A cohort of 91 DLBCL patients at diagnosis from the BMS-LyTRANS clinical trial (ClinicalTrials.gov Identifier: NCT01287923) was used in this study. Clinical characteristics of DLBCL patients enrolled in this training cohort are listed in Table 1 . Patients with previous corticosteroid treatment were excluded from this study. As controls, age-matched heathy donors (HD, n = 49), follicular lymphomas (n = 9), mantle cell lymphomas (n = 9), chronic lymphocytic leukemias (n = 11), and marginal zone lymphomas (n = 10) were included. Part of these samples (DLBCL and HD) were used in a previous work (22). Prognosis scores were validated in a second cohort of 155 DLBCL patients from the recently published GAINED trial (ClinicalTrials.gov Identifier: NCT01659099) (24). Clinical characteristics of DLBCL patients enrolled in this validation cohort are listed in Table 1 . The research protocol was conducted under French legal guidelines and fulfilled the requirements of the local institutional ethics committee and biosecurity procedures.

Blood samples were collected on heparin tubes. Flow cytometry analysis of M-MDSCs, cMO, iMO, and ncMO were performed on whole blood (300 µL/tube) with the antibody panel shown Table S1 and the gating strategy defined Figure S1 . Absolute counts were obtained by using Flow-Count beads (Beckman Coulter, Brea, CA). An erythrocytes lysis (Uti-Lyse Dako, Carpinteria, CA) was performed before analysis by flow cytometry (Navios, Beckman Coulter). Analyses were performed using Kaluza software (Beckman Coulter).

OCI-Ly3 and OCI-Ly19 cell lines were cultured in OCI-Ly medium (IMDM supplemented with 10% human AB serum,1% penicillin–streptomycin (Invitrogen) and 50 µM of β-Mercaptoethanol). For supernatants, OCI-Ly 3 and 19 were seeded at 3 x10⁶ cells/ml in OCI-Ly medium for 24h (37°C, 5% CO₂). Supernatants were obtained after centrifugation and were frozen at -80°C until migration assay. For control of migration assay, OCI-Ly medium were incubated at (37°C, 5% CO₂) without cells for 24h before centrifugation and freezing.

Monocytes were obtained from PBMCs by elutriation before cryopreservation (plate-forme DTC; CIC Biotherapie, Nantes, France). Monocytes were thawed and resuspended at 4 x10⁶ cells/mL in RPMI 1640 (Invitrogen, Carlsband, CA, USA) supplemented with 10% FCS and antibiotics (Invitrogen) and then diluted at 2 x10⁶ cells/mL by adding, as control, the OCI-Ly medium (IMDM supplemented with 10% human AB serum,1% penicillin–streptomycin and 50 µM of β-Mercaptoethanol, or OCI-Ly3 or OCI-Ly19 supernatant. Two mL of cell suspension were seeded in a 6-well plate during 4 days before mass cytometry analysis or migration assay.

Cell labeling and mass-cytometry analysis were performed as previously described. (25–27) Briefly, cells were incubated with 25 µM cisplatin (Fluidigm San Francisco, CA, USA). Then, 5 x10⁶ cells were washed in PBS (HyClone Laboratories, Logan, UT, USA) containing 1% BSA (Thermo Fisher Scientific) and stained in 100 µL PBS and BSA 1%-containing Antibody cocktail. Cells were stained for 30 min at RT with the antibodies ( Table S2 ) Cells were washed twice in PBS - BSA 1% before fixation in 1.6% PFA, and permeabilization with methanol (Electron Microscopy Sciences, Hatfield, PA, USA). After incubating overnight at -20°C in MeOH, cells were washed twice with PBS -BSA 1% and stained 20 min with iridium intercalator (Fluidigm, Sunnyvale, CA, US). Finally, cells were washed twice with PBS and twice with diH2O before acquisition a CyTOF 2.0 mass cytometer (Fluidigm). Mass cytometry raw data were deposited in Flow Repository (http://flowrepository.org/id/FR-FCM-Z4N6).

Data analysis was performed using the workflow previously developed (23). Briefly, after acquisition, intrafile signal drift was normalized and.fcs files were obtained using CyTOF software. To diminish batch effects, all files were normalized on EQ Beads (Fluidigm) using the premessa R package (https://github.com/ParkerICI/premessa). Raw median intensity values were transformed to a hyperbolic arcsine (arcsinh) and then analysis was performed using Cytobank software (Beckman Coulter, Brea, CA, USA).

At day 4 of monocyte culture with OCI-Ly3, or OCI-Ly19 supernatant, or control culture medium, cells were collected and washed twice in PBS before starvation during 1 hour (37°C, 5% CO₂) at 10⁶ cells/mL in RPMI 1% HSA. Then, cells were washed once, counted and 100 µL of cells at 10⁶ cells/mL were added to the upper compartment of Transwell chambers with 5 µM pore filters (Corning Incorporated, Kennebunk, ME, USA). The lower chamber contained a chemiokine among CCL2 (R&D Systems, 30 ng/ml), CCL3 (R&D Systems, 20 ng/ml), CCL5 (R&D Systems, 30 ng/ml), CCL22 (R&D Systems, 20 ng/ml), CXCL5 (R&D Systems, 20 ng/ml), CXCL12 (R&D Systems, 20 ng/ml), or RPMI 1640 1% HSA as control (corresponding to Basal migration). Cells in the lower chamber were collected after 5h (37°C, 5% CO₂) and the absolute number of viable (DAPI negative) monocytes was quantified by flow cytometry using Precision Count Beads™ (Biolegend Inc, San Diego, CA, USA). The absolute number of cells (Total Cells) added to the upper compartment at the beginning of migration assay was also determined for each condition (control, OCI-Ly3 and OCI-Ly19) with Precision Count Beads™. Percentage of cells that have specifically migrated against a chemokine was calculated as follow: Specific migration = (Lower chamber - Basal migration)/Total cells x 100. The migration represents the variation of this percentage with OCI-Ly supernatants compared to control culture medium (% of specific migration with supernatant – % of specific migration with medium) for each chemokine.

cMO (CD19^neg CD3^neg CD335^neg CD45^pos CD14^high CD16^neg), iMO (CD19^neg CD3^neg CD335^neg CD45^pos CD14^high CD16^pos), ncMO Slan^pos (CD19^neg CD3^neg CD335^neg CD45^pos CD14^low CD16^pos Slan^pos), and ncMO Slan^neg (CD19^neg CD3^neg CD335^neg CD45^pos CD14^low CD16^pos Slan^neg) were sorted from thawed PBMC of DLBCL patients and HD using an ARIA II (FACSAria, BD Biosciences).

Total RNA was extracted usingNucleospin^® RNA XS kit (Macherey-Nagel, Duren, Germany). cDNA was then generated using Fluidigm Reverse Transcription Master Mix (Fluidigm). The qPCR were performed in triplicate using 96.96 Dynamic Array™ IFCs and the BioMark™ HD System from Fluidigm. For each sample, the mean CT value for the gene of interest was calculated, normalized to the geometric mean value of the 2 housekeeping genes (CDKN1B and ELF1) ( Table S3 ), and compared to the median value obtained from the reference population (HD cMO or iMO, and DLBCL ncMO) using the 2-ddCT method. Results were expressed as the ratio of sample mean to reference mean for each gene.

Statistical analyses were performed with GraphPad Prism 8.4.3 software (GraphPad Software, San Diego, CA, USA) using Pearson correlation, Wilcoxon, Mann-Whitney, Ordinary one-way ANOVA with Tukey’s multiple comparisons test, and Fishers’s exact tests as appropriate. Optimal thresholds were defined with the maxstat package, log-rank tests were performed with the survminer package, cox model for univariate and multivariate analysis were performed with the survival package. Analyses were generated with R v4.0.3, using Rstudio v1.3.1093.

For original data, please contact the corresponding author. Mass cytometry raw data were deposited in Flow Repository (http://flowrepository.org/id/FR-FCM-Z4N6. QPCR raw data are included in Table S4 .

We have previously shown that M-MDSCs accumulated in DLBCL peripheral blood (22). However, this increase accounts for only a part of the total monocyte accumulation suggesting that additional monocyte subsets are also increased in DLBCL samples ( Figure 1A ). We quantified the absolute count of the 4 circulating monocyte subsets M-MDSC, cMO, iMO, and ncMO. M-MDSC, cMO, and iMO were increased in DLBCL when compared to HDs (P <.05, median: 5.75 x10⁶ cells/L vs 2.8 x10⁶ cells/L, 348.8 x10⁶ cells/L vs 274.1 x10⁶ cells/L, and 34.4 x10⁶ cells/L vs 26.1 x10⁶ cells/L; respectively). Conversely, ncMO were significantly decreased in DLBCL when compared to HD (P <.0001, median: 17.1 x10⁶ cells/L vs 36.1 x10⁶ cells/L; Figure 1B ). Noteworthy, whereas the increase of cMO and iMO was also found in other B cell lymphomas, this decrease in ncMO was specific of DLBCL ( Figure S2 ). Then, we wondered in which monocyte subset M-MDSCs were included. Of note M-MDSC count was correlated with total monocyte (R = .57, P <.0001), cMO (R = .61, P <.0001), but was not correlated with iMO and ncMO ( Figure 1C ) No correlation were observed between MO subsets in HD samples (data not shown). Regarding CD14 and CD16 expression, M-MDSCs were essentially aligned with the cMO phenotype and to a lesser extent to iMO ( Figure 1D ). Altogether, these results confirmed that in addition to MDSCs, cMO and iMO were also involved in the monocyte increase observed in DLBCL patients.

To further identify the immune properties of monocyte subsets, we sorted cMO and iMO from DLBCL (n=7) and HD (n=4) samples. Gene expression was assessed by high throughput qPCR on 71 genes involved in myeloid biology (22) ( Tables S3 , S4 , and Figure S3 ). Of note, 6 cMO and 5 iMO out of 7 DLBCL were clustered ( Figure 2A and Figure S3 ). DLBCL cMO and DLBCL iMO were significantly enriched for inflammatory genes (FCGR3A, CD36, FCGR1A, CYBB, AIM2, STAT6, FCGR2A, CCR2, NLRC4, S100A8, and CD14 genes), when compared to the corresponding subsets in HDs (P <.05, ∣log2FC∣>1) ( Figure 2B ). In addition, S100A9 and CD163 were also increased in DLBCL cMO, whereas CD33 and ITGAM were enriched only in DLBCL iMO. For both subsets, SLC7A11, CD274, and CXCL1 were expressed at lower levels in DLBCLs ( Figure 2B ). Biological processes enriched in DLBCL cMO and iMO included apoptosis, production of ROS, immune response, and phagocytosis ( Figure 2C ). By flow cytometry, we showed that cMO and iMO from DLBCL displayed a higher expression of CD64 and CCR2 (P <.05), without variation in HLA-DR and CD163 ( Figure 2D ). Altogether, these results suggested that, in DLBCL, cMO and iMO share a common deregulated inflammatory phenotype.

We then focused on ncMO in DLBCL samples and found that both subsets of ncMO (ncMO Slan^pos and ncMO Slan^neg) were decreased in DLBCLs when compared to HDs [ncMO Slan^pos median at 3.3 x10⁶ cells/L vs 9.1 x10⁶ cells/L (P <.001) and ncMO Slan^neg median at 19 x10⁶ cells/L vs 23.7 x10⁶ cells/L (P <.05), respectively] ( Figure 3A ). No increase of apoptosis was detected in ncMO from DLBCL patients (data not shown). Then, sorted ncMO Slan^pos and ncMO Slan^neg from DLBCL and HD samples were analyzed by high-throughput Q-PCR ( Figure S4 ). To explore the similarities between ncMO Slan^pos, ncMO Slan^neg, iMO, and cMO, we performed an unsupervised hierarchical clustering on the DLBCL samples. For 6 out of 7 patients, cMO and iMO were separated from ncMO independently of Slan expression ( Figure 3B and Figure S5 ). In addition, ncMO Slan^pos and ncMO Slan^neg exhibited tolerogenic genes (PDCD1LG2, IL10, IDO, CD274, AGER, TNFAIP6) ( Figure 3C and Figure S5 ), most of these genes were not expressed in HD ncMO ( Figure S6 ). In DLBCL, cMO and iMO in one hand and ncMO Slan^pos and ncMO Slan^neg in the other hand shared similar gene expression ( Figure 2A and Figure S5 ), thus we compared the gene expression between ncMO, irrespectively of the Slan status, and both cMO and iMO. DLBCL ncMO were enriched for both inflammatory (CXCL10, AIM2, IL12A) and tolerogenic (PDCD1LG2, IL10, IDO, CD274, AGER) genes (P <.05, ∣log2FC∣ > 1) compared to cMO and iMO ( Figure 3C ). Biological processes involved by genes enriched in ncMO from DLBCL patients were growth of tumor, inhibition of cells, and chemotaxis ( Figures 3C, D ).

In order to evaluate how tumor B cells contribute directly to the phenotype of DLBCL monocytes, we cultured monocytes from HD with supernatants from the DLBCL cell lines OCI-Ly3 and OCI-Ly19. After coculture, we analyzed the monocyte phenotype by mass cytometry (23, 25) and noticed an increased expression of CD16, Slan, CD64, CD163. We concluded that tumor conditioned monocytes triggered the cMO to ncMO transition ( Figure 4A ). Then we wondered if these cells were prone to migrate into tissue. Tumor-conditioned monocytes demonstrated an increase in in vitro migration in response to CCL5 and CXCL12 when compared to non-conditioned monocytes ( Figure 4B ).

Then, we evaluated the prognosis value of cMO, iMO, and ncMO in DLBCL. We used i) the proportion of ncMO to other monocytes (ratio ncMO to sum of cMO and iMO) and ii) the absolute count of circulating cMO, iMO, and ncMO. Analysis was performed on 52 patients for which clinical data were available. cMO and iMO were not associated with prognosis (data not shown). By contrast, patients with high proportion of ncMO and high absolute count of circulating ncMO were associated with a lower event-free survival probability (P = .043 and P = .0061, respectively) using thresholds (ratio at 0.06 and ncMO at 20.58 x10⁶ cells/L) defined with the maxstat package ( Figure 5A and Figure S7 ). To validate the prognosis value of ncMO obtained on this training cohort, we analyzed by flow cytometry the proportion of monocyte subsets in an independent cohort of 155 DLBCL samples from the recently published GAINED trial (NCT01659099) (24). With the previously calculated thresholds, high proportion of ncMO and high absolute count of ncMO was associated with a lower overall survival (P = .017 and P = .011, respectively) ( Figure 5A ). A univariate analysis on the validation cohort showed that Ann Arbor Stage III-IV, ECOG status >1, elevated LDH, PET4 positivity, and increase in circulating ncMO were associated with lower OS ( Table 2 ). In a multivariable analysis Ann Arbor Stage III-IV, PET4 positivity, increase in circulating ncMO remained statistically significant ( Table 2 ). We previously demonstrated the accumulation of M-MDSC in DLBCL (22) and since no phenotypic overlap existed between M-MDSC and ncMO ( Figure 1D ), we wondered if patients’ characteristics were different between M-MDSC^high and ncMO^high DLBCLs. Both M-MDSC and ncMO were infrequently increased together [11 cases out of 155 (7.1%)]; ncMO were increased alone in 51 cases (32.9%), and M-MDSC were increased alone in 28 cases (18.1%) ( Figure 5B ). Interestingly, ncMO^high and M-MDSC^high patients corresponded to different types of patients. In particular when compared to ncMO^low, ncMO^high were enriched in ABC DLBCL subtypes [37.5 vs 15.9% (P = .014)] and in older patients [median age at 50 vs 46 years (P = .044)]. On the other hand, when compared to ncMO^high, M-MDSC^high patients were younger [median age at 42 vs 50 years (P = .0027)] and had higher levels of soluble PD-L1 [sPD-L1 at 1849 vs 1142 pg/mL (P = .008)] ( Figure 5B ).