D on single cell subsets via RNA-seq.

(A) Uniform Manifold Approximation and Projection (UMAP) 2-dimensional reduction of annotated cells split by protocol. (B) Differential cellular proportionality based on protocol; to the right of the solid line denotes increased proportion from protocol 3, to the left is increased proportion from protocol 5. (C) Volcano plot comparing T cells isolated from the collagenase D versus collagenase P protocols, red dots denote genes with log₂fold change >|0.8| and p-value < 0.05. (D) Volcano plot comparing Myeloid cells isolated from the collagenase D versus collagenase P protocols, red dots denote genes with log₂fold change >|0.8| and p-value < 0.05.

Discussion

Obtaining representative single cell suspensions from human lungs for phenotypic and functional assays requires the combination of mechanical and enzymatic digestion. A variety of enzymes are available for tissue digestion, each with distinct advantages and disadvantages that can impact the quantity and viability of the cellular yield as well as disrupt cell surface epitopes. Herein we compare 5 distinct, previously published and commonly used approaches to mechanical and enzymatic digestion of human lungs and compare them with mechanical digestion alone. We found a substantial degree of intra-protocol variability in cellular yield across human lungs, likely a reflection of the heterogeneity of human lung cellular composition confounded by age, gender, and cause of death. However, we found a consist increase in cellular viability in protocols containing collagenase P with shorter digestion times. All methods of digestion had increased absolute number of viable cells than mechanical digestion alone.

Importantly, we found distinct and consistent differences in the type of cells, both immune and non-immune, obtained across protocols. The dispase^hi protocol consistently yielded more monocytes than macrophages and this difference did not appear to be due to cleavage of canonical macrophage cell surface markers. All other protocols resulted in the expected predominance of alveolar macrophages; lymphocytes proportions remained consistent across protocols. We found a significantly disparate population of epithelial cells, with increased number of cells positive for tetraspanin 8 (TSPN8) and integrin alpha 6 (ITGA6), and negative for carcinoembryonic antigen-related cell adhesion molecule 5 (CD66e). The increase in TSPN8⁺ and ITGA6⁺ epithelial cells as well as a decrease proportion of CD66e⁺ cells suggest preferential liberation of goblet and secretory cells at the expense of basal cells by using a dispase^hi protocol²². The lack of dose response with the dispase^lo protocol suggests this is a true difference and not a results of epitope cleavage.

Similar to previous studies, we found that the dispase^hi protocol cleaved both CD4 and CD8 epitopes in a dose dependent manner¹². We also found that the presence of dispase contributed to cleavage of the canonical tissue resident memory markers, CD69 and CD103; CD69 cleavage has been previously reported¹². When studying tissue residency and the experimental readout is flow cytometry, dispase should be avoided. However, epitope cleavage should not impact scRNA unless you plan to perform CITE-seq for these above-listed epitopes. Directly comparing collagenase D with collagenase P, we found vastly different proportions of AT1, AT2, and endothelial cells, with Col_P yielding higher numbers of AT1 and AT2 cells, and Col_C higher endothelial cells. However, there did not seem to be a substantial difference in differential gene expression of immune populations, suggesting neither the type of collagenase nor the duration of digestion impacted gene expression in a meaningful way.

This study has limitations. Due to the number of enzymes available and the paucity of donor lung tissue, we were unable to vary conditions within each protocol, including enzymatic digestion time, and testing without the presence of Dnase. Protocols were tested on distal parenchymal tissue, excluding larger airways, and on lungs without appreciable disease (with the exception of the COVID fibrosis lung). Likewise, spatial heterogeneity of the human lung may have contributed to cellular composition variability. scRNA was utilized to directly compare collagenase P and D, and not dispase. Finally, our flow cytometry analyses were performed on cryopreserved cells. We are not aware of evidence suggesting that cryopreservation changes the relative proportion of distinct cellular subsets. On the contrary, cryopreservation appears to preserve cellular heterogeneity ²³. However, the freeze thaw cycle may impact cellular composition following enzymatic digestion. Despite these limitations, we believe that these results can help guide the selection of an enzymatic digestion protocol for future studies and provide a framework for comparing results across studies. Of particular importance, these results reinforce the limitations of interpreting cellular populations derived from scRNA analyses. Comparing the results of scRNA with spatial transcriptomics may provide a more precise representation of cellular heterogeneity within the human lung in both health and disease.

Supplementary Material

Figure S1

Suppl fig 1. Lung immune cell flow cytometry gating strategy. (A) Flow cytometry gating strategy utilized to isolate lymphocytes, SSC = side scatter, FSC = forward scatter, A=area, H=height. CD4⁺ T cells defined as live, CD3⁺CD56⁺CD4⁺CD8⁻ cells; CD8⁺ T cells defined as live, CD3⁺CD56⁺CD4⁻CD8⁺ cells; Natural killer cells defined as live, CD3⁻CD56⁺; Natural killer T cells defined as live, CD3⁺CD56⁺; TEM = effector memory T cells, defined as CCR7⁻CD45RA⁻, TEMRA = terminally differentiated effector T cells, defined as CCR7⁻CD45RA⁺, Naïve T cells defined as CCR7⁺CD45RA⁺, TCM = central memory T cells, defined as CCR7⁺CD45RA⁻. (B) Flow cytometry gating strategy utilized to isolate granulocytes/myeloid cells.

Figure S2

Suppl fig 2. Non-immune cell flow cytometry gating strategy. (A) Flow cytometry gating strategy to isolate endothelial cells, defined as live, CD45⁻CD31⁺ cells; fibroblasts, defined as live, CD45⁻CD31⁻EPCAM⁻CD90⁺ cells; epithelial cells, defined as live, CD45-CD31-EPCAM+ cells. (B) Gating strategy within the non-immune panel with confirmation of immune population proportions in comparison to non-immune cells. (C) Histograms of different epithelial subset markers, from live, CD45⁻CD31⁻EPCAM⁺ cells (positive gating determined using negative internal controls, cells known to be negative for the marker of interest, displayed as shaded histogram).

Abbreviations:

FC

Multiparameter Flow Cytometry

scRNA

Single cell RNA sequencing