Comprehensive characterization of the multiple myeloma immune

microenvironment using integrated scRNA-seq, CyTOF, and CITE-seq

analysis

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

Lijun Yao1, Reyka G. Jayasinghe1, Brian Lee2, Swati Sharma Bhasin3, William Pilcher3, Deon Bryant Doxie3, Edgar

Gonzalez-Kozlova2, Surendra Dasari4, Mark A. Fiala1, Yered Pita-Juarez6, Michael Strausbauch4, Geoffrey Kelly2,

Beena E. Thomas3, Shaji K. Kumar4, Hearn Jay Cho2,5, Emilie Anderson4, Michael C. Wendl1, Travis Dawson2,

Darwin D’souza2, Stephen T. Oh1, Giulia Cheloni6, Ying Li4, John F. DiPersio1, Adeeb H. Rahman2, Kavita M.

Dhodapkar3, Seunghee Kim-Schulze2, Ravi Vij1, Ioannis S. Vlachos6, Shaadi Mehr5, Mark Hamilton5, Daniel Auclair5,

Taxiarchis Kourelis4*, David Avigan6*, Madhav V. Dhodapkar3*, Sacha Gnjatic2*, Manoj K. Bhasin3*, Li Ding1*

*corresponding authors

1 Washington University School of Medicine, Saint Louis, MO

2 Icahn School of Medicine at Mt. Sinai, New York, NY

3 Emory University, Atlanta, GA

4 Mayo Clinic, Rochester, MN

5 Multiple Myeloma Research Foundation, Norwalk, CT

6 Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA

Running title:

Examining the MM TME with complementary single-cell methods

Correspondence to:

Li Ding, Ph.D, Washington University School of Medicine, Saint Louis, MO, 63110. Email: [email protected]. Phone:

3142861848

Competing interests

The authors declare that they have no competing interests.

Keywords

Multiple Myeloma; Bone marrow; Immune cells; Single-cell technologies

1
Abstract

As part of the Multiple Myeloma Research Foundation (MMRF) immune atlas pilot project, we

compared immune cells of Multiple Myeloma (MM) bone marrow samples from 18 patients

assessed by single-cell RNA-seq (scRNA-seq), mass cytometry (CyTOF), and Cellular Indexing

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

of Transcriptomes and Epitopes by Sequencing (CITE-seq) to understand the concordance of

measurements among single-cell techniques. Cell type abundances are relatively consistent

across the three approaches, while variations are observed in T cells, macrophages, and

monocytes. Concordance and correlation analysis of cell type marker gene expression across

different modalities highlighted the importance of choosing cell type marker genes best suited to

particular modalities. By integrating data from these three assays, we found International

Staging System (ISS) stage 3 patients exhibited decreased CD4+ T/ CD8+ T cells ratio.

Moreover, we observed upregulation of RAC2 and PSMB9, in NK cells of fast progressors (FP)

compared to those of non-progressors (NP), as revealed by both scRNA-seq and CITE-seq

RNA measurement. This detailed examination of the immune microenvironment in MM using

multiple single cell technologies revealed markers associated with MM rapid progression which

will be further characterized by the full-scale immune atlas project.

Significance

scRNA-seq, CyTOF, and CITE-seq are increasingly used for evaluating cellular heterogeneity.

Understanding their concordances is of great interest. To date, this study is the most

comprehensive examination of the measurement of the immune microenvironment in MM using

the three techniques. Moreover, we identified markers predicted to be significantly associated

with MM rapid progression.

2
Introduction

Single-cell sequencing technologies offer advantages over traditional bulk methods in cancer

genomics research for evaluating cellular heterogeneity and investigating evolution of cellular

subpopulations between the tumor and its microenvironment. For example, single-cell methods

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

have been extensively applied to Multiple Myeloma (MM), a highly heterogeneous disease

marked by uncontrolled clonal expansion of plasma cells. Single-cell RNA sequencing (scRNA-

seq) has been used to examine tumor and immune cell populations [1,2] and mass cytometry

(CyTOF) to evaluate the impact of drugs on immune populations in MM [3]. The third technology,

cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq), is a more recent,

multimodal approach with simultaneous quantification of single-cell transcriptomes and surface

proteins. All three approaches enable identification of cell types, cell states, and characterization

of cellular heterogeneity at transcriptomic and/or protein levels. Consequently, understanding

their concordances across technologies is of great practical interest.

In addition, the bone marrow microenvironment plays an important role in the evolution of

premalignant MM, MM progression and treatment response. Single-cell transcriptomics analysis

of the tumor microenvironment revealed compositional alterations begin at the Monoclonal

gammopathy of undetermined significance (MGUS) stage, including enrichment of T cells, NK

cells and CD16+ monocytes [2]. Specifically, the percentage of CD4+ T cells was significantly

reduced in bone marrow of MM patients, leading to altered CD4+ T/CD8+ T ratio [4]. When

comparing the clinical status, the ratio decreased in International Staging System (ISS) stage 3

patients compared to stage 1 patients [5]. With respect to treatment, the proportion of CD3+ T

cells was lower in treated patients compared to chemo-naive MM patients [6]. Further work is

needed to expand initial findings using various assays and reveal candidate markers for

characterizing clinical features of MM patients and optimizing treatment.

3
Combining the timeliness of the technology concordance question with furtherance of MM

research, we subjected bone marrow samples from 18 MM patients to scRNA-seq, CyTOF, and

CITE-seq, examining the similarities across the aforementioned single cell techniques. We used

the results to investigate the relationship between immune population compositional alterations

and disease stages and revealed a set of markers associated with MM rapid progression.

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

Materials and Methods

Ethics approval and consent to participate

All procedures performed in studies involving human participants were in accordance with the

ethical standards of the MMRF research committee. These samples provided by MMRF were all

from the MMRF’s CoMMpass clinical trial (NCT NCT01454297). Written informed patient

consent was obtained from all patients for the collection and analysis of their samples by the

MMRF. The CoMMpass study was conducted in accordance with recognized ethical guidelines

in the US and EU. The Institutional Review Board at each participating center approved the

study protocol.

Ammonium-chloride-potassium (ACK) Lysis of Bone Marrow Aspirates (BMA)

BMA samples obtained from subjects enrolled in the MMRF CoMMpass study

(NCT01454297). Any blood clots were removed from BMA samples via passage through

70 mM cell strainer. BMA samples were aliquoted into 5mL aliquots in 50mL conical

tubes and 45mL of 22mM-filtered ACK lysing buffer (155mM Ammonium Chloride/10mM

Potassium Bicarbonate/0.1mM EDTA/pH7.4) was added to each 5 mL aliquot and the

tune gently inverted several times to mix. Tubes were then centrifuged at 400xg for 5

minutes. The supernatant was removed and the cell pellet resuspended with 5 mL of

RPMI-1640 and transferred to a clean tube. All aliquots of ACK-lysed BMA aliquots were

4
combined into 1x 50 mL tube, the volume adjusted to 50 mL with RPMI-1640. The cells

were then mixed by gentle inversion and the tube centrifuged at 400xg for 5 minutes.

The supernatant was then removed by aspiration. Depending on the size of the BMA cell

pellet, the cell pellet resuspended in 1-10mL of EasySep buffer (Phosphate-buffered

saline (PBS) containing 2%FBS (v/v) and 1mM EDTA (PBS/FCS/EDTA buffer). 25mL of

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

cell suspension was removed for cell counting.

Isolation of CD138-positive and CD138-negative cells from BMA

CD138-negative immune cell mononuclear (CD138-) cells in bone marrow aspirates

from subjects enrolled in the MMRF CoMMpass study (NCT NCT01454297) were

isolated via negative selection from CD138-positive (CD138+) myeloma cells using the

EasySepTM immunomagnetic bead technology (EasySepTM Human CD138 Positive

Selection Kit: Stem Cell Technologies) in accordance with the manufacturers protocol.

Briefly, 100x106 cell/mL bone marrow mononuclear cells (MNC) in a sterile 17x100mm

(14mL) tube were gently mixed and incubated with 100mL/mL CD138 selection antibody

cocktail for 15 minutes at room temperature. 50mL/mL of EasySep magnetic

nanoparticles was then added to the cell suspension, gently mixed, and incubated for a

further 10 minutes at room temperature. The volume of the cell suspension was then

adjusted to 8mL with phosphate-buffered saline (PBS) containing 2%FBS (v/v) and 1mM

EDTA (PBS/FCS/EDTA buffer) and the cell suspension mixed by gentle pipetting (2-3x).

The tube was then placed in the magnetic separator. After 5 minutes incubation at room

temperature, the magnet and tube were carefully inverted to pour off the supernatant

into a sterile 50mL conical tube. This supernatant contains the heterogeneous CD138-

negative immune cell mononuclear population (MNC). The tube was then removed from

the magnet and an additional 8mL of PBS/FCS/EDTA added, gently mixed, and returned

to the magnetic separator. Again, after 5 minutes incubation in the magnetic separator,

5
the tube and magnet were carefully inverted to pour of the supernatant into the 50mL

collection tube. This PBS/FCS/EDTA ‘wash’ step was repeated once more resulting in

~24mL suspension of CD138-negative bone marrow MNC cells. CD138- MNC cells were

then pelleted by centrifugation at 400xg for 5 minutes and the supernatant removed by

aspiration. The CD138-MNC cell pellet was resuspended in freezing medium

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

(90%FCS/10%DMSO) at a concentration of ~8-10x106 cells/mL prior to cryogenic

storage in liquid nitrogen.

Processing of BMMC and library prep from MMRF CoMMpass study for scRNA-seq at

WUSTL

WUSTL Cell Thawing: Multiple Myeloma bone marrow mononuclear cells (BMMC) aliquots were

thawed in 37⁰C water bath. Cells were then pelleted by centrifugation at 300g for 5 min and all

supernatant was removed. To prepare cells for the Miltenyi Dead Cell Removal Kit, cells were

resuspended in 100 uL of beads and incubated at room temperature for 15 minutes. Dead cells

were depleted using the autoMACS®Pro Separator. Live cells were pelleted by centrifugation at

450g for 5 minutes. Cells were finally resuspended in ice cold phosphate buffer saline (PBS)

and 0.5% BSA and loaded onto the 10x Genomics Chromium Controller and using the

Chromium Next GEM Single Cell 3’ GEM, Library & Gel Bead Kit v3.3. Utilizing the 10x

Genomics Chromium Single Cell 3’v3 Library Kit and Chromium instrument, approximately

16,500 to 20,000 cells were partitioned into nanoliter droplets to achieve single cell resolution

for a maximum of 10,000 individual cells per sample. The resulting cDNA was tagged with a

common 16nt cell barcode and 10nt Unique Molecular Identifier during the RT reaction. Full

length cDNA from poly-A mRNA transcripts was enzymatically fragmented and size selected to

optimize the cDNA amplicon size (approximately 400 bp) for library construction (10x

Genomics). The concentration of the 10x single cell library was accurately determined through

qPCR (Kapa Biosystems) to produce cluster counts appropriate for the HiSeq 4000 or NovaSeq

6
6000 platform (Illumina). 26x98bp (3'v2 libraries) sequence data were generated targeting

between 25K-50K read pairs/cell, which provided digital gene expression profiles for each

individual cell.

ISMMS BMMC processing differences from WUSTL: BMMC aliquots were partially

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

thawed in 37⁰C water bath. 1mL of warm thawing media (RPMI + 10% FBS) was added to

the partially thawed BMMC aliquot and the entire volume was transferred to a 15 mL conical

tube containing 10 mL of warm thawing media. The empty BMMC tube was rinsed with

another 1 mL of thawing media which was then also transferred to the 15 mL conical tube.

Cells were processed using the EasySep Dead Cell Removal (Annexin V) Kit (StemCell

Technologies, Cat# 17899).

scRNA-seq data quantification preprocessing

For single cell RNA-seq analysis, the proprietary software tool Cell Ranger v3.0.0 from 10x

Genomics was used for de-multiplexing sequence data into FASTQ files, aligning reads to the

human genome (GRCh38), and generating gene-by-cell UMI count matrix.

Seurat v3.0.0 [7],[8] was used for all subsequent analysis. First, a series of quality filters was

applied to the data to remove those barcodes which fell into any one of these categories

recommended by Seurat: too few total transcript counts (< 300); possible debris with too few

genes expressed (< 200) and too few UMIs (< 1,000); possible more than one cell with too

many genes expressed (> 50,000) and too many UMIs (> 10,000); possible dead cell or a sign

of cellular stress and apoptosis with too high proportion of mitochondrial gene expression over

the total transcript counts (> 20%). Finally, predicted doublets were also removed using scrublet

V0.2.3.

7
We constructed a Seurat object using the unfiltered feature-barcode matrix for each sample.

Each sample was scaled and normalized using Seurat’s ‘SCTransform’ function to correct for

batch effects (with parameters: vars.to.regress = c("nCount_RNA", "percent.mito"),

return.only.var.genes = F).

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

scRNA-seq cell type annotation

Cell types were assigned to each cluster by manually reviewing the expression of marker genes.

The marker genes for main cell types were CD79A, CD79B, MS4A1 (B cells); CD8A, CD8B,

CD7, CD3E (CD8+ T cells); CD4, IL7R, CD7, CD3E (CD4+ T cells); NKG7, GNLY, KLRD1,

NCAM1 (NK cells); MZB1, SDC1, IGHG1 (Plasma cells); CLEC4C, IL3RA, IRF8, GZMB

(Dendritic cells); FCGR3A (Macrophages); CD14, LYZ, S100A8, S100A9 (Monocytes); AZU1,

ELANE, MPO (Neutrophils); COL1A1, COL3A1, TNC, S100A4 (Fibroblasts); and AHSP1, HBA,

HBB (Erythrocytes). Detailed cell type markers are listed in Supplementary Table S1A. All cells

that were labeled as erythrocytes and plasma cells were removed from subsequent analysis.

Processing of BMMC from MMRF CoMMpass study for CITE-seq

Samples were thawed in the water bath at 37°C for 2-3 min and the cell concentration, viability

were determined using a Bio-Rad T20 Cell Counter (Cat# 145-0102). Samples were blocked by

incubation with TruStain fcX (Biolegend, cat# 422301) in a 50 μL cell labeling buffer. Next,

samples were labeled with Total-seq antibodies (Biolegend, Supplementary Table S1B) for 30

minutes. Cells were washed and resuspended to obtain a cell concentration of 700-1,200

cells/μl and gently pipette mix using a regular-bore pipette tip until a single cell suspension is

achieved. We then proceed immediately to Single cell Gene Expression Library (3’GEX)

construction using 10X Chromium Single Cell 3' Reagent Kits v3 (Cat# 1000075) and Chromium

i7 Sample Index Plate with Barcoding technology for Cell Surface Protein. For each sample,

8
5000 cells were injected for CITE-Seq. The libraries were sequenced on NovaSeq S4 platform

in pair end sequencing and a single index with at least 50,000 read pairs per cell.

CITE-seq data quantification preprocessing

We used Cell Ranger to demultiplex, map to the human reference genome (grch38), and count

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

UMIs in the mRNA libraries, and CITE-seq-Count to count UMIs in the ADT libraries. We filtered

out cells with more than 10% UMIs from mitochondrially-encoded genes or less than 1,200

mRNA UMIs in total. We then constructed a Seurat object using the feature-barcode matrix for

each sample (Seurat v3.0.0). Each sample was scaled and normalized using Seurat’s

‘SCTransform’ function to correct for batch effects (with parameters: vars.to.regress =

c("nCount_RNA", "percent.mito"), return.only.var.genes = F). Next, the protein expression

levels were added to the Seurat object, followed by normalization and scaling for ADT assay.

CITE-seq data multi-modal integration and cell type annotation

Using Citefuse v1.2.0, expression was normalized by function normaliseExprs(sce,

altExp_name = "ADT", transform = "log"). We then integrated RNA and ADT matrix by an

integration algorithm called similarity network fusion (SNF) and clustered cells by louvain

clustering. Then, cell types were assigned to each cluster by manually reviewing the expression

of marker genes at RNA levels (same as scRNA-seq, Supplementary Table S1A) and ADT

levels (if available). All cells that were labeled as erythrocytes and plasma cells were removed

from subsequent analysis.

Processing of BMMC from MMRF CoMMpass study for CyTOF at ISMMS

BMMC aliquots were thawed in a 37⁰C water bath and immediately transferred into RPMI+10%

FBS. Cells were pelleted by centrifugation at 300g for 5 minutes and all supernatant was

9
removed. Cells were then incubated for 20 minutes in a 37⁰C water bath with Cell-ID Rh103

Intercalator (Fluidigm, Cat# 201103A) to label non-viable cells. Samples were then blocked with

Fc receptor blocking solution (Biolegend, Cat# 422302) and stained with a cocktail of surface

antibodies for 30 minutes on ice. All antibodies were either conjugated in-house using

Fluidigm's ×8 polymer conjugation kits or purchased commercially from Fluidigm. Next, samples

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

were fixed and barcoded using Fluidigm’s 20-Plex Pd barcoding kit (Cat# 201060) and pooled

into a single tube. The pooled sample was then fixed and permeabilized using BD's

Cytofix/Cytoperm Fixation/Permeabilization Kit (Cat# 554714), blocked with heparin at a

concentration of 100U/ml to prevent non-specific staining of eosinophils and stained with a

cocktail of intracellular antibodies. Finally, the sample was re-fixed with freshly diluted 2.4%

formaldehyde in PBS containing 0.02% saponin and Cell-ID Intercalator-Ir (Fluidigm, Cat#

201192A) to label nucleated cells. The sample was then stored as a pellet in PBS until

acquisition.

Immediately prior to acquisition, the pooled sample was washed with Cell Staining Buffer and

Cell Acquisition Solution (Fluidigm, Cat# 201240) and resuspended in Cell Acquisition Solution

at a concentration of 1 million cells per ml containing a 1:20 dilution of EQ normalization beads

(Fluidigm, Cat# 201078). The sample was acquired on the Fluidigm Helios mass cytometer

using the wide bore injector configuration at an acquisition speed of < 400 cells per second.

Processing of BMMC from MMRF CoMMpass study for CyTOF at Mayo

BMMC aliquots were thawed in a 37⁰C water bath and immediately transferred into 15ml tubes

and slowly diluted with 10 mL of pre warmed RPMI+10% FBS+25U/ml Benzonase((Sigma-

Aldrich; Catalog Number-E1014-5KU; 250U/mL). Cells were pelleted by centrifugation (all spins

at 500g for 5 minutes) and supernatant was removed. Cells were then incubated for 1hr in a

37⁰C water bath in 10mL of RPMI+10% FBS. Cells were counted and 3-4 million cells were

aliquoted into microfuge 2mL conical tubes, pelleted and washed 2X with 2mL CSB Maxpar®

10
Cell Staining Buffer (Fluidigm; Catalog number-201068; 500 mL) and resuspended in 300uL of

Cell-ID™ Cisplatin (Fluidigm; Catalog Number: 201064) 5 min/RT, to label dead cells.

Immediately quenched with 1.5 mL CSB, pelleted, and washed with CSB 2X.

For staining, the cell pellet was gently resuspended in 50uL CSB and the addition of an equal

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

volume of diluted surface antibody cocktail, for a final staining volume of 100uL. The staining

reaction was incubated on a rocker platform for 45min at RT. 1mL of CSB was used to wash

and pellet the cells 2X. Cell pellet was resuspended in the residual volume and then gently

resuspended in 500uL of 1X PBS. An equal volume of 4% PFA in PBS was added to fix cells for

a minimum of 20 minutes at a final concentration of 2% PFA in PBS. The sample was labeled

overnight at 4⁰C on a rocker platform with Cell-ID Intercalator-Ir (Fluidigm, Cat# 201192A) in

Maxpar Fix and Perm Buffer (Fluidigm; Catalog Number-201067; 100 mL) to label nucleated

cells.

The following day the sample was washed 1X with CSB (all cell pelleting performed at 800g for

5 minutes after fixation) and twice with Cell Acquisition Solution (Fluidigm, Cat# 201240). Final

resuspension was in Cell Acquisition Solution at a concentration of 0.7 million cells per mL

containing a 1:10 dilution of EQ normalization beads (Fluidigm, Cat# 201078). The sample was

acquired on the Fluidigm Helios mass cytometer using the wide bore injector configuration at a

targeted acquisition speed of 300 events per second. A cryopreserved specimen of 3-4 million

Ficoll enriched PBMC derived from a pool of 4 anonymous platelet donors was included with

every batch of MMRF samples [24]. This sample was treated and analyzed in parallel

throughout the entire experiment as a process control.

Processing of BMMC from MMRF CoMMpass study for CyTOF at Emory

11
BMMC aliquots were thawed in a 37⁰C water bath and immediately transferred into RPMI+10%

FBS. Cells were pelleted by centrifugation at 300g for 5 minutes and all supernatant was

removed. Cells were then incubated for 20 minutes in a 37⁰C incubator. Cells were pelleted by

centrifugation at 300g for 5 minutes and all supernatant was removed. Cells were resuspended

in PBS and incubated with cisplatin for 1 min (Fluidigm, Cat# 201195) to label non-viable cells.

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

Samples were washed with Maxpar cell staining buffer (Fluidigm, Cat# 201068) and stained

with a cocktail of surface antibodies for 15 minutes at room temperature. All antibodies were

either conjugated in-house using Fluidigm's X8 polymer conjugation kits or purchased

commercially from Fluidigm. Next, samples were washed and fixed and permed with TF

Fix/Perm and Perm/Wash Kit (BD Pharmigen, Cat# 51-9008100 and # 51-9008102) using

manufacturer’s recommendations. Permeabilized samples were incubated for 30 min in

Perm/Wash with a cocktail of intracellular antibodies. After washing and centrifugation at 800g

for 5 minutes, the sample was re-fixed with Maxpar Fix I buffer (Fluidigm, Cat# 201065) and

Cell-ID Intercalator-Ir (Fluidigm, Cat# 201192A) to label nucleated cells. The sample was then

stored as a pellet in PBS until acquisition. Immediately prior to acquisition, the sample was

washed with Cell Staining Buffer and Maxpar Water (Fluidigm, Cat# 201069) and resuspended

in Maxpar Water at a concentration of 1 million cells per ml containing a 1:10 dilution of EQ

normalization beads (Fluidigm, Cat# 201078). The sample was acquired on the Fluidigm Helios

mass cytometer using the HT injector configuration at an acquisition speed of < 500 cells per

second.

CyTOF data preprocessing

The resulting FCS files were normalized and concatenated using Fluidigm's CyTOF software

and then de-multiplexed using the Zunder lab single-cell debarocder

(https://github.com/zunderlab/single-cell-debarcoder). The FCS files were further cleaned on

Cytobank by removing EQ beads, low DNA debris, and gaussian multiplets. Barcoding

12
multiplets were also removed based on the Mahalanobis distance and barcode separation

distance parameters provided by the Zunder lab debarcoder.

CyTOF cell type annotation and expression normalization

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

Gating and data analysis were done using WUSTL Cytobank. Live, single cells are selected by

gating out cells/debris with outlier cisplatin and DNA intercalator staining. Cell populations were

determined based on gating of cell type marker expression. ISMMS: CD3+CD19-CD56-CD33-

(T cells); CD3-CD19-CD56-CD33-CD123+HLA_DR+CD11c+ (pDC); CD3-CD19+CD56-CD33-

(B cells); CD56+CD3-CD19-CD33- (NK cells); CD33+CD3-CD19-CD14+ (Monocytes);

CD33+CD3-CD19-CD14-CD16+ (Macrophages). Mayo: CD3+CD19- (T cells); CD3-

CD19+CD56- (B cells); CD56+CD3-CD16+HLADR-/CD56+CD3-CD16-CD123-CD11c- (NK

cells); CD3-CD19-CD20-CD14+ (Monocytes); CD3-CD19-CD20-CD14-CD16+ (Macrophages);

CD3-CD19-CD20-CD123+ (pDC); Emory: CD3+CD19- (T cells); CD3- CD19+ (B cells); CD3-

CD19-CD14+ (Monocytes); CD3-CD19-CD14-CD16+ (Macrophages). For T cell subtypes,

ISMMS and Mayo used the same gating strategy: CD4+CD8- (CD4+ T cells); CD8+CD4- (CD8+

T cells); CD4+CD8-CD25+CD127- (Treg); CD45RA+CCR7+ (Naive T cells); CD45RA+CCR7-

(EMRA T cells); CD45RA-CCR7+ (Central memory T cells), CD45RA-CCR7- (Effector memory

T cells), Emory: CD4+CD8- (CD4+ T cells); CD8+CD4- (CD8+ T cells); CD45RO-CCR7+ (Naive

T cells). Next, we performed t-SNE analysis for 18 samples from ISMMS. We used the scaled

expression of markers, including CD57, CD11c, Ki67, CD19, CD45RA, KLRG1, CD4, CD8,

ICOS, CD16, CD127, CD1c, CD123, CD66b, TIGIT, TIM3, CD27, PD-L1, CD33, CD14, CD56,

NKG2A, CD5, CD45RO, NKG2D, CD25, CCR7, CD3, Tbet, CD38, CD39, CD28, DNAM1, HLA-

DR, PD-1, GranzymeB, CD11b. For expression normalization in CyTOF analysis, we followed

instructions from Cytobank and used transformed ratios itself compared to its control, which is

13
the table’s minimum of median of channel (described here https://support.cytobank.org/hc/en-

us/articles/206147637-How-to-create-and-configure-a-Heatmap).

Bland-Altman analysis

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

R package Blandr (v0.5.3) was used to calculate mean difference and 95% confidence interval

in Bland-Altman analyses [25]. Parameter sig.level = 0.95.

Differential expression analysis

Differential expression analysis was performed using the default test (Wilcoxon Rank Sum test)

of function FindMarkers (from the Seurat package) with the specified parameters: min.pct=0.25,

logfc.threshold = 0.25, and only.pos = T.

Data availability statement

The sequence data generated in this study has been submitted to the NCBI BioProject

database PRJNA765009 (https://www.ncbi.nlm.nih.gov/bioproject/).

14
Results

Patient characteristics and overview of CD45+ immune cells measured by scRNA-seq,

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

CyTOF and CITE-seq

We used eighteen cryopreserved Multiple Myeloma (MM) samples of CD138-negative “immune

cell” fractions from patients enrolled in the Multiple Myeloma Research Foundation (MMRF)

CoMMpass study (NCT01454297). Nine were fast progressors (FP, progressed within 6 months)

and nine were non-progressors (NP, progressed >6 months but within 5 years) with patient

ages ranging from 37 to 83 years. Twelve patients were in the International Staging System

(ISS) stage III, eight underwent Autologous Stem Cell Transplantation (ASCT), eleven were

females and fifteen were Caucasions (Fig. 1A and Supplementary Table S1C). Each sample

was subjected to scRNA-seq, CyTOF, and CITE-seq at three different respective academic

research centers, namely Washington University in St. Louis (WUSTL), Icahn School of

Medicine at Mount Sinai (ISMMS), and Beth Israel Deaconess Medical Center (BIDMC). All

sites received aliquots from the same sample and technical replicates were conducted for 2

samples for each assay (Fig. 1A).

To assess immune cell composition of MM patients, bone marrow (BM) baseline samples

(collected at the initial diagnosis) from these eighteen patients were subjected to scRNA-seq,

with immune cells clustered based on their transcriptome profiles using the Louvain clustering

algorithm implemented by Seurat [7,8] (Fig. 1B). We then investigated immune cells of these

same samples by mass cytometry (CyTOF) using a 39-marker panel (Supplementary Table

S1D). Cell populations were characterized by expression of markers, clustered by the flowsom

algorithm [9], and visualized with vi-SNE in the Cytobank [10] platform (Fig. 1C). Given the

15
discordance between RNA expression and protein expression that is known to exist [11], it is

informative to characterize cell populations by measuring RNA and protein at the same time.

Finally, we utilized CITE-seq with antibody-oligonucleotide conjugates and 29 protein markers

(Supplementary Table S1B) to simultaneously quantify single-cell transcriptomes and surface

proteins. Following standard scRNA-seq quality filtering protocols, immune cells were clustered

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

based on integrated multi-omic profiles by the similarity network fusion (SNF) integration

algorithm in CiteFuse [12] (Fig. 1D). From CD138-negative BM aliquots, we detected, on

average, 1,051 immune cells/sample using scRNA-seq, >64K CD45+ cells/sample using

CyTOF, and 718 immune cells/sample using CITE-seq.

Advantages of CITE-seq in distinguishing T cell subtypes in multiple myeloma

To assess the potential advantages of simultaneous quantification of RNA and protein

expression in CITE-seq as compared to standard scRNA-seq, we labelled immune cell identities

determined by integrated transcriptome and protein expression, but clustered cells by

transcriptional profiles alone (Fig. 1E). Interestingly, most cell types, including B cells,

monocytes, macrophages, neutrophils, and plasmacytoid dendritic cells (pDCs), formed distinct

clusters, while T cell subtypes mixed together. To further understand the difference of cell type

marker expression between the RNA and protein levels, we visualized the expression of some

canonical markers in Uniform Manifold Approximation and Projection (UMAP) and investigated

the concordance of the sample-level average expression of the 29 CITE-seq protein markers

between RNA level and Antibody-Derived Tags (ADT) level (Fig. 1F, G; Supplementary Fig.

S1A). As expected, expression levels of markers are generally concordant (R =0.72, p < 10-4),

with some exceptions where protein-level expression is higher than RNA-level expression and

vice versa. One impressive example is CD4 (Fig. 1F, G), which is highly expressed at ADT

measurement, but minimally expressed at the RNA level, mainly because mRNAs are produced

16
at much lower rates and have much shorter half-lives than proteins [13]. This observation is

consistent with previous studies showing low CD4 mRNA expression compared to surface CD4

protein [14]. Last, since Naive CD8+ T cells were clustered together with CD4+ T cells based on

transcriptome profiles (Fig. 1E), we investigated whether reclustering T cells alone could help to

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

distinguish subtypes at the RNA level. Due to the high similarities of transcriptional profiles

among T cells [14] and different surface protein markers could be encoded by the same gene

[15], reclustering CD4+ and naive CD8+ T cells did not provide additional resolution of T cell

subtypes (Fig. 1H). Consistent with a published study about renal T subtype identification using

CITE-seq [16], our observation emphasizes the advantage of integrating protein-level

expression of cell type markers for MM T cell subtype identification in CITE-seq as compared to

standard scRNA-seq.

Data reproducibility and comparisons of cell populations measured by the same

technologies across different centers

To examine data reproducibility, percentages of cell subsets in CD45+ populations were

compared between technical replicates for 2 samples in each assay. The technical replicate

pairs are strongly correlated in all 3 assays (average Pearson correlation coefficient r=0.94 in

scRNA-seq, 0.89 in CyTOF, and 0.92 in CITE-seq) (Supplementary Fig. S1B-D). Next, to

examine the consistency of immune cell populations measured by the same techniques at

different sites, we evaluated the percentage of immune populations captured by three centers

using four samples. scRNA-seq data was generated in ISMMS, WUSTL and BIDMC using

aliquots of the same samples and CyTOF data was generated in ISMMS, Mayo Clinic and

Emory University (panels are shown in Supplementary Table S1D-F). BIDMC scRNA-seq data

is from CITE-seq data analyzed with RNA signal alone. (Supplementary Fig. S1E). We

observed that the percentages of B cells, pre-B cells, NK cells, pDCs, monocytes and

17
macrophages are generally consistent, while the T cell subset varies across centers in scRNA-

seq measurement (Supplementary Fig. S1F). This suggests that T cell composition could vary

by aliquots and potential sample processing differences across centers while other cell types

are more similar in scRNA-seq measurement. The cell type abundance measured by CyTOF is

less variable than that measured by scRNA-seq, with smaller differences observed in T cell

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

subsets across centers (Supplementary Fig. S1G, mean difference calculated by Bland-Altman

analysis, shown in Supplementary Table S1G). Moreover, cell subset abundances of ISMMS

samples tend to have less variation likely due to the benefit of barcoding samples (Methods).

The cell type frequencies calculated by one center (Emory) tend to be lower overall compared to

other centers in CyTOF, probably because wide bore injector assembly with cell acquisition

solution was not used to maintain cell integrity (Methods). It is worthwhile noting that including

reference samples in CyTOF is very helpful for identifying potential artifacts. For example, we

observed a big proportion of CD66b/CD3+ cells in patient samples while these were absent in

the reference sample from a healthy donor (data not shown). We hypothesized that this CD66b

staining artifact (CD66b is not expressed on CD3+ T cells) was likely due to non-specific

staining from dead cells. Indeed, the percentage of CD66b/CD3+ cells dropped dramatically

after dead cell depletion. Lastly, to evaluate the similarity of expression profiles across different

samples and centers, we calculated the Pearson correlation coefficient of expression of the B

cell markers between populations detected from different centers using scRNA-seq

(Supplementary Fig. S1H). We observed that B cells clustered according to patients instead of

centers, suggesting patient-dependence of B cell transcriptome profiles, likely because B cells

are potential reservoirs of plasma cells [17]. Overall, we observed that cell type abundances are

generally consistent across centers for most cell types and that similarity of transcriptome

profiles of immune populations is center-independent, suggesting absence of strong batch

effects across centers. These observations imply that our cross-technique comparisons should

be valid.

18
Comparisons of cell type abundances and correlations of cell type marker expression

across the three techniques

To evaluate the concordance of cell type composition determined by the three methods, we

calculated the cell subset frequency of each immune population relative to the CD45+

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

populations (Fig. 2A). Overall, all three approaches were concordant, though there is somewhat

stronger concordance between scRNA-seq and CITE-seq for all cell types except NK cells

(mean difference calculated by Bland-Altman analysis, shown in Supplementary Table S1H).

Cell type abundance is especially consistent for B cells, plasmacytoid dendritic cells (pDC), and

neutrophils. Interestingly, the cell frequency decreased and increased for T cells and

macrophages/monocytes, respectively, in CyTOF as compared with scRNA-seq and CITE-seq.

The mean differences between CyTOF and CITE-seq were -13.6% (95% CI: -24.02% to -3.11%)

for T cells and 11.07% (95% CI: 3.19% to 18.95%) for macrophages/monocytes. This finding is

consistent with a previous study where fewer T cells were detected in CyTOF compared to

scRNA-seq in healthy bone marrow samples [18]. To further investigate which subpopulations

were discordant, the frequencies of T cell subsets, monocytes, and macrophages were

evaluated (Fig. 2B, mean difference calculated by Bland-Altman analysis). Interestingly, CITE-

seq detected far more CD4+ T cells compared to CyTOF and scRNA-seq, while CyTOF

detected far fewer CD8+ T cells compared to the other two techniques. In terms of T cell

subtypes, regulatory T (Treg) cell frequency increased and memory CD8+ T cells reduced in

scRNA-seq, as compared to CyTOF. In addition, scRNA-seq detected far more macrophages

than the other 2 methods, while monocyte frequency was the lowest in CyTOF.

To further evaluate concordance between scRNA-seq and CITE-seq, we examined expression

of cell type marker genes, including both the RNA and ADT levels. Average expressions of each

marker gene at the transcriptional level (blue dots) between scRNA-seq and CITE-seq are

19
generally concordant (Fig. 2C). By contrast, we observed drastic differences of some marker

genes between RNA and ADT expression in CITE-seq, probably due to the RNA dropout [19]

and shorter half-lives of mRNAs versus proteins [13]. For example, expression of CD4_adt is

higher than that of transcriptional CD4, whereas CD127/IL7R tends to be highly expressed at

the transcriptional level. This dynamic explains why IL7R is often differentially expressed in

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

CD4+ T cell population, while CD4 is weakly expressed in scRNA-seq. Taken together, these

observations highlight the importance of choosing cell type marker genes best suited to

particular modalities.

We also correlated expressions of marker genes among scRNA-seq, CyTOF, and CITE-seq.

The vast majority are positively correlated in protein-protein comparison (Fig. 2D) and RNA-

RNA comparison (Fig. 2E). Next, we investigated the correlations of expressions of marker

genes between the transcriptome and protein levels (Fig. 2F, G; Supplementary Fig. S2A, B).

As expected, the overall correlation between different modalities is lower than that of the same

modalities. We observed significant correlation for some markers, including CCR7 in CD4+

naive T cells, IL7R in CD4+ memory T cells, and FCGR3A in NK cells, between RNA and

protein level of CITE-seq, while no markers are significantly correlated between scRNA-seq and

CyTOF (Fig. 2G). We also found that FCGR3A in macrophages has a strong correlation, while

some markers are significantly anti-correlated between CITE-seq transcriptional level and

CyTOF, such as CD3D, CD3G, IL7R, CD8A, etc. (Supplementary Fig. S2A, C; Supplementary

Table S1I).

20
Decreased ratio of CD4+/CD8+T cells from ISS stage 2 to ISS stage 3 patients and fast

progression-related gene signatures

Further, we sought to investigate the relationship between clinical features and immune cell

composition of MM patients by examining the ratio of CD4+/CD8+ T cells of patients at different

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

disease stages. A previous study used flow cytometry to reveal that this ratio was significantly

lower in peripheral blood mononuclear cells (PBMC) of MM patients as compared to that of

normal controls and the ratio decreased with the MM progression [5]. By integrating 3 assays,

we found the ratio tends to decrease from ISS stage 2 to ISS stage 3 patients (Fig. 3A). Further,

CITE-seq and CyTOF analyses revealed significant downregulation of CD45RA in stage 3

patients, suggesting that CD8+ T cells tend to be activated rather than naive in stage 3 patients

(Fig. 3B). In addition, we then identified several DEGs of NK cells from FPs relative to NPs,

including ARPC5, XAF1, RAC2 and PSMB9, as revealed by both scRNA-seq and CITE-seq

assays (Fig. 3C). ARPC5, Actin-Related Protein 2/3 Complex Subunit 5, has been revealed to

be highly expressed in patients with poor overall survival and could be treated as an

independent biomarker for patients with MM [20], consistent with our observations. A previous

microarray-based study found that RAC2, Rac Family Small GTPase 2, is significantly

upregulated in MM as compared to MGUS [21]. One subunit of the proteasome (PSMB9), was

remarkably highly expressed in cell groups with t(4;14) translocations versus cells from MGUS

[22]. In summary, previous studies indicated RAC2 and PSMB9 are associated with disease

development from MGUS to MM and our analysis suggested that they might also be related to

MM progression. Taken together, we observed the ratio of CD4+ T/CD8+ T cells decreased in

stage 3 patients relative to stage 2 patients, suggesting an increased population of CD8+ T cells

in bone marrow microenvironment (BMME) of patients in stage 3. We also found that RAC2 and

PSMB9 are upregulated in NK cells in FPs relative to NPs at transcriptional level, which could

potentially serve as MM progression markers.

21
Discussion

Single-cell sequencing technologies have been widely used in studying tissue heterogeneity,

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

tumorigenesis and metastasis given their advantages of being able to depict genome,

transcriptome, proteome and other mutli-omics profiles of single cells [23]. However, the

similarities of measurements across the various single cell techniques remains to be fully

elucidated. Herein, we integrated scRNA-seq, CyTOF, and CITE-seq to perform a detailed

comparison of their measurements for MM BM microenvironment. From CD138-negative BM

aliquots of 20 samples from 18 patients, we detected, on average, 1,051 immune cells/sample

using scRNA-seq, >64K CD45+ cells/sample using CyTOF, and 718 immune cells/sample using

CITE-seq. By clustering cells with or without protein profiles in CITE-seq, we showed the

advantages of multimodal measurement over transcriptional measurement alone of cell type

markers when characterizing T cell subtypes in MM (Fig. 1E, H). This observation is in line with

a study to investigate renal T cell subtypes by CITE-seq [16].

Next, to examine the consistency of cell populations measured by the same techniques at

different sites, we evaluated the cell subset abundances captured by three centers using four

samples. Cross-center comparisons (Supplementary Fig. S1F, G) suggested no strong batch

effect across centers and there are some important factors to consider in order to obtain

reproducible and reliable results: 1) It is important to include reference samples in CyTOF to

help identify marker non-specific staining artifacts; 2) Barcoding samples, sample delivery

mechanism, and using lyophilized panels is important in CyTOF experiments. Further, cross-

technique comparisons revealed that the percentages of immune populations measured by

scRNA-seq, CyTOF, and CITE-seq are generally concordant, except some variations in T cells,

22
macrophages, and monocytes (Fig. 2A, B). Analysis revealed relatively high correlations of most

markers between the same modalities, though some markers are negatively correlated. (Fig.

2C-G). This observation highlighted the importance of choosing marker genes best suited to

particular modalities.

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

Previous studies have found MM patients have lower CD4+ T/CD8+ T ratios relative to healthy

donors and these ratios are further decreased in ISS stage 3 versus ISS stage 1 patients [5].

Here, we confirmed this trend using 3 single cell technologies, finding that this ratio tends to

decrease even in stage 3 versus stage 2 patients (Fig. 3A). We also observed the decreased

ratio in stage 2 compared to stage 1 patients based on CyTOF and CITE-seq measurement but

not in scRNA-seq, probably due to the limited number of patients in stage 1. Future study could

further investigate how immune cell composition changes along with ISS stages with expanded

sample size. In addition, we observed upregulation of ARPC5, XAF1, RAC2, and PSMB9 in NK

cells of FPs compared to those of NPs, as suggested by both scRNA-seq and CITE-seq RNA

measurements (Fig. 3C). RAC2 and PSMB9 have been revealed to be associated with disease

development from MGUS to MM [21,22] and our analysis suggested that they might also be

related to MM rapid progression, supported by both scRNA-seq and CITE-seq. Due to the

limited number of protein markers in CITE-seq, we were unable to evaluate the protein-level

expression of these MM progression-related genes identified from RNA measurement, which

requires further validation. It would also be interesting to investigate MM progression-related

markers after controlling for treatments in future studies.

This analysis is just a small sampling of the larger work being conducted by the MMRF and their

associated academic research centers to provide a sufficiently broad, deep, and technologically

diverse vast dataset for accurately characterizing BMME and to help interrogate MM tumor

23
microenvironment (TME) using different single-cell technologies. We hope this study will help

researchers refine cell population characterization strategies and provide insights to those

considering integrating multiple single-cell methods to comprehensively address biological

questions.

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

Acknowledgements

This study was funded through the Multiple Myeloma Research Foundation (MMRF) Immune

Atlas initiative. We thank the multiple myeloma patients, families, and professionals who have

contributed to this study. We thank Upadhyaya Bhaskar,, Nicolas Fernandez, and Laura

Walker for their contribution to the initial work of MMRF immune atlas pilot study. We thank

John Leech for administrative support.

Author contributions

L.D., M.B., S.G., M.D., D.A., and S.D. led project design. L.Y. led data analysis, interpretation

and Figure generation. R.J., B.L., S.S.B., M.F., D.B.D., M.S., T.D., S.K. , H.C., E. A., G.K., and

D.D. contributed to sample processing and data generation. L.Y., R.J., S.S.B., T.K., E.G.,

D.B.D., T.D. G.K., D.D., B.T., Y.P., and W.P. contributed to data processing and data

interpretation. S. O., J.D., A.R., S.K., R. V., I.V., S.M., M. H., K.D., and D.A. contributed to study

design and supervision. L.Y wrote the manuscript and made figures. M.A.W helped polish

Figures. R.J., M.W, T.K., and D.B.D. helped edit the manuscript. L.D. and S.G reviewed the

manuscript.

24
References

1. Liu R, Gao Q, Foltz SM, Fowles JS, Yao L, Wang JT, et al. Co-evolution of tumor and
immune cells during progression of multiple myeloma. Nat Commun. 2021;12: 2559.

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

2. Zavidij O, Haradhvala NJ, Mouhieddine TH, Sklavenitis-Pistofidis R, Cai S, Reidy M, et al.
Single-cell RNA sequencing reveals compromised immune microenvironment in precursor
stages of multiple myeloma. Nat Cancer. 2020;1: 493–506.

3. Adams HC 3rd, Stevenaert F, Krejcik J, Van der Borght K, Smets T, Bald J, et al. High-
Parameter Mass Cytometry Evaluation of Relapsed/Refractory Multiple Myeloma Patients
Treated with Daratumumab Demonstrates Immune Modulation as a Novel Mechanism of
Action. Cytometry A. 2019;95: 279–289.

4. Redoglia V, Boccadoro M, Battaglio S, Dianzani U, Massaia M, Pileri A. Multiple myeloma:

altered CD4/CD8 ratio in bone marrow. Haematologica. 1990;75: 129–131.

5. Koike M, Sekigawa I, Okada M, Matsumoto M, Iida N, Hashimoto H, et al. Relationship

between CD4(+)/CD8(+) T cell ratio and T cell activation in multiple myeloma: reference to
IL-16. Leuk Res. 2002;26: 705–711.

6. Zelle-Rieser C, Thangavadivel S, Biedermann R, Brunner A, Stoitzner P, Willenbacher E, et

al. T cells in multiple myeloma display features of exhaustion and senescence at the tumor
site. J Hematol Oncol. 2016;9: 116.

7. Butler A, Hoffman P, Smibert P, Papalexi E, Satija R. Integrating single-cell transcriptomic

data across different conditions, technologies, and species. Nat Biotechnol. 2018;36: 411–
420.

8. Hafemeister C, Satija R. Normalization and variance stabilization of single-cell RNA-seq

data using regularized negative binomial regression. Genome Biol. 2019;20: 296.

9. Van Gassen S, Callebaut B, Van Helden MJ, Lambrecht BN, Demeester P, Dhaene T, et
al. FlowSOM: Using self-organizing maps for visualization and interpretation of cytometry
data. Cytometry A. 2015;87: 636–645.

10. Kotecha N, Krutzik PO, Irish JM. Web-based analysis and publication of flow cytometry
experiments. Curr Protoc Cytom. 2010;Chapter 10: Unit10.17.

11. Vogel C, Marcotte EM. Insights into the regulation of protein abundance from proteomic
and transcriptomic analyses. Nat Rev Genet. 2012;13: 227–232.

12. Kim HJ, Lin Y, Geddes TA, Yang JYH, Yang P. CiteFuse enables multi-modal analysis of
CITE-seq data. Bioinformatics. 2020;36: 4137–4143.

13. Schwanhäusser B, Busse D, Li N, Dittmar G, Schuchhardt J, Wolf J, et al. Global

quantification of mammalian gene expression control. Nature. 2011;473: 337–342.

25
14. Ding J, Smith SL, Orozco G, Barton A, Eyre S, Martin P. Characterisation of CD4+ T-cell
subtypes using single cell RNA sequencing and the impact of cell number and sequencing
depth. Sci Rep. 2020;10: 19825.

15. Ntranos V, Yi L, Melsted P, Pachter L. A discriminative learning approach to differential

expression analysis for single-cell RNA-seq. Nat Methods. 2019;16: 163–166.

16. Krebs CF, Reimers D, Zhao Y, Paust H-J, Bartsch P, Nuñez S, et al. Pathogen-induced
tissue-resident memory TH17 (TRM17) cells amplify autoimmune kidney disease. Sci

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

Immunol. 2020;5. doi:10.1126/sciimmunol.aba4163

17. Calame KL. Plasma cells: finding new light at the end of B cell development. Nat Immunol.
2001;2: 1103–1108.

18. Oetjen KA, Lindblad KE, Goswami M, Gui G, Dagur PK, Lai C, et al. Human bone marrow
assessment by single-cell RNA sequencing, mass cytometry, and flow cytometry. JCI
Insight. 2018;3. doi:10.1172/jci.insight.124928

19. Qiu P. Embracing the dropouts in single-cell RNA-seq analysis. Nat Commun. 2020;11:
1169.

20. Xiong T, Luo Z. The Expression of Actin-Related Protein 2/3 Complex Subunit 5 (ARPC5)
Expression in Multiple Myeloma and its Prognostic Significance. Med Sci Monit. 2018;24:
6340–6348.

21. Liu Z, Huang J, Zhong Q, She Y, Ou R, Li C, et al. Network-based analysis of the molecular
mechanisms of multiple myeloma and monoclonal gammopathy of undetermined
significance. Oncol Lett. 2017;14: 4167–4175.

22. Jang JS, Li Y, Mitra AK, Bi L, Abyzov A, van Wijnen AJ, et al. Molecular signatures of
multiple myeloma progression through single cell RNA-Seq. Blood Cancer J. 2019;9: 2.

23. Tang X, Huang Y, Lei J, Luo H, Zhu X. The single-cell sequencing: new developments and
medical applications. Cell Biosci. 2019;9: 53.

24. Dietz AB, Bulur PA, Emery RL, Winters JL, Epps DE, Zubair AC, et al. A novel source of
viable peripheral blood mononuclear cells from leukoreduction system chambers.
Transfusion . 2006;46: 2083–2089.

25. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods
of clinical measurement. Lancet. 1986;1: 307–310.

Figure legends

Fig. 1. Overview of cell populations of 18 multiple myeloma patient samples subject to

scRNA-seq, CyTOF, and CITE-seq.

26
A) Patient characteristics and single-cell data collection. FP and NP denote fast progressors

and non-progressors, respectively. ISS=International Staging System. ASCT=Autologous Stem

Cell Transplantation.

B) Umap projection of integrated scRNA-seq data, with cells colored by immune cell types.

C) t-SNE projection of integrated CyTOF data, with cells colored by immune cell types.

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

D) UMAP projection of integrated CITE-seq data, with cells clustered by integrated RNA and

ADT expression, colored by immune cell types.

E) UMAP projection of integrated CITE-seq data, with cells clustered by transcriptional level

alone, colored by immune cell identities from panel D.

F) Comparison of canonical cell type marker gene expressions between protein level (ADT, top)

and transcriptional level (RNA, bottom). Cells are colored by normalized expression.

G) Concordance of sample-level average expressions of CITE-seq protein markers measured at

RNA level and ADT level. The grey shaded area represents the 95% confidence interval around

the line of best fit. R = Pearson correlation coefficient

H) UMAP projection of CD4+ T cells and naive CD8+ T cells, which is the subset of integrated

data in panel E, with cells clustered by transcriptional level alone, colored by immune cell

identities from panel D and E.

Fig. 2. Comparison of cell subset frequencies and correlations of expression of canonical

cell type markers across different modalities.

A) Main immune cell population (CD45+) frequencies observed by CITE-seq, CyTOF, and

scRNA-seq. Each boxplot is colored by assay. CITE-seq populations are determined based on

integrated RNA and ADT expressions.

B) Immune cell subtype frequencies for CITE-seq, CyTOF, and scRNA-seq. Each boxplot is

colored by assay. CITE-seq populations are determined based on integrated RNA and ADT

expressions.

27
C) Concordance of sample-level average expressions of canonical cell type markers in main cell

subsets between scRNA-seq and CITE-seq. CITE-seq RNA and protein (ADT) level

expressions are represented by blue and red dots, respectively.

D) Spearman correlation coefficients of protein level expressions of cell type markers between

CyTOF and CITE-seq. Each dot represents a marker gene and the color of the dot represents

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

the p value of correlation. Markers are highlighted with an outer circle if the p value is less than

0.05.

E) Spearman correlation coefficients of transcriptional level expressions of cell type markers

between scRNA-seq and CITE-seq. Each dot represents a marker gene and the color of the dot

represents the p value of correlation. Markers are highlighted with an outer circle if the p value is

less than 0.05.

F) Spearman correlation coefficients of cell type markers between transcriptional level and

protein level expressions in CITE-seq. Each dot represents a marker gene and the color of the

dot represents the p value of correlation. Markers are highlighted with an outer circle if the p

value is less than 0.05.

G) Spearman correlation coefficients of cell type markers between transcriptional level

expressions from scRNA-seq and protein level expressions from CyTOF. Each dot represents a

marker gene and the color of the dot represents the p value of correlation. Markers are

highlighted with an outer circle if the p value is less than 0.05.

Fig. 3. Ratio of CD4+ T/CD8+ T of patients in different ISS stages and markers associated

with ISS disease stages and MM progression.

A) Violin plots showing the ratio of CD4+ T/CD8+ T of patients in ISS stage 2 and 3 in scRNA-

seq, CyTOF and CITE-seq. Horizontal lines indicate the median of data points in each group.

28
B) Violin plots showing single cell-level normalized expression of CD45RA in CITE-seq ADT

measurement and CyTOF. The difference is significant at p<=0.0001 based on Wilcoxon rank

sum test.

C) Heatmaps showing DEGs of NK cells of FP versus NP patients in CITE-seq RNA

measurement (left) and scRNA-seq measurement (right). The samples are ordered based on

Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022

hierarchical clustering of expression profiles of these genes in CITE-seq RNA measurement.

Expression values are scaled such that for each gene, the average of the scaled expression is 0

and the standard deviation is 1. Adjusted p values and log Fold Change in CITE-seq and

scRNA-seq were shown on the left and right side of DEGs respectively. FC=fold change.

29
Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022
Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022
Downloaded from http://aacrjournals.org/cancerrescommun/article-pdf/doi/10.1158/2767-9764.CRC-22-0022/3204323/crc-22-0022.pdf by guest on 01 September 2022