Available via license: CC BY 4.0
Content may be subject to copyright.
RESEARCH
Robust detection of clinically relevant features in single-cell
RNA profiles of patient-matched fresh and formalin-fixed
paraffin-embedded (FFPE) lung cancer tissue
Alexandra Trinks1· Miha Milek2· Dieter Beule2· Julie Kluge3· Stefan Florian3,4 ·
Christine Sers3,5 · David Horst3,5 · Markus Morkel1,3,5 · Philip Bischo3,4,5
Accepted: 23 January 2024
© The Author(s) 2024
Abstract
Purpose Single-cell transcriptional proling reveals cell heterogeneity and clinically relevant traits in intra-operatively
collected patient-derived tissue. So far, single-cell studies have been constrained by the requirement for prospectively
collected fresh or cryopreserved tissue. This limitation might be overcome by recent technical developments enabling
single-cell analysis of FFPE tissue.
Methods We benchmark single-cell proles from patient-matched fresh, cryopreserved and archival FFPE cancer tissue.
Results We nd that fresh tissue and FFPE routine blocks can be employed for the robust detection of clinically relevant
traits on the single-cell level. Specically, single-cell maps of fresh patient tissues and corresponding FFPE tissue blocks
could be integrated into common low-dimensional representations, and cell subtype clusters showed highly correlated
transcriptional strengths of signaling pathway, hallmark, and clinically useful signatures, although expression of single
genes varied due to technological dierences. FFPE tissue blocks revealed higher cell diversity compared to fresh tissue.
In contrast, single-cell proling of cryopreserved tissue was prone to artifacts in the clinical setting.
Conclusion Our analysis highlights the potential of single-cell proling in the analysis of retrospectively and prospectively
collected archival pathology cohorts and increases the applicability in translational research.
Keywords Single-cell RNA sequencing �Single-cell transcriptomics �Lung cancer �FFPE tissue analysis �
Tumor heterogeneity
1 Introduction
High-throughput single-cell RNA sequencing (scRNA-seq)
has enabled researchers to study the various aspects of
cellular heterogeneity of tissues, including human clinical
samples. In particular, scRNA-seq has been used exten-
sively to characterize the complex cell compositions of
tumor tissue cohorts [1–3], and can reveal features which
contribute to patient prognosis [4,5] and therapy resistance
[6]. So far, single-cell RNA proling required the prospec-
tive collection of fresh tissue samples during surgery or
biopsy. Retrospective analysis of frozen or formalin-xed
and paran-embedded (FFPE) tissue cohorts would allow
faster correlation of scRNA-seq proles to clinical fea-
tures, as many important clinical characteristics, such as
tumor genetics, therapy response and patient survival, are
only obtainable weeks to years after sample acquisition. In
addition, this could unlock archival FFPE tissue collections
for single-cell proling.
Markus Morkel and Philip Bischo have equally contributed to this
article.
✉Markus Morkel
markus.morkel@charite.de
✉Philip Bischo
philip.bischo@charite.de
1Bioportal Single Cells, Berlin Institute of Health at Charité –
Universitätsmedizin Berlin, Berlin, Germany
2Core Unit Bioinformatics (CUBI), Berlin Institute of Health at
Charité – Universitätsmedizin Berlin, Berlin, Germany
3Institute of Pathology, Charité – Universitätsmedizin Berlin,
Corporate Member of Freie Universität Berlin and Humboldt-
Universität zu Berlin, Charitéplatz 1, 10117, Berlin, Germany
4BIH Biomedical Innovation Academy, BIH Charité Clinician
Scientist Program, Berlin Institute of Health at Charité –
Universitätsmedizin Berlin, Berlin, Germany
5German Cancer Research Center (DKFZ), German Cancer
Consortium (DKTK) Partner Site Berlin, Heidelberg, Germany
Cellular Oncology
https://doi.org/10.1007/s13402-024-00922-0
Recently, scRNA-seq chemistry has been developed
for frozen and formalin-xed single-nucleus suspensions.
Early studies have shown that this allows performing
single-nucleus RNA sequencing (snRNA-seq) of nuclei
isolated from frozen and FFPE tissue samples [7–10] at
read depths that allow a similarly ne-grained analysis
compared to fresh cell suspensions. However, to our
knowledge, no side-by-side comparisons have been per-
formed on fresh, frozen, and FFPE clinical tissue samples
from the same patients. To answer the question which
features of fresh tissue are preserved in archival samples,
we proled three lung adenocarcinomas using fresh, cryo-
preserved and FFPE tissue samples side-by-side. We
found that cell-intrinsic and clinically relevant features
of cancers are robustly preserved in single-cell transcrip-
tomes of FFPE samples.
2 Methods
2.1 Sample acquisition
Tissue samples were collected from therapy-naïve lung
adenocarcinoma patients undergoing primary surgery.
Informed consent was obtained from all participants before
sample acquisition. The ethics committee of Charité -
Universitätsmedizin Berlin approved the use of tissue sam-
ples for single-cell gene expression analysis (vote EA4/
164/19). Fresh tumor tissue close to the invasive front was
sampled by a pathologist during intraoperative pathology
consultation. From the same portion of tissue, tissue sam-
ples of approx. 5 × 5 × 5 mm each were either immediately
snap-frozen in liquid nitrogen and stored at −80 °C, or
placed in Tissue Storage Solution (Miltenyi) on ice for
max. 30 min before subsequent further processing of the
fresh tissue. The remaining surgery specimen was sub-
jected to routine histological examination, including for-
malin xation and paran embedding. FFPE tissue was
sampled from archive tissue blocks after storage under
standard archive conditions (dry, room temperature) for
4–5 months. For FFPE tissue sampling, 4–6 tissue cores
of 1 mm diameter and 1–3 mm length, depending on the
block thickness, were punched out of FFPE tissue blocks.
Tissue cores were obtained from a tumor area of approx.
5 × 10 mm close to the invasive front, which had been
marked by a pathologist, in order to closely resemble the
intraoperatively collected fresh and cryopreserved tissue
(Fig. S1a).
2.2 Processing of fresh tissue samples
Fresh tissue samples were cut into small pieces of 1 mm
diameter and dissociated using the Tumor Dissociation Kit,
human (Miltenyi) and a gentleMACS Octo Dissociator
with heaters (Miltenyi), using preinstalled program
37C_h_TDK_1 for 30–45 min. All subsequent steps were
performed at 4 °C or on ice. Dissociated tissue was ltered
through 100 µm lters, pelleted by centrifugation at 300×g
in BSA-coated low-binding tubes, incubated with 1 ml
ACK erythrocyte lysis buer (Gibco) for 1 min, washed
with DMEM, pelleted, resuspended in PBS and ltered
through 20 µm lters. Finally, debris was removed using
the Debris Removal Solution (Miltenyi) and cells were
counted using a Neubauer chamber. Single-cell suspen-
sions were further processed using the Chromium Single
Cell 3′Reagent Kit v3 and the Chromium Controller (10x
Genomics, Pleasanton, California, USA) according to the
manufacturer’s protocol without any adjustments.
2.3 Processing of cryopreserved tissue samples
Cryopreserved tissue samples were cracked using a pestle
and mortar placed on dry ice and pre-chilled with liquid
nitrogen into small pieces of 1 mm diameter. All subse-
quent steps were performed at 4 °C or on ice. Tissue pieces
were homogenized in homogenization buer [1x Nuclei
EZ lysis buer (Sigma), 0.6 U/ml RNAse Inhibitor
(Ambion), 0.3 U/ml Superasin (Ambion)] using a pestle
and douncer by approx. 10 strokes with a loose pestle and
5 strokes with a tight pestle. Homogenized tissue was
ltered through a 30 µm lter, stained with DAPI (0.1
µg/mL) for 5 min and sorted using a BD FACSAria
Fusion (100 µm nozzle) to remove debris and doublets.
Cell concentration was determined using a Neubauer
chamber before single-nuclei suspensions were further pro-
cessed using the Chromium Single Cell 3′Reagent Kit v3
and the Chromium Controller iX (10x Genomics,
Pleasanton, California, USA) according to the manufac-
turer’s protocol without any adjustments.
2.4 Processing of FFPE tissue samples
FFPE tissue samples were processed according to
a published protocol [8] with adjustments according to
another published protocol [9]. For a detailed version of
the protocol see Supplementary Methods. In short, tissue
cores were obtained as described above and cut into small
pieces of 1 mm diameter, washed at room temperature in
xylene 3 times for 10 min, in 100% ethanol 2 times for
30 s, in 70% ethanol for 30 s, in 50% ethanol for 30 s, in
distilled water for 30 s, and in Buer V (FFPE Tissue
Dissociation Kit, Miltenyi) for 30 s. Tissue was dissociated
using the FFPE Tissue Dissociation Kit (Miltenyi) and
a gentleMACS Octo Dissociator with heaters (Miltenyi),
using preinstalled program 37C_FFPE_1. After 20 min of
dissociation and after complete dissociation, dissociated
Cellular Oncology
tissue was pipetted through a 20 G needle for 10–20 times.
Next, samples were ltered through 70 µm lters, placed
on ice, washed with chilled Buer V, resuspended in
Resuspension Buer (0.5x PBS, 50 mM Tris pH8,
0.02% BSA, 0.24 U/µl RNasin (Ambion, AM2684) in
H2O), and cells were counted using a Neubauer chamber.
Single-nuclei suspensions were further processed using
the Chromium Fixed RNA Proling Reagent Kit and the
Chromium Controller Xi (10x Genomics, Pleasanton,
California, USA) according to the manufacturer’s proto-
col. After step 2.1.m in the manufacturer’s protocol, cells
were stained with DAPI (0.1 µg/mL) for 5 min and sorted
using a BD FACSAria Fusion (100 µm nozzle) to
remove debris and doublets. After cell sorting, samples
were further processed according to the manufacturer’s
protocol.
2.5 Library preparation and sequencing
After isolation, 10,000 cells/nuclei were subjected to
barcoding and library preparation. Libraries of fresh and
cryopreserved samples were prepared using the Chromium
Single Cell 3′Reagent Kit v3 (10x Genomics) according
to the manufacturer’s protocol. Libraries of FFPE sam-
ples were prepared using the Chromium Fixed RNA
Proling Reagent Kit (10x Genomics) according to the
manufacturer’s protocol. Libraries were sequenced on
a NovaSeq (Illumina) at approx. 400 mio. reads per
library.
2.6 Data analysis
Sequencing reads were aligned against reference tran-
scriptome GRCh38 and UMIs were quantied using
Cellranger, version 7.1.0 (10x Genomics). Subsequent
analyses were performed using R version 4.1.1 and the
Seurat package version 4.3.0 [11], if not stated other-
wise. First, signal from ambient RNA was removed
using the SoupX package version 1.6.2 [12], assuming
a contamination fraction of up to 0.2. Gene expression
data of all samples were merged and ltered for the
following quality parameters: 300–10,000 genes per
cell, 500–100,000 UMIs per cell, fraction of mitochon-
drial reads lower than 15% and fraction of hemoglobin
reads lower than 5% (see Fig. S1b for cell counts before
and after ltering). Next, gene expression data was log-
normalized and dimensionality reduction was performed
by principal component analysis (PCA). Uniform mani-
fold approximation projection (UMAP) based on the top
10 principal components (PCs) was used for data visua-
lization. Gene expression data was integrated by sample
type (fresh, frozen or FFPE) using reciprocal PCA based
on the top 30 PCs. Clustering was performed using
shared nearest neighbor graph calculation. In fresh and
FFPE data, main cell types and cell subtypes were
manually annotated using canonical marker genes
selected from the literature. Cell type labels were trans-
ferred based on the rst 30 PCs to the frozen data using
the FindTransferAnchors and TransferData functions of
the Seurat package. Clusters containing cell doublets
were identied by discrepant marker gene expression
and removed prior to further analysis. Signaling path-
way activity scores were calculated using the PROGENy
package version 1.17.3 [13]. Gene signature expression
scores were calculated using Seurat. The inferCNV
package version 1.10.1 (https://github.com/broadinsti
tute/inferCNV) was used for copy number analysis in
epithelial transcriptomes.
2.7 Immunohistochemistry and image analysis
FFPE tissue sections of 2–3 µm were prepared for immu-
nohistochemistry (IHC). Staining procedures were per-
formed on Bond MAX (Leica) and BenchMark XT
(Ventana) immunostainers according to manufacturer’s
protocols. Tissue sections were deparanized, rehydrated,
and antigen retrieval was performed either by protease
incubation for 8 min (for Cytokeratin), or incubation for
30 min at 100 °C in Epitope Retrieval ER2 Solution
(Leica) (for CD45, PD-L1), or CC1 buer (Ventana) (for
ERG), or without antigen retrieval (for smooth muscle
Actin). Incubation with primary antibodies was performed
for 30 min at room temperature. The following primary
antibodies were used: monoclonal mouse anti-Cytokeratin
(clone AE1/AE3, 1:500, Dako/Agilent, cat. no. M3515),
monoclonal mouse anti-CD45 (clone X16/99, 1:100, Leica,
cat. no. NCL-L-LCA), monoclonal mouse anti-smooth
muscle Actin (clone 1A4, 1:400, Dako/Agilent, cat. no.
M0851), monoclonal rabbit anti-ERG (clone EPR3864,
ready-to-use dilution, Ventana/Roche, cat. no. 790-4576),
and monoclonal mouse anti-PD-L1 (clone E1L3N, 1:200,
Cell Signaling, cat. no. 13684). Subsequently, tissue sec-
tions were incubated with HRP-conjugated secondary anti-
body (Leica) and DAB, and nuclei were counterstained
with hematoxylin.
After IHC staining was completed, tissue sections were
digitized using a DP600 slide scanner (Ventana/Roche).
For quantication of positive cells, the open-source soft-
ware QuPath (version 0.3.2) was used. Three representa-
tive tumor regions of 1 mm² each adjacent to the sampling
area were selected by a pathologist. Within these areas,
cells were detected based on nuclei counterstaining and
positive cells were identied via IHC intensity based on
cut-o values set by a pathologist. The percentage of
Cellular Oncology
positive cells was exported from QuPath for further
analysis.
2.8 Statistical analysis
The quantity of main cell types in fresh and FFPE
samples as well as IHC images were compared using
the paired t test. The correlation of PROGENy pathway
and Hallmark signature scores in dierent sample types
(fresh, frozen, FFPE) was analyzed by calculating the
Pearson correlation coecient. Dierentially expressed
genes were identied using the FindAllMarkers function
of the Seurat package with the following parameters:
include only positive markers, proportion of expressing
cells inside the cluster ≥ 0.2, dierence between pro-
portions of expressing cells inside and outside the
cluster ≥ 0.2.
3 Results
3.1 Quality metrics of single-cell transcriptomes
from fresh, frozen and archival FFPE samples
We selected cryopreserved and FFPE tissue samples of
three lung adenocarcinoma patients (P075, P078, P079)
that were previously analyzed by fresh tissue scRNA-seq
for snRNA-seq analysis (Figs. 1a, S1a and S1b). Initial
analysis of fresh and FFPE tissue-derived libraries showed
expected high quality parameters (Figs. 1b and S1c); how-
ever, results gained from cryopreserved tissue showed
unequal and lower quality scores in two subsequent rounds
of library preparation and optimization of lab workows
(Fig. S1d). Across the patients, numbers of detected genes
per sample were highest for fresh tissue, but lower for
FFPE or frozen tissue (Fig. 1b), even when adjusting for
Fig. 1 Workow and quality metrics of fresh, frozen and FFPE tissue
single-cell analysis. aWorkow of tissue specimens used for the
study. In short, fresh tissues were procured intra-operatively, disso-
ciated enzymatically, and cell suspensions were used for single-cell
library preparation. Cryopreserved tissues were stored for 4–5 months
at −80 °C, homogenized in the frozen state for nuclei isolation, and
nucleus suspensions were used for library preparation. For FFPE
analysis, 4–5 months old routine FFPE blocks were dissociated
enzymatically and mechanically, and nucleus suspensions were used
for library preparation. bNumbers of genes called per cell in
the various libraries. Full colors: all genes; lighter colors: genes
limited by FFPE probe set. cUMAPs based on the top 10 principal
components of all single-cell transcriptomes after ltering and
data integration, color-coded by fresh, frozen or FFPE tissue origin.
dQuantication of main cell type, by clustering and calling of cell
type-specic marker genes. eModule score of gene signatures related
to dissociation stress in the various main cell types, by fresh, frozen
and FFPE origin, mean scores per patient indicated by circles
Cellular Oncology
the restricted gene set used in the targeted FFPE sequen-
cing approach. Single-cell information of the fresh, frozen
and FFPE samples could be integrated into a common
UMAP with clusters sharing transcriptomes of fresh, fro-
zen and FFPE tissue origin, indicating that cell type infor-
mation was stable across the methods (Figs. 1c and S1e).
When calling main cell types, we found that immune cell
transcriptomes were enriched, but epithelial and stromal
cells transcriptomes were rather depleted from fresh tissue
single-cell libraries (Fig. 1d), indicating potential negative
eects of tissue dissociation on cell representation. In line
with this interpretation, a recently published gene signature
for dissociation stress [14] was highest in fresh tissue
transcriptomes, in particular in epithelial cells (Fig. 1e).
3.2 Cell type abundancies in FFPE versus fresh
samples
As FFPE tissue blocks are most relevant for clinical appli-
cations and frozen tissue libraries were hampered by low
quality parameters, we integrated fresh and FFPE data for
in-depth analysis (Fig. 2a), and again determined fractions
of transcriptomes of epithelial, immune and stromal origin
(Figs. 2b and S2). The proportion of main cell types as
quantied by IHC was better recapitulated in FFPE data
compared to fresh tissue data, although in both fresh and
FFPE data, we observed a trend towards overrepresenta-
tion of immune cells and underrepresentation of epithelial
and stromal cells. Assessment of epithelial marker genes
showed similar expression scores across fresh and FFPE
tissue samples, as EPCAM, KRT5, SCGB3A2, FOXJ1,
AGER and SFTPC marked tumor epithelial, Basal,
Transitional, Ciliated, AT1, and AT2 cells, respectively,
across the datasets (Fig. 2c). However, we also saw that
some marker genes were preferentially detected in fresh
versus FFPE transcriptomes, such as MUC5B and SFTPC
in fresh or FFPE-derived tumor epithelial cells, respec-
tively. This was likely due to the dierent absolute num-
bers of epithelial cell transcriptomes across fresh versus
FFPE tissues (Fig. 2b), but also due to higher epithelial cell
type diversity in the FFPE tissue blocks of patients P078
and P079 (Fig. 2d). The latter result suggests that normal
epithelial cells are most easily lost during fresh tissue
dissociation; however, we cannot rule out compositional
bias in the fresh versus FFPE tissue specimens used.
Detection of immune cell marker genes and cell types
was more even across fresh and FFPE tissue specimens
(Fig. 2e, f), while many more broblasts were detected
among the stromal cell types in FFPE tissue-derived
sequence libraries (Fig. 2g, h). Cell type label transfer to
the cryopreserved tissue samples showed often skewed
representation of cell types (Fig. S3a, b), in particular for
immune cells such as regulatory or CD8+ T cells that are
dened by few key marker genes which could not be
reliably detected in the frozen samples.
3.3 Cell trait heterogeneity in FFPE versus fresh
samples
We asked whether archival FFPE tissue analysis allows
characterization of cell types to a similar extent as fresh
tissue single-cell analysis. We performed this analysis on
gene signature level, taking into account sets of pathway
target gene signatures [13] and relevant Hallmark gene sets
[15]. Technological dierences in sample preparation
workows resulted in dierences of normalized gene
expression scores between the fresh tissue and the FFPE-
derived libraries (Fig. S4a, b), where, for instance, the
MAPK target gene and feedback regulator DUSP4 was
found higher expressed in fresh tissue libraries, whereas
the MAPK-regulated methyltransferase KMT5A scored
higher in FFPE-derived libraries (Fig. 3a). Despite these
dierences on individual gene expression levels, we
found high correlations between inferred pathway activ-
ities (Fig. 3b) and tumor-related cell traits (Fig. 3c) from
fresh and FFPE samples throughout. In line with prior
studies, we found high TGFβ activity in broblasts, and
this cell type also had the highest score for mesenchymal
traits judged by activity of the EMT and Angiogenesis
Hallmark signature. As expected, highest activity of the
G2/M checkpoint and E2F targets Hallmark signatures
was found in proliferating T cells and macrophages,
which correlated with high MAPK activity in proliferat-
ing T cells. We detected high activity of TNFα and NFκB
signaling in monocytes, and high scores of the Interferon
Gamma Response Hallmark signature in monocytes
and tumor-associated monocyte-derived macrophages
(Mo-Macs), while tissue-resident alveolar macrophages
were characterized by high scores of the Oxidative
Phosphorylation Hallmark signature. In summary, corre-
lations between signature activities derived from fresh or
FFPE tissue samples were strong on the cell type level with
Pearson correlation coecients mostly larger than 0.9 and
signicances mostly smaller than p = 0.001 across all signa-
tures analyzed, showing that biology-related cell type char-
acteristics can be inferred from fresh and archival FFPE
tissue specimens alike. Scores were lower when comparing
correlations of cell type traits between fresh versus frozen
libraries, in particular for pathways with smaller variation
between cell types (Fig. S3c).
3.4 Assessment of clinical features from FFPE
versus fresh tissue single-cell transcriptomes
We investigated whether patient-specic dierences in
tumor and microenvironmental characteristics can be
Cellular Oncology
Fig. 2 Cell type diversity in fresh versus FFPE tissue single-cell
analysis. aUMAPs based on the top 10 principal components of
fresh or FFPE single-cell transcriptomes, as indicated, color-coded by
main cell type. bRelative proportions of epithelial, immune or stromal
cells, compared between fresh and FFPE-derived single-cell libraries,
and as quantied by IHC staining of FFPE tissue sections, see Fig. S2
for representative images, paired t-test per main cell type. c,dAnalysis
of epithelial transcriptomes. cEpithelial marker gene expression per
cell type in fresh or FFPE tissue-derived libraries. dUMAP and
absolute cell numbers, color-coded by cell type. e,fAnalysis of
immune cell transcriptomes. eImmune marker gene expression per
cell type in fresh or FFPE tissue-derived libraries. fUMAP and
absolute cell numbers, color-coded by cell type. g,hAnalysis of
stromal cell transcriptomes. gStromal marker gene expression per
cell type in fresh or FFPE tissue-derived libraries. hUMAP and
absolute cell numbers, color-coded by cell type
Cellular Oncology
Fig. 3 Cell trait quantication in fresh versus FFPE single-cell ana-
lysis. aExpression of selected PROGENy signature MAPK target
genes in fresh tissue and FFPE libraries. Expression was normalized
to scale across libraries. Data shown for cells assigned as
Proliferating T cells (high MAPK activity, see Fig. 3b) or Ciliated
cells (low MAPK activity, see Fig. 3b). bCorrelations of PROGENy
pathway scores between FFPE and fresh tissue gene expression per
cell type, Pearson correlation coecient and p-value indicated per
pathway. cCorrelations of selected biologically-relevant Hallmark
gene signatures between FFPE and fresh tissue gene expression per
cell type, Pearson correlation coecient and p-value indicated per
gene signature
Cellular Oncology
determined in fresh and FFPE tissue samples to the same
degree. The histological grade of tumors (Fig. 4a) perfectly
correlated with recently identied gene signatures of tumor
dierentiation [16] (Fig. 4b), regardless of whether the
signatures were called from fresh or FFPE tissue-derived
epithelial cell transcriptomes. The lowest histological
tumor grade in patient P079 was further associated with
higher activity of CAF-related pathways in broblasts [17]
(Fig. 4c) and high expression of typical CAF marker genes
on the single-cell mRNA level (Fig. 4d). With regard to
oncogenic pathways, tumor epithelial cells of patient P079
scored highest for MAPK and WNT target expression,
while P075 had the highest values of p53 pathway activity,
in both fresh and FFPE tissue samples (Fig. S5a). In con-
trast, no correlation was observed for EGFR pathway
activity. We inferred similar copy number proles in
tumor epithelial cells from both fresh and FFPE tissue-
derived transcriptomes (Fig. S5b).
Histologically, tumors diered in expression of the
immune evasion marker PD-L1 in tumor cells (Fig. 4e)
which was similarly observed on the RNA level in fresh as
well as FFPE tissue-derived tumor epithelial cell transcrip-
tomes (Fig. 4f). In both fresh and FFPE transcriptome data,
high PD-L1 expression in patient P079 correlated with
high expression of the Interferon Gamma Response
Hallmark signature in tumor cells and the tumor microen-
vironment, in particular in Mo-Macs [18] (Fig. 4g). In Mo-
Macs, this feature was accompanied by high TNFα, NFκB
and JAK/STAT pathway activity and high expression
scores of the Inammatory Response Hallmark signature
in patient P079 (Fig. 4h) [19,20]. Among T cells, high
PD-L1 expression in P079 correlated with low cytotoxicity
and high naiveness scores in CD4+ T cells, and high scores
for exhaustion in CD8+ T cells (Fig. 4i) [21]. In summary,
these correlations are in agreement with previous analysis
of the lung cancer microenvironment [16,22], and indicate
that cell type characteristics relevant for clinical stratica-
tion can be retrieved faithfully from the FFPE snRNA-seq
approach.
4 Discussion
Single-cell sequencing of surgical tissues has uncovered
clinically relevant information on the patient and cohort
levels. However, the use of fresh tissues comes with ser-
ious limitations in the clinic, such as the requirement for
a rapid tissue handling pipeline. Recent technological
advances allow single-nucleus sequencing from frozen
and FFPE tissue, which promises broader availability of
tissue, including routine pathology specimens. Here, we
present a systematic comparison of fresh, frozen and
FFPE single-cell analysis of clinical tissue. We nd that
FFPE tissue robustly preserved clinically relevant informa-
tion on cell types and patient characteristics comparable to
fresh tissue.
Single-cell analysis of fresh solid tissue has been the
gold standard to survey cell type composition of human
tissues in health and disease [23–25]. Generally, the pro-
cedure relies on cell dissociation, most often by proteases,
which can result in disproportionate release of cell types
and preparation-related transcriptome artifacts, although
procedures have been developed to minimize such eects.
Moreover, fresh tissue requires timely processing, which
can complicate standardization and result in batch eects.
Potentially, analysis of transcriptionally inert nuclei from
frozen or FFPE tissues could result in more even tran-
scriptome representation across cell types in the absence
of dissociation artifacts. We have here tested available
procedures side-by-side on tissues of the same origin
and found that single-cell sequencing of FFPE tissue
results in transcriptomes representing a higher diversity
of cell types compared to analysis of fresh tissue cell
suspensions, an equally good representation of biological
and clinical transcriptome features on a signature level,
and lower induction of stress-related gene expression. In
contrast, frozen tissue analysis was less reliable in our
hands, probably due to handling of unxed nuclei suspen-
sions that are prone to RNA degradation or RNA
diusion.
In the clinic, fresh tissue single-cell analysis allowed for
prospective collection of samples only. Importantly, single
nucleus transcriptomics of FFPE tissues allow for retro-
spective analysis of cohorts that are annotated with long-
term clinical follow-up data such as therapy response,
relapse, metastasis, and patient survival. The FFPE blocks
under analysis here were stored under standard archival
conditions at room temperature for 4–5 months. It remains
to be determined whether older FFPE blocks with longer
follow-up periods perform equally well, as RNA quality
could potentially deteriorate over time even in the xed
and embedded state. Moreover, it needs to be investigated
to which degree prolonged time until xation and tissue
autolysis in large surgery specimens impair snRNA-seq
analysis.
Pathology review is usually performed using FFPE sec-
tions, as these yield high-quality histology and immuno-
histochemistry information. Therefore, FFPE blocks can be
annotated regionally with high condence, unlike intra-
operative fresh tissue samples. Thus, FFPE blocks allow
analysis of cell composition of circumscribed areas of
interest. We envision that careful selection of tissue regions
in pathology, in conjunction with multiplexing techniques
for cost-eectiveness, will bring single-cell analysis aiding
therapy prediction or prognosis a step closer to clinical
routine for cancer patients.
Cellular Oncology
Fig. 4 Quantication of clinically relevant gene expression patterns
in fresh versus FFPE single-cell analysis. aHematoxylin and eosin
stained FFPE tumor sections of patients P075, P078, P079, showing
a predominant lepidic, acinar or solid growth pattern, respectively,
corresponding to well-dierentiated (G1), moderately-dierentiated
(G2) or poorly-dierentiated (G3) tumor morphology and grade, as
indicated. bCorrelation between FFPE and fresh tissue of Alveolar/
club-like and Undierentiated tumor cell signature scores in tumor
epithelial cells. cCorrelation between FFPE and fresh tissue of
PROGENy pathway scores in broblasts. dCorrelation between
FFPE and fresh tissue of CAF marker expression in broblasts.
eImmunohistochemistry on FFPE tumor section for PD-L1 expres-
sion. TPS: Tumor proportion score. fCorrelation between FFPE and
fresh tissue of PD-L1 gene expression in tumor epithelial cells.
gCorrelation between FFPE and fresh tissue of Interferon Gamma
Response Hallmark gene signature scores in various cell types, as
indicated. Mo-Macs = Monocyte-derived macrophages. hCorrelation
between FFPE and fresh tissue of clinically and biologically relevant
signature scores in Mo-Macs, as indicated. iCorrelation between
FFPE and fresh tissue of functional cell state scores in CD4+ and
CD8+ T cells, as indicated
Cellular Oncology
Supplementary Information The online version contains supplemen-
tary material available at https://doi.org/10.1007/s13402-024-00922-0.
Acknowledgements We thank Caroline Braeuning at the BIH Core
Unit Genomics for excellent technical support. Figure 1a was created
with BioRender.com.
Author contributions Conceptualization: PB, SF, MM; Investigation:
AT, PB, JK; Data Curation: MiM, DB, PB; Formal Analysis: PB;
Validation: PB; Visualization: PB, MM; Writing – Original Draft: PB,
MM, AT; Resources: MM, PB, CS, DH.
Funding Open Access funding enabled and organized by Projekt
DEAL. PB is participant in the BIH Charité Clinician Scientist
Program funded by the Charité – Universitätsmedizin Berlin, and
the Berlin Institute of Health at Charité (BIH).
Data availability Single-cell gene expression data generated in this
study and analysis scripts are available on Zenodo at https://doi.org/
10.5281/zenodo.7852154.
Declarations
Conflict of interest The authors declare that they have no conict of
interest.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format,
as long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons licence, and indicate
if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons licence and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this licence, visit http://creativecommons.
org/licenses/by/4.0/.
References
1. A. Maynard, C.E. McCoach, J.K. Rotow, L. Harris, F. Haderk,
D.L. Kerr et al., Therapy-induced evolution of human lung
cancer revealed by single-cell RNA sequencing. Cell 182(5),
1232–1251.e22 (2020)
2. B. Chen, C.R. Scurrah, E.T. McKinley, A.J. Simmons, M.A.
Ramirez-Solano, X. Zhu et al., Dierential pre-malignant pro-
grams and microenvironment chart distinct paths to malignancy
in human colorectal polyps. Cell 184(26), 6262–6280.e26
(2021)
3. Y.C. Cohen, M. Zada, S.Y. Wang, C. Bornstein, E. David, A.
Moshe et al., Identication of resistance pathways and therapeu-
tic targets in relapsed multiple myeloma patients through single-
cell sequencing. Nat. Med. 27(3), 491–503 (2021)
4. L. Ma, M.O. Hernandez, Y. Zhao, M. Mehta, B. Tran,
M. Kelly et al., Tumor cell biodiversity drives microenviron-
mental reprogramming in liver cancer. Cancer Cell 36(4),
418–430.e6 (2019)
5. C.J. Hanley, S. Waise, M.J. Ellis, M.A. Lopez, W.Y. Pun, J.
Taylor et al., Single-cell analysis reveals prognostic broblast
subpopulations linked to molecular and immunological subtypes
of lung cancer. Nat. Commun. 14(1), 387 (2023)
6. C.A. Stewart, C.M. Gay, Y. Xi, S. Sivajothi, V. Sivakamasundari,
J. Fujimoto et al., Single-cell analyses reveal increased intratu-
moral heterogeneity after the onset of therapy resistance in small-
cell lung cancer. Nat. Cancer 1, 423–436 (2020)
7. M. Slyper, C.B.M. Porter, O. Ashenberg, J. Waldman, E.
Drokhlyansky, I. Wakiro et al., A single-cell and single-nucleus
RNA-Seq toolbox for fresh and frozen human tumors. Nat. Med.
26(5), 792–802 (2020)
8. A. Janesick, R. Shelansky, A.D. Gottscho, F. Wagner,
M. Rouault, G. Beliako et al., High resolution mapping of
the breast cancer tumor microenvironment using integrated sin-
gle cell, spatial and in situ analysis of FFPE tissue. bioRxiv
2022.10.06.510405 (2022). https://doi.org/10.1101/2022.10.06.
510405
9. A.F. Vallejo, K. Harvey, T. Wang, K. Wise, L.M. Butler, J. Polo
et al., snPATHO-seq: unlocking the FFPE archives for single
nucleus RNA proling. bioRxiv 2022.08.23.505054 (2022).
https://doi.org/10.1101/2022.08.23.505054
10. H. Chung, A. Melnikov, C. McCabe, E. Drokhlyansky, V.W. N.,
E.M. Magee et al., SnFFPE-Seq: towards scalable single nucleus
RNA-Seq of formalin-xed paran-embedded (FFPE) tissue.
bioRxiv 2022.08.25.505257 (2022). https://doi.org/10.1101/
2022.08.25.505257
11. Y. Hao, S. Hao, E. Andersen-Nissen, W.M. Mauck, S. Zheng, A.
Butler et al., Integrated analysis of multimodal single-cell data.
Cell 184(13), 3573–3587.e29 (2021)
12. M.D. Young, S. Behjati, SoupX removes ambient RNA contam-
ination from droplet-based single-cell RNA sequencing data.
Gigascience 9(12), giaa151 (2020)
13. M. Schubert, B. Klinger, M. Klünemann, A. Sieber, F. Uhlitz,
S. Sauer et al., Perturbation-response genes reveal signaling
footprints in cancer gene expression. Nat. Commun. 9(1),
20 (2018)
14. A. Neuschulz, O. Bakina, V. Badillo-Lisakowski, P. Olivares-
Chauvet, T. Conrad, M. Gotthardt et al., A single-cell RNA
labeling strategy for measuring stress response upon tissue dis-
sociation. Mol. Syst. Biol. 19(2), e11147 (2023)
15. A. Liberzon, C. Birger, H. Thorvaldsdóttir, M. Ghandi, J.P.
Mesirov, P. Tamayo, The molecular signatures database hall-
mark gene set collection. Cell Syst. 1, 417–425 (2015)
16. P. Bischo, A. Trinks, B. Obermayer, J.P. Pett, J. Wiederspahn,
F. Uhlitz et al., Single-cell RNA sequencing reveals distinct
tumor microenvironmental patterns in lung adenocarcinoma.
Oncogene 40(50), 6748–6758 (2021)
17. G.J. Yoshida, Regulation of heterogeneous cancer-associated
broblasts: the molecular pathology of activated signaling path-
ways. J. Exp. Clin. Cancer Res. 39(1), 112 (2020)
18. F. Castro, A.P. Cardoso, R.M. Gonçalves, K. Serre, M.J.
Oliveira, Interferon-gamma at the crossroads of tumor immune
surveillance or evasion. Front. Immunol. 9, 847 (2018)
19. M.G. Dorrington, I.D.C. Fraser, NF-κB signaling in macro-
phages: dynamics, crosstalk, and signal integration. Front.
Immunol. 10, 705 (2019)
20. M. Orecchioni, Y. Ghosheh, A.B. Pramod, K. Ley, Macrophage
polarization: dierent gene signatures in M1(Lps+) vs. classi-
cally and M2(LPS−) vs. alternatively activated macrophages.
Front. Immunol. 10, 1084 (2019)
21. X. Guo, Y. Zhang, L. Zheng, C. Zheng, J. Song, Q. Zhang et al.,
Global characterization of T cells in non-small-cell lung cancer
by single-cell sequencing. Nat. Med. 24(7), 978–985 (2018)
22. A.M. Gocher, C.J. Workman, D.A.A. Vignali, Interferon-γ:
teammate or opponent in the tumour microenvironment? Nat.
Rev. Immunol. 22(3), 158–172 (2021)
23. R.C. Jones, J. Karkanias, M.A. Krasnow, A.O. Pisco, S.R. Quake,
J. Salzman et al., The Tabula Sapiens: a multiple-organ, single-cell
Cellular Oncology
transcriptomic atlas of humans. Science 376(6594), eabl4896
(2022)
24. J.M. Chan, Á. Quintanal-Villalonga, V.R. Gao, Y. Xie, V. Allaj,
O. Chaudhary et al., Signatures of plasticity, metastasis, and
immunosuppression in an atlas of human small cell lung cancer.
Cancer Cell 39(11), 1479–1496.e18 (2021)
25. K.J. Travaglini, A.N. Nabhan, L. Penland, R. Sinha, A. Gillich,
R.V. Sit et al., A molecular cell atlas of the human lung from
single-cell RNA sequencing. Nature 587(7835), 619–625 (2020)
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional aliations.
Cellular Oncology
