Quorum-sensing mechanisms that sense the density of immune cells at the site of inflammation to initiate inflammation resolution have recently been demonstrated as a major determinant of the inflammatory response. We observed a density-dependent increase in expression of the inflammatory tumor suppressor protein programmed cell death 4 (PDCD4) in mouse macrophage cells. Conditioned medium from high-density cells upregulated PDCD4 expression, revealing the presence of a secreted factor(s) acting as a macrophage quorum sensor. Secreted gelsolin (GSN) was identified as the quorum-sensing autoinducer. Alteration of GSN levels changed PDCD4 expression and the density-dependent phenotype of cells. LPS induced the expression of microRNA miR-21, which downregulated both GSN and PDCD4 expression, and reversed the high-density phenotype. The high-density phenotype was correlated with an anti-inflammatory gene expression program, which was counteracted by inflammatory stimulus. Together, our observations establish the miR-21–GSN-PDCD4 regulatory network as a crucial mediator of a macrophage quorum-sensing mechanism for the control of inflammatory responses.

This article is featured in Top Reads, p.1223

Inflammation is the primary protective response of the body to infection and injury; however, unresolved inflammation is the major driver of many diseases, such as atherosclerosis, rheumatoid arthritis, inflammatory bowel disease, Alzheimer disease, multiple sclerosis, and cancer (1). Resolution of inflammation is an active and regulated process aimed at restoring tissue homeostasis and function (2). Although the decrease of the pathogenic load or tissue injury is an important factor in limiting inflammation, the active role of immune cells and molecular mediators is also an important determinant of inflammation resolution, demonstrating that resolution of inflammation is an active process and is not merely a result of the cessation of the inflammatory stimulus (3). These include processes that cause neutrophil apoptosis and clearance by macrophages, regulation of cytokine and chemokine release and activity, release of inflammation-resolving cytokines such as TGF-β and IL-10, and regulation by reactive oxygen and nitrogen intermediates (3, 4).

An important step in inflammation resolution is the transition of macrophages from a proinflammatory to an anti-inflammatory phenotype (5). Tissue-resident and monocyte-derived macrophages play a crucial role both in the initiation and resolution of the inflammatory response. Proinflammatory macrophages, at the onset of inflammation, respond to pathogen-derived molecules such as bacterial LPS to trigger specific signaling and gene-regulatory pathways resulting in the release of proinflammatory cytokines, cell migration, and phagocytic activity. Later in the inflammatory response, the macrophages switch to an anti-inflammatory phenotype marked by the secretion of anti-inflammatory cytokines and an anti-inflammatory gene expression program, contributing to inflammation resolution (6). In contrast, anti-inflammatory macrophages can also attain an alternative phenotype that directly influence multiple steps in tumor development, including cancer cell survival, proliferation, stemness, invasiveness, angiogenesis, and immunosuppression (7). This necessitates a fine-tuning of macrophage activity in response to inflammatory stimuli. Although this is well known, mechanisms that select the appropriate time or context for inflammation resolution, and the transition of macrophages from the pro- to the anti-inflammatory phenotype, are not well understood. Importantly, mechanisms that sense when a sufficient density of macrophages has been attained at the site of inflammation to initiate inflammation resolution have mostly remained unexplored (8).

Recently, quorum-sensing mechanisms that can sense the density of macrophages at the site of inflammation have been demonstrated as a major determinant of the inflammatory response (8, 9). Quorum sensing is a well-known mechanism in bacteria that initiates specific group behaviors only when a sufficient cell density has been reached and is mediated by specific diffusive molecule(s) that act as reporters for cell density (10). Although quorum sensing is not known in mammalian systems, recently, several regulatory processes akin to quorum-sensing mechanisms have been demonstrated in the mammalian immune system. Specifically, quorum-sensing mechanisms appear to play an important role in determining macrophage function and termination of inflammation (9, 11). A potential pathway based on the central role of CSF-1 receptor signaling has been proposed, as it regulates the survival and proliferation of macrophages at steady state (12, 13). Regeneration of hair follicles in response to patterned hair plucking has been proposed to be regulated by a quorum-sensing mechanism involving TNF-α–producing macrophages (14). Recently, TNF-α has also been implicated in a quorum-licensing mechanism in regulating macrophage function and cytokine production in groups versus in single cells (15). Also, a metabolism-based quorum-sensing mechanism relying on NO as a regulator of macrophage number and activity has been demonstrated for the control of inflammation resolution in response to parasite infection (16).

Although these important discoveries have demonstrated the importance of quorum-sensing mechanisms dependent on cell density in the regulation of macrophage function, molecular markers of cell density in macrophages and secreted molecule(s) that function as surrogates for cell density have not been identified. In this study, we have demonstrated programmed cell death 4 (PDCD4), a proinflammatory tumor suppressor protein, as a marker of high cell density of macrophages. PDCD4 has been shown to play an important role in macrophage function and exhibits both inflammatory and anti-inflammatory activities (17). PDCD4 has been previously shown to be upregulated in a cell density–dependent manner in keratinocytes and regulate keratinocyte cell proliferation and contact inhibition in vitro (18). We have further shown that the expression of PDCD4 in high-density macrophages is regulated by secreted gelsolin (GSN) in a density-dependent manner, thereby demonstrating GSN as a bona fide quorum sensor and autoinducer. Finally, we have shown that activation by the inflammatory agonist LPS inhibits both GSN and PDCD4 expression via microRNA-21 (miR-21) thereby causing a reversal from the anti-inflammatory phenotype of the high-density macrophages to a proinflammatory phenotype. Together, our observations establish the GSN–PDCD4 axis as a central mediator of a macrophage quorum-sensing mechanism for the control of inflammatory responses.

Bacterial LPS (Escherichia coli, serotype 026:B6), actinomycin D, MG132, pyrrolidine dithiocarbonate (PDTC), SB203580, FR180204, and GW4869 were from Sigma-Aldrich (St. Louis, MO). Cycloheximide was purchased from Amresco. Corning Costar Transwell inserts were from Corning. Small interfering RNAs (siRNAs; siGENOME SMARTpool) were purchased from Dharmacon. The full-length, 236-bp GSN 3′ untranslated region (UTR) was amplified from RNA isolated from RAW264.7 cells by RT-PCR. It was cloned downstream of firefly luciferase gene in pCDNA3.1-Fluc vector between EcoRI and NotI sites. The putative miR-21 target site in the GSN 3′ UTR was mutated, and the mutant 3′ UTR was cloned downstream of firefly luciferase gene between the same sites. miR-21 miRNA was expressed from pSUPER–miR-21, which contains an miR-21 short hairpin RNA sequence cloned in pSUPER vector (19). pCDNA3.1_HSA GSN-DYK (clone: OHu20043D, accession no.: NM_198252.3, https://www.ncbi.nlm.nih.gov/nuccore/NM_198252) was purchased from GenScript. pGEM-T and pRL-CMV vectors were from Promega.

The human GSN protein-coding sequence was PCR amplified from the mammalian over expression construct (pCDNA3.1_HSA GSN_DYK) and cloned into bacterial expression plasmid pET28b. The plasmid pET28b_HSA GSN was transformed into E. coli BL21 (DE3) strain for expression of N-terminally His-tagged recombinant GSN (rGSN) protein. The bacterial culture was induced with 0.6 mM isopropyl β-d-thiogalactoside for 5 h, and the cell pellet was resuspended in lysis buffer (50 mM NaH2P04, 300 mM NaCl, and 10 mM imidazole) along with 1 mg/ml lysozyme and kept in ice for 30 min. Cells were lysed by sonication, and lysate was collected after centrifugation (10,000 rpm for 30 min). The His-tagged rGSN protein was purified by Ni-affinity chromatography using Ni-NTA agarose (QIAGEN) and eluted using increasing concentrations of imidazole.

Bacterial endotoxins were removed from the purified protein preparation as described previously (20). Briefly, eluted protein fractions were mixed with 1% Triton X-114 (Sigma-Aldrich) by vigorous vortexing and immediately kept in ice for 5 min to achieve a homogeneous solution. The samples were then incubated at 37°C for 5 min to allow phase separation into a lower micellar phase and an upper aqueous phase. Finally, phases were separated by a short centrifugation, and the upper aqueous layer was collected carefully in a separate tube leaving behind the endotoxin encapsulated micellar phase. The same process was repeated thrice to ensure a maximum removal of endotoxin from the protein samples. To reconstitute the disulfide bonds characteristic of plasma GSN (pGSN), the protein samples were mildly oxidized by overnight incubation in 25 mM Tris-HCl (pH 8), 2 mM CaCl2, and 50 mM NaCl in presence of 2 mM oxidized glutathione (Sigma-Aldrich) as described previously (21).

Mouse macrophage cell line RAW264.7 was maintained in RPMI 1640 media (Thermo Fisher Scientific) supplemented with 10% heat-inactivated FBS (Thermo Fisher Scientific), 100 U/ml penicillin and streptomycin, and 4 mM l-glutamine in a humidified incubator at 37°C and 5% CO2. C57BL/6J mice were obtained from pathogen-free animal facility of University of Alabama at Birmingham (UAB). All animal procedures were reviewed and approved by the UAB Institutional Animal Care and Use Committee in compliance with the National Research Council Guide for the Care and Use of Laboratory Animals. Bone marrow–derived macrophages (BMDM) were isolated from 8–12-wk-old healthy mice and maintained in 1× macrophage serum-free media (Thermo Fisher Scientific) supplemented with 10% FBS, 30% L929-conditioned media, and 1% penicillin–streptomycin as previously described (22). For each treatment, fresh 1× macrophage serum-free media supplemented with 10% FBS, and 1% penicillin–streptomycin were added. Cells were treated with LPS (500 ng/ml), 100 µg/ml cycloheximide, 1.5 µg/ml actinomycin D, 5 µM MG132, 20 µM PDTC, 10 µM GW4869, 10 µM SB203580, 30 µM FR180204, or bacterially expressed and glutathione-oxidized recombinant human GSN protein, as indicated in results. Cells were transfected with different plasmid DNA, siRNAs, and antagomiR-21 using TurboFect Transfection Reagent (Thermo Fisher Scientific) in serum-free Opti-MEM medium. DNA amount for transfection was normalized using pGEM-T vector DNA.

A total of 4 × 105 cells were seeded in 35-mm dish. After 24 h of seeding, supernatant medium (conditioned medium [CM]) was collected as 0 h CM, PBS wash was given, and fresh medium was added and kept for another 24 h, following which the supernatant was collected as 24 h CM. The cell supernatant was centrifuged to remove cells and debris and then concentrated by centrifugation in 5-kDA cutoff centrifugal concentrator. For inhibitor treatment, cells were seeded and after 24 h of seeding, cells were washed with 1× PBS, followed by incubation in different inhibitors for 24 h in fresh media. After 24 h, CM was collected as described before. The CM was sterilized by passing through 0.22 μm sterile filters and stored in aliquots at −20°C until use.

For exosome isolation, cells were seeded in low-density and high-density conditions in media containing exosome-free FBS. After 24 h of seeding, cells were treated with or without exosome inhibitor and incubated for another 24 h. Ten milliliters of collected CM was lyophilized and then resuspended in 500 µl 1× PBS, which was passed through qEV Exosome isolation column (Izon Science) for exosome isolation following manufacturer’s protocol.

Cells were washed with precooled PBS and then lysed on ice in S10 lysis buffer (10 mM HEPES [pH −7.4], 15 mM KCl, 1 mM PMSF, 1 mM DTT, and 0.1% Triton ×100), followed by centrifugation at 10,000 × g for 15 min for cytoplasmic lysate preparation. Protein concentration was measured using Bradford assay (Amresco). Proteins were separated by 12% SDS–PAGE and transferred onto polyvinylidene difluoride membranes (MilliporeSigma) using semidry transfer apparatus (TE70×; Hoefer). Blots were then incubated with primary Abs, PDCD4 (D29C6; Cell Signaling Technology), GSN (D9W8Y; Cell Signaling Technology), GAPDH (FL-335; Santa Cruz Biotechnology), flotillin (D2V7J; Cell Signaling Technology), HRP-conjugated β-actin Ab (A00192-40; GenScript), and HRP-conjugated anti-Rabbit (Cell Signaling Technology) as secondary Ab. The bands were detected using femtoLUCENT chemiluminescence detection kit (G-Biosciences).

Total RNA was isolated using RNAiso Plus (Takara Bio) according to the manufacturer’s protocol. Total RNA (1 µg) was subjected to reverse transcription using Oligo-dT primer (Thermo Scientific) or miRNA-specific adaptor primer by M-MLV Reverse Transcriptase (Invitrogen) in a 20 µl reaction volume for 60 min at 37°C. Quantitative PCR for GSN mRNA was performed using primer set ACCTGGGACCAGGTCTTTGTCTG (forward) and CCACGAAGGAAGGAGGCTCAAAG (reverse) complementary to sequences in the last exon junction and 3′ UTR, respectively. Quantitative PCR specific to pGSN mRNA was performed using primer set CAAAGTCGGGTGTCTGAGGC (forward) and CAGGCACCAGGTCAAACTTCTCC (reverse) complementary to sequences present in the 5′ 151-nt region specific to pGSN mRNA and the first exon junction, respectively. Quantitative real-time PCR was performed using PowerUp SYBR Green Master Mix (Applied Biosystems) in StepOnePlus Real-Time PCR System (Thermo Fisher Scientific). U6B small nuclear RNA (snRNA) and GAPDH mRNA was used for miRNA and mRNA quantity normalization, respectively.

CM from low-density and high-density cells were collected and lyophilized. The lyophilized samples were diluted with 6 M urea and 100 mM Tris (pH 8). The samples were reduced and alkylated and then diluted to give a final urea concentration of <1 M for digestion with trypsin. The digestion was carried out by adding ∼10 μl of 0.1 μg/μl trypsin in 100 mM Tris and incubating overnight at room temperature. The samples were then cleaned up on a C18 spin column prior to liquid chromatography–mass spectrometry analysis on a Finnigan LTQ-Obitrap Elite hybrid mass spectrometer system. Five-microliter volumes of the samples were injected into HPLC column, and the peptides were eluted from the column by an acetonitrile/0.1% formic acid gradient at a flow rate of 0.25 μl/min and introduced into the source of the mass spectrometer online. The microelectrospray ion source was operated at 2.5 kV. The digest was analyzed using the data-dependent multitask capability of the instrument acquiring full scan mass spectra to determine peptide molecular weights and product ion spectra to determine amino acid sequence in successive instrument scans. The data were searched against the mouse and bovine UnipProtKB database using Sequest.

Cell migration assay was performed using 8-µm pore-sized Transwell cell culture inserts (Corning). Zero- and twenty-four-hour LPS (500 ng/ml) treated or untreated RAW264.7 cells (1.0 × 105) were seeded on the upper well in 100 µl serum-free RPMI media. The lower well was filled with 200 µl medium with 10% FBS. Control siRNA and PDCD4 siRNA transfected RAW264.7 cells were similarly seeded in the upper chamber. For investigating the effect of CM or rGSN on cell migration, cells were treated with rGSN or CM collected from low-density and high-density cells with and without LPS and also from CM of GSN-overexpressing and -knockdown cells. After incubation for 16 h, cells that had not migrated were scraped off with a cotton swab from the upper surface of the membrane, and the membrane was excised carefully. The membrane was fixed in 100% methanol and stained with crystal violet 0.5% to visualize the migrated cells. Filters were photographed, and the total number of cells was counted using a microscope (Olympus IX70).

CM collected from low-density and high-density cells were incubated at 4°C with rotation for 4 h with anti-GSN Ab (D9W8Y; Cell Signaling Technology) coupled to protein G-agarose beads in NT2 buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM MgCl2, 0.05% NP-40). The beads were pelleted down by centrifugation, and the supernatant was collected. The immunodepletion of GSN from the supernatant was determined by immunoblot with anti-GSN Ab.

Low-density cells were incubated with a blocking Ab (ITGA5, A6209, 1:200 dilution; ABclonal Science) against integrin α5 subunit of integrin α5β1 receptor for 1 h, following which they were incubated with CM from low-density or high-density cells for 12 h. Cells were then harvested, washed with PBS, and lysed, and lysates were immunoblotted.

Cells were seeded in 96-well plates and after 24 h of seeding, cells were transfected with different luciferase constructs and/or antagomiR-21. Twenty-four hours posttransfection, cells were washed with 1× PBS, and LPS treatment was given to cells in fresh medium. After 24 h of LPS treatment, cells were washed with 1× PBS and lysed in 1× passive lysis buffer (Promega). Luciferase activity was measured using multimode microplate reader (Chameleon, Hidex) using dual luciferase reporter assay system (Promega) according to the manufacturer’s protocol.

Cells were seeded in 96-well plates, and after 24 h of seeding, cells were treated with or without LPS (500 ng/ml). The culture supernatants were collected, and the concentrations of NO were measured according to the Griess reaction using o-phosphoric acid and N-(1-Naphthyl)ethylenediamine dihydrochloride (Sigma-Aldrich) and sulphanilamide (Sisco Research Laboratories) as described earlier (23).

Total RNA isolation was performed as described in the previous section. cDNA synthesis, sequencing, and analysis of RNA sequencing (RNAseq) data were performed. In brief, paired-end sequence data were generated from samples using Illumina HiSeq, and quality was checked using FastQC software. Adapter sequences, low-quality reads, and small fragmented reads (read size below 50 bp) were removed using fastp. Sequence reads were mapped to mouse reference genome (GRCm38.98) using TopHat2, and subsequently, Cufflinks and Cuffmerge package were used for transcript assembly and quantification of gene expression in fragments per kilobase of transcript per million (FPKM) (24). For differential gene expression analysis, Cuffdiff package was used (25). Genes with log2 fold change ≥1.5 was considered as upregulated genes and log2 fold change ≤−1.5 as downregulated genes. All p values in volcano plots were calculated from the mean and variance of paired-end reads for each transcript. Genes that did not have reads in any of the samples were removed from the analysis. Data represented were obtained from one experiment. Gene ontology (GO) analysis was performed using DAVID (26). The GO terms defining gene clusters are ordered based on enrichment p values.

Cell viability was measured at 0 and 24 h using MTT assay (Sigma-Aldrich) as per manufacturer’s protocol. For cell cycle analysis, cells were harvested, fixed, and stained with propidium iodide with RNase treatment. DNA content was analyzed using flow cytometry (FACSCalibur; Beckton Dickinson). Apoptosis was detected by caspase assay using CaspaseGlo caspase-3/7 assay kit (Promega) following manufacturer’s protocol.

All graphical data represent mean ± SD of at least three independent experiments (biological replicates). A single asterisk (*), single hashmark (#), and single dollar sign ($) signify p values ≤0.05; double asterisk (**), double hashmark (##), and double dollar sign ($$) signify p values ≤0.01; and triple asterisk (***), triple hashmark (###), and triple dollar sign ($$$) signify p values ≤0.001 (paired two-tailed Student t test) between controls and samples as indicated in the figures.

LPS treatment has been shown to reduce PDCD4 expression in human PBMCs and mouse macrophages (27). We treated RAW264.7 mouse macrophage cells with LPS for 24 h and observed a gradual decline in PDCD4 expression. Remarkably, untreated cells showed a sharp increase in PDCD4 expression 24 h after incubation (Fig. 1A). As this enhancement was nonlinear and showed a sharp increase after 24 h of cell growth, it pointed at regulation by a quorum-sensing type mechanism. To determine whether this increase in PDCD4 expression was density dependent, we seeded the same number of cells (4 × 105) in dishes of increasing diameter and, therefore, at decreasing density. After 24 h of incubation, the highest level of PDCD4 expression was in 35-mm dishes with a decrease in 60-mm dishes and with no change in expression in 90-mm dishes (Fig. 1B). The cells were exposed to equal volume of media in all dishes. This clearly indicated a density-dependent expression of PDCD4. This expression of PDCD4 was abrogated upon 24 h of LPS treatment in dishes of all diameters (Fig. 1B). We also determined that this increase in PDCD4 was actually a function of cell density, and not cell number, by taking equal number of cells after harvesting from different diameter dishes and lysing and immunoblotting for PDCD4. In this case too, cells from the 35-mm dishes showed a high expression of PDCD4 with no increase in cells from 90-mm dishes, indicating that the expression of PDCD4 was determined by cell density and not cell number (Supplemental Fig. 1).

FIGURE 1.

Cell density–dependent upregulation of PDCD4 expression in macrophages and its reversal by LPS. (A) A total of 4 × 105 RAW264.7 mouse macrophage cells were seeded and, after 24 h of seeding, were treated or not treated with 500 ng/ml LPS. Cells were harvested at indicated time points posttreatment, and PDCD4 protein levels were analyzed by Western blot. GAPDH protein levels were used as loading control (ctrl.). Densitometric quantification of PDCD4 protein bands, normalized to GAPDH protein bands, is included below. The data represent mean ± SD from three independent experiments. *p ≤ 0.05 from 0-h no LPS ctrl.; #p ≤ 0.05, ##p ≤ 0.01, ###p ≤ 0.001 from LPS-treated ctrl. (B) A total of 4 × 105 RAW264.7 cells were seeded on 35-, 60-, and 90-mm-diameter plastic dishes and incubated for 24 h to attain high, medium, and low densities, respectively. Cells were treated or not treated with 500 ng/ml LPS for designated time points and then harvested and analyzed by immunoblotting with PDCD4 and GAPDH Abs. Densitometric quantification of PDCD4 protein bands, normalized to GAPDH protein bands, is included below. The data represent mean ± SD from three independent experiments. *p ≤ 0.05 from 35-mm dish/no LPS; ##p ≤ 0.01, ###p ≤ 0.001 from LPS-treated dishes. (C) RAW264.7 cells were first treated or not treated with LPS for 0 and 24 h (left panel). Both LPS treated and untreated cells were harvested and then reseeded (4 × 105 cells) and kept for 24 and 48 h post-reseeding to attain low and high density, respectively. Cells were then harvested and immunoblotted using anti-PDCD4 and anti-GAPDH Abs (right panel). Representative blot from three independent experiments. (D) RAW264.7 cells were seeded and after 24 h of seeding were transfected with ctrl. siRNA and PDCD4 siRNA. After 24 h of transfection, the cells were seeded in the upper chamber of Transwell inserts. The medium in the bottom chamber contained 10% FBS as chemoattractant. At the indicated time points, the Transwells were removed, the membranes incised out, and the cells on the lower side of the membranes were stained with crystal violet (left panel) and counted. Microscopic images were obtained at original magnification ×20. The data represent mean ± SD of cells counted from three different fields in three independent experiments. *p ≤ 0.05, **p ≤ 0.01 from ctrl. siRNA treatment.

FIGURE 1.

Cell density–dependent upregulation of PDCD4 expression in macrophages and its reversal by LPS. (A) A total of 4 × 105 RAW264.7 mouse macrophage cells were seeded and, after 24 h of seeding, were treated or not treated with 500 ng/ml LPS. Cells were harvested at indicated time points posttreatment, and PDCD4 protein levels were analyzed by Western blot. GAPDH protein levels were used as loading control (ctrl.). Densitometric quantification of PDCD4 protein bands, normalized to GAPDH protein bands, is included below. The data represent mean ± SD from three independent experiments. *p ≤ 0.05 from 0-h no LPS ctrl.; #p ≤ 0.05, ##p ≤ 0.01, ###p ≤ 0.001 from LPS-treated ctrl. (B) A total of 4 × 105 RAW264.7 cells were seeded on 35-, 60-, and 90-mm-diameter plastic dishes and incubated for 24 h to attain high, medium, and low densities, respectively. Cells were treated or not treated with 500 ng/ml LPS for designated time points and then harvested and analyzed by immunoblotting with PDCD4 and GAPDH Abs. Densitometric quantification of PDCD4 protein bands, normalized to GAPDH protein bands, is included below. The data represent mean ± SD from three independent experiments. *p ≤ 0.05 from 35-mm dish/no LPS; ##p ≤ 0.01, ###p ≤ 0.001 from LPS-treated dishes. (C) RAW264.7 cells were first treated or not treated with LPS for 0 and 24 h (left panel). Both LPS treated and untreated cells were harvested and then reseeded (4 × 105 cells) and kept for 24 and 48 h post-reseeding to attain low and high density, respectively. Cells were then harvested and immunoblotted using anti-PDCD4 and anti-GAPDH Abs (right panel). Representative blot from three independent experiments. (D) RAW264.7 cells were seeded and after 24 h of seeding were transfected with ctrl. siRNA and PDCD4 siRNA. After 24 h of transfection, the cells were seeded in the upper chamber of Transwell inserts. The medium in the bottom chamber contained 10% FBS as chemoattractant. At the indicated time points, the Transwells were removed, the membranes incised out, and the cells on the lower side of the membranes were stained with crystal violet (left panel) and counted. Microscopic images were obtained at original magnification ×20. The data represent mean ± SD of cells counted from three different fields in three independent experiments. *p ≤ 0.05, **p ≤ 0.01 from ctrl. siRNA treatment.

Close modal

To determine whether the high-density–dependent expression of PDCD4 was reversible on reduction of cell density, we took cells that have been grown for 24 h and then treated or untreated with LPS for 24 h and reseeded them at low density. The reseeded cells were then grown for 24 and 48 h, to low- and high-density, respectively. The reseeded cells grown for 24 h did not show a high level of PDCD4 expression, although the cells that had been seeded were high PDCD4-expressing cells. However, the cells that were grown for 48 h again showed high PDCD4 expression that was reduced on LPS treatment (Fig. 1C). Together, these data demonstrate that the high-density–dependent expression of PDCD4 is reversible on reduction of cell density and is recovered once the cells reach high density again.

As migration is a major characteristic of macrophages, we determined the effect of high PDCD4 expression on the migration potential of the high-density cells in Transwell assays using 10% serum as the chemoattractant. However, the migration of the high-density cells was significantly enhanced when the cells were transfected with siRNA against PDCD4, even in absence of any stimulus, as observed over a 24-h period (Fig. 1D). These observations showed that high-density cells have low migration potential, which is at least partly due to high expression of PDCD4. Together, these results demonstrated that PDCD4 expression may be considered as a bona fide marker for high macrophage cell density.

The cell density–dependent upregulation of PDCD4 expression can be either due to changes in extracellular matrix (ECM) composition and concentration or due to threshold levels of a secreted factor. To determine whether changes in ECM can influence the high-density–dependent expression of PDCD4, cells were grown to high density in uncoated plates and plates coated with different ECM components fibronectin or gelatin. In all three cases, there was similar upregulation of PDCD4 expression in high-density cells (Fig. 2A). This showed that differences in ECM composition did not affect the expression of PDCD4 in high-density cells.

FIGURE 2.

A secreted factor causes the cell density–dependent upregulation of PDCD4 expression. (A) A total of 4 × 105 RAW 264.7 cells were seeded on culture dishes either uncoated or coated with fibronectin or gelatin and, after 24 h of seeding, were incubated for the indicated time points. Cells were then lysed and immunoblotted with PDCD4 and GAPDH Abs. (B) A total of 4 × 105 RAW 264.7 cells and mouse BMDM were seeded and, after 24 h of seeding, were incubated for 24 h either continuously in the same medium or with changes every 6 h in medium. Cells were then lysed and immunoblotted with PDCD4 and GAPDH Abs. (C) CM from high-density (HD) RAW264.7 and BMDM cells were collected and added to respective cells at low density in three increasing volumes and incubated for 12 h. CM, which has been heat denatured for 15 min at 100°C, was also added in the two higher volumes. Cells were then harvested and PDCD4 and GAPDH protein levels were analyzed by Western blot. (D) CM from HD cells or low-density (LD) cells were added to LD cells and incubated for 12 h. After 6 h of CM addition, actinomycin D (1.5 μg/ml) was added. Heat-inactivated (HI) HD cell CM was also added and similarly incubated. Cells were then lysed and immunoblotted with PDCD4 and GAPDH Abs. (A)–(D) are representative blots from three independent experiments. (E) Total RNA was extracted from the cell lysates obtained in (D), and PDCD4 mRNA levels were estimated using quantitative RT-PCR (qRT-PCR). GAPDH mRNA was used as normalization control. The data represent mean ± SD from three independent experiments. *p ≤ 0.05 from media-treated cells; ##p ≤ 0.01 from HD cell, CM-treated cells. (F) Cells treated for 12 h with LD, HD, and HI HD CM were seeded in Transwell chambers and assayed for 12 h for migration as described in (Fig. 1. The data represent mean ± SD of cells counted from three different fields in three independent experiments. **p ≤ 0.01 from LD CM-treated cells. (G) Cells treated with LD, HD, and HI HD CM in presence of 500 ng/ml LPS were seeded in Transwell chambers and assayed for 12 h for migration as described in (Fig. 1. Microscopic images were obtained at original magnification ×20. The data represent mean ± SD of cells counted from three different fields in three independent experiments. **p ≤ 0.01 from LD CM-treated cells.

FIGURE 2.

A secreted factor causes the cell density–dependent upregulation of PDCD4 expression. (A) A total of 4 × 105 RAW 264.7 cells were seeded on culture dishes either uncoated or coated with fibronectin or gelatin and, after 24 h of seeding, were incubated for the indicated time points. Cells were then lysed and immunoblotted with PDCD4 and GAPDH Abs. (B) A total of 4 × 105 RAW 264.7 cells and mouse BMDM were seeded and, after 24 h of seeding, were incubated for 24 h either continuously in the same medium or with changes every 6 h in medium. Cells were then lysed and immunoblotted with PDCD4 and GAPDH Abs. (C) CM from high-density (HD) RAW264.7 and BMDM cells were collected and added to respective cells at low density in three increasing volumes and incubated for 12 h. CM, which has been heat denatured for 15 min at 100°C, was also added in the two higher volumes. Cells were then harvested and PDCD4 and GAPDH protein levels were analyzed by Western blot. (D) CM from HD cells or low-density (LD) cells were added to LD cells and incubated for 12 h. After 6 h of CM addition, actinomycin D (1.5 μg/ml) was added. Heat-inactivated (HI) HD cell CM was also added and similarly incubated. Cells were then lysed and immunoblotted with PDCD4 and GAPDH Abs. (A)–(D) are representative blots from three independent experiments. (E) Total RNA was extracted from the cell lysates obtained in (D), and PDCD4 mRNA levels were estimated using quantitative RT-PCR (qRT-PCR). GAPDH mRNA was used as normalization control. The data represent mean ± SD from three independent experiments. *p ≤ 0.05 from media-treated cells; ##p ≤ 0.01 from HD cell, CM-treated cells. (F) Cells treated for 12 h with LD, HD, and HI HD CM were seeded in Transwell chambers and assayed for 12 h for migration as described in (Fig. 1. The data represent mean ± SD of cells counted from three different fields in three independent experiments. **p ≤ 0.01 from LD CM-treated cells. (G) Cells treated with LD, HD, and HI HD CM in presence of 500 ng/ml LPS were seeded in Transwell chambers and assayed for 12 h for migration as described in (Fig. 1. Microscopic images were obtained at original magnification ×20. The data represent mean ± SD of cells counted from three different fields in three independent experiments. **p ≤ 0.01 from LD CM-treated cells.

Close modal

We then investigated whether a secreted factor or factors acting in a paracrine manner is responsible for the density-dependent expression of PDCD4. RAW264.7 cells were grown for 24 h to high density or with media changes every 6 h to wash out any secreted factor(s), and PDCD4 expression was checked. There was upregulation of PDCD4 expression in the high-density cells, which was abrogated by the media changes every 6 h, suggesting the involvement of a secreted factor(s) in the upregulation of PDCD4 expression (Fig. 2B). The same observation was also made in mouse BMDM cells treated in a similar manner (Fig. 2B, lower panels). We then collected CM from RAW264.7 or BMDM cells grown to high density for 24 h and added it to less-dense cells. The low-density cells treated with high-density cell CM showed increase in PDCD4 expression in dose-dependent manner in the case of both RAW264.7 and BMDM (Fig. 2C). Furthermore, the increase in PDCD4 expression was partially abrogated on addition of heat-inactivated cell CM, suggesting the role of a heat-labile secreted factor from high-density cells that causes upregulation of PDCD4 expression. The increase in PDCD4 expression was due to enhanced transcription, as treatment with the transcription inhibitor actinomycin D resulted in the decrease in both PDCD4 protein and mRNA levels in cells treated with high-density cell CM (Fig. 2D, 2E).

Finally, we checked the effect of high-density cell CM on migration potential of cells. Treatment with high-density cell CM caused a reduction in cell migration in comparison with treatment with low-density cell CM (Fig. 2F). However, heat-inactivated, high-density cell CM failed to cause a reduction in cell migration. CM from high-density cells treated with LPS showed increased cell migration, which was abrogated by heat inactivation (Fig. 2G). Together, these observations showed the presence of a heat-labile secreted factor in the CM of high-density cells, which caused transcriptional upregulation of PDCD4 expression and a reduction in cell migration.

To identify the secreted factor in the CM of high-density cells that acted as a quorum sensor and inducer of PDCD4 expression, we performed subtractive mass spectrometry of CM from high-density versus low-density cells. Peptides from a total of 38 proteins were found either solely in the CM from high-density cells or were more than 2-fold enriched in the CM from high-density cells in comparison with CM from low-density cells. The sequence mapping of these peptides showed many of them to be of bovine origin, signifying presence of residual proteins from FBS used to culture the cells. However, because of significant homology between bovine and murine protein sequences, many of these peptides also match with murine homologs of the proteins. We did a systematic analysis, first identifying the secretory proteins from this set, thereafter selecting the proteins that are specific to immune cells and then identifying those proteins that are either of murine origin or share more than 90% identity between murine and bovine sequences (Supplemental Table I). Only two proteins, antithrombin-III (AT-III) and GSN, matched these criteria. We then checked for AT-III and GSN mRNA expression in high-density versus low-density cells in absence and presence of LPS stimulation. AT-III mRNA was found to be downregulated in high-density cells compared with low-density cells and did not show any significant change on LPS treatment (Supplemental Fig. 2A). On the contrary, GSN mRNA expression was found to be significantly upregulated in high-density cells and was significantly inhibited on LPS treatment (Supplemental Fig. 2A). These experiments and analyses strongly suggested GSN as the potential quorum-sensing molecule secreted by high-density macrophages.

GSN is a highly conserved intracellular actin-capping protein with a secreted isoform, pGSN that differs by an extra 25-aa N-terminal sequence (28, 29). GSN was first identified in the cytosol of macrophages and later in various cell types and extracellular fluids (30). GSN plays an anti-inflammatory role by binding to inflammatory mediators and enhances pathogen clearance by improved macrophage uptake (31). These characteristics make GSN an excellent potential candidate as a quorum sensor in determining macrophage function and resolution of inflammation. We checked for GSN protein and mRNA expression in high-density RAW264.7 macrophages and found GSN protein and mRNA to be significantly upregulated in high-density cells, which also show enhanced PDCD4 expression (Fig. 3A). We also found significantly increased expression of the mRNA of the secreted form of GSN, pGSN, by using primers specific to the extra 25-aa sequence of pGSN (Fig. 3A, extreme right panel). GSN protein and mRNA expression were also found to be enhanced in high-density BMDM cells (Fig. 3B). LPS treatment downregulated GSN expression as it did for PDCD4 in both RAW264.7 and BMDM cells. Overexpression of exogenous Flag-tagged GSN in low-density cells resulted in enhancement of PDCD4 expression, whereas siRNA-mediated knockdown of GSN in high-density cells resulted in downregulation of PDCD4 expression (Fig. 3C, 3D). We then collected CM from GSN-overexpressing and -knockdown cells and added them to freshly seeded macrophage cells. CM from GSN-overexpressing cells enhanced PDCD4 expression in a dose-dependent manner, whereas CM from GSN-knockdown cells reduced PDCD4 expression (Fig. 3E, 3F). We immunodepleted secreted GSN from the high-density cell CM by immunoprecipitation using an Ab against GSN, and the GSN immunodepleted CM failed to enhance PDCD4 expression on addition to low-density cells, whereas a CM incubated with control IgG showed enhanced PDCD4 expression in low-density cells (Fig. 3G). CM from GSN-overexpressing cells significantly reduced macrophage cell migration, whereas CM from GSN-knockdown cells enhanced cell migration (Fig. 3H). Therefore, CM from cells overexpressing GSN recapitulated the behavior of high-density cells, whereas CM from GSN-knockdown cells recapitulated that of low-density cells, demonstrating the role of GSN as a surrogate for high macrophage density. Finally, to confirm the identity of GSN, we cloned human GSN protein-coding sequence into a bacterial expression vector and expressed and purified the rGSN protein from bacteria (Supplemental Fig. 2B, 2C). The purified rGSN was treated with oxidized l-glutathione to reconstitute the disulfide linkages characteristic of the secreted form of GSN (pGSN) (21). Treatment of low-density cells with purified rGSN protein enhanced PDCD4 expression in a dose-dependent manner, which was abrogated on LPS treatment (Fig. 3I). Treatment with rGSN protein also enhanced GSN expression, supporting its autocrine function as a quorum-sensing autoinducer. Interestingly, we also observed that only the rGSN protein treated with oxidized l-glutathione was able to enhance PDCD4 and GSN expression, whereas the protein that has not been treated similarly failed to do so, confirming the role of the secreted form of GSN (pGSN) as the autoinducer (Fig. 3J). Purified rGSN also significantly reduced migration of RAW264.7 cells and counteracted the effect of LPS on cell migration, recapitulating the effect of CM from high-density cells (Fig. 3K).

FIGURE 3.

Secreted GSN is the quorum sensor in high-density macrophages. (A) RAW264.7 cells were treated or untreated with 500 ng/ml LPS and incubated for 0 and 24 h. Cells were lysed, and the lysates were immunoblotted with PDCD4, GSN, and GAPDH Abs used as loading control (left panel). Densitometric quantification of GSN protein bands, normalized to GAPDH, from three independent experiments, is included (middle panel). Total RNA was extracted from the cell lysates, and GSN mRNA and pGSN mRNA levels were determined by quantitative RT-PCR (qRT-PCR) using mRNA-specific primers (two right panels). GAPDH mRNA was used as normalization control. The data represent mean ± SD from three independent experiments. *p ≤ 0.05 from 0-h, −LPS cells; ##p ≤ 0.01 from 0-h, +LPS cells. (B) BMDM cells were treated or untreated with 500 ng/ml LPS and incubated for 0 and 24 h. Cells were lysed, and the lysates were immunoblotted with PDCD4, GSN, and GAPDH Abs used as loading control (left panel). Total RNA was extracted from the cell lysates, and PDCD4 and GSN mRNA levels were determined by qRT-PCR (two right panels). GAPDH mRNA was used as normalization control. The data represent mean ± SD from three independent experiments. *p ≤ 0.05 from 0-h, −LPS cells; #p ≤ 0.05, ##p ≤ 0.01 from 0-h, +LPS cells. (C) RAW264.7 cells, either mock transfected or transfected with two increasing concentrations of pCDNA3.1-GSN-DYK plasmid DNA were harvested after 30 h of transfection and lysates immunoblotted with FLAG, PDCD4, and GAPDH Abs. (D) RAW264.7 cells, either transfected with control siRNA or with two increasing concentration of GSN siRNA were lysed after 30 h of transfection and lysates immunoblotted with FLAG, PDCD4, and GAPDH Abs. (E) LD RAW264.7 cells were treated with CM collected from cells either mock transfected or transfected with two increasing concentrations of pCDNA3.1-GSN-DYK DNA. After 12 h of incubation in CM, cells were lysed and immunoblotted with PDCD4 and GAPDH Abs. (F) LD RAW264.7 cells were treated with CM collected from cells transfected with either control siRNA or two increasing concentration of GSN siRNA. After 12 h of incubation in CM, cells were lysed and immunoblotted with PDCD4 and GAPDH Abs. (G) LD RAW264.7 cells were treated with either IgG or GSN immunodepleted CM from low-density (LD) and high-density (HD) cells and incubated for 12 h. Cells were then harvested, and PDCD4 and GAPDH protein levels were analyzed by immunoblotting. (H) RAW264.7 cells treated for 12 h with CM collected from cells overexpressing GSN at two increasing concentrations or expressing GSN siRNA at two increasing concentrations were seeded in Transwell inserts, and migration assay was done for 12 h. The data represent mean ± SD of cells counted from three different fields in three independent experiments. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.005 from cells treated with CM from mock-transfected cells. (I) RAW264.7 cells grown at low density were treated without and with three increasing concentrations of recombinant GSN (rGSN) protein that has been oxidized by oxidized glutathione, for 12 h with or without LPS. Cells were then harvested, lysed, and immunoblotted with PDCD4, GSN, and GAPDH Abs. (J) Low-density RAW264.7 cells were treated with glutathione-oxidized and nonoxidized rGSN protein and incubated for 12 h. Cells were then harvested, lysed, and immunoblotted with PDCD4, GSN, and GAPDH Abs. (K) RAW264.7 cells seeded in Transwell inserts were treated or not treated with LPS in combination with and without rGSN protein and migration assay was done for 12 h. Microscopic images were obtained at original magnification ×20. The data represent mean ± SD of cells counted from three different fields in three independent experiments. **p ≤ 0.01, ***p ≤ 0.005 from cells treated without LPS and rGSN. (L) Exosomes were isolated by fractionation from CM of LD and HD RAW264.7 cells. After isolation, exosomes were resuspended in equal volume. An equal volume of each sample was analyzed by Western blotting using GSN and exosomal marker flotillin 1 Abs. (M) Low-density and high-density RAW264.7 cells were treated with or without the exosome inhibitor GW4869 for 24 h. Cells were then harvested, and lysates were immunoblotted using GSN and GAPDH Abs. (N) CM from HD cells treated with or without GW4869 were added to low-density cells and incubated for 12 h. Heat-inactivated CM from GW4869 treated/untreated cells was also added. Cells were then harvested and immunoblotted using PDCD4 and GAPDH Abs. All immunoblotting data are representative blots from three independent experiments.

FIGURE 3.

Secreted GSN is the quorum sensor in high-density macrophages. (A) RAW264.7 cells were treated or untreated with 500 ng/ml LPS and incubated for 0 and 24 h. Cells were lysed, and the lysates were immunoblotted with PDCD4, GSN, and GAPDH Abs used as loading control (left panel). Densitometric quantification of GSN protein bands, normalized to GAPDH, from three independent experiments, is included (middle panel). Total RNA was extracted from the cell lysates, and GSN mRNA and pGSN mRNA levels were determined by quantitative RT-PCR (qRT-PCR) using mRNA-specific primers (two right panels). GAPDH mRNA was used as normalization control. The data represent mean ± SD from three independent experiments. *p ≤ 0.05 from 0-h, −LPS cells; ##p ≤ 0.01 from 0-h, +LPS cells. (B) BMDM cells were treated or untreated with 500 ng/ml LPS and incubated for 0 and 24 h. Cells were lysed, and the lysates were immunoblotted with PDCD4, GSN, and GAPDH Abs used as loading control (left panel). Total RNA was extracted from the cell lysates, and PDCD4 and GSN mRNA levels were determined by qRT-PCR (two right panels). GAPDH mRNA was used as normalization control. The data represent mean ± SD from three independent experiments. *p ≤ 0.05 from 0-h, −LPS cells; #p ≤ 0.05, ##p ≤ 0.01 from 0-h, +LPS cells. (C) RAW264.7 cells, either mock transfected or transfected with two increasing concentrations of pCDNA3.1-GSN-DYK plasmid DNA were harvested after 30 h of transfection and lysates immunoblotted with FLAG, PDCD4, and GAPDH Abs. (D) RAW264.7 cells, either transfected with control siRNA or with two increasing concentration of GSN siRNA were lysed after 30 h of transfection and lysates immunoblotted with FLAG, PDCD4, and GAPDH Abs. (E) LD RAW264.7 cells were treated with CM collected from cells either mock transfected or transfected with two increasing concentrations of pCDNA3.1-GSN-DYK DNA. After 12 h of incubation in CM, cells were lysed and immunoblotted with PDCD4 and GAPDH Abs. (F) LD RAW264.7 cells were treated with CM collected from cells transfected with either control siRNA or two increasing concentration of GSN siRNA. After 12 h of incubation in CM, cells were lysed and immunoblotted with PDCD4 and GAPDH Abs. (G) LD RAW264.7 cells were treated with either IgG or GSN immunodepleted CM from low-density (LD) and high-density (HD) cells and incubated for 12 h. Cells were then harvested, and PDCD4 and GAPDH protein levels were analyzed by immunoblotting. (H) RAW264.7 cells treated for 12 h with CM collected from cells overexpressing GSN at two increasing concentrations or expressing GSN siRNA at two increasing concentrations were seeded in Transwell inserts, and migration assay was done for 12 h. The data represent mean ± SD of cells counted from three different fields in three independent experiments. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.005 from cells treated with CM from mock-transfected cells. (I) RAW264.7 cells grown at low density were treated without and with three increasing concentrations of recombinant GSN (rGSN) protein that has been oxidized by oxidized glutathione, for 12 h with or without LPS. Cells were then harvested, lysed, and immunoblotted with PDCD4, GSN, and GAPDH Abs. (J) Low-density RAW264.7 cells were treated with glutathione-oxidized and nonoxidized rGSN protein and incubated for 12 h. Cells were then harvested, lysed, and immunoblotted with PDCD4, GSN, and GAPDH Abs. (K) RAW264.7 cells seeded in Transwell inserts were treated or not treated with LPS in combination with and without rGSN protein and migration assay was done for 12 h. Microscopic images were obtained at original magnification ×20. The data represent mean ± SD of cells counted from three different fields in three independent experiments. **p ≤ 0.01, ***p ≤ 0.005 from cells treated without LPS and rGSN. (L) Exosomes were isolated by fractionation from CM of LD and HD RAW264.7 cells. After isolation, exosomes were resuspended in equal volume. An equal volume of each sample was analyzed by Western blotting using GSN and exosomal marker flotillin 1 Abs. (M) Low-density and high-density RAW264.7 cells were treated with or without the exosome inhibitor GW4869 for 24 h. Cells were then harvested, and lysates were immunoblotted using GSN and GAPDH Abs. (N) CM from HD cells treated with or without GW4869 were added to low-density cells and incubated for 12 h. Heat-inactivated CM from GW4869 treated/untreated cells was also added. Cells were then harvested and immunoblotted using PDCD4 and GAPDH Abs. All immunoblotting data are representative blots from three independent experiments.

Close modal

As pGSN is reported to be secreted and transported via exosomes (32), we checked for the presence of GSN in exosomal fractions from high-density cells. Exosome fractionation from CM of high- and low-density cells showed the enrichment of GSN in exosomal fractions from high-density cells (Fig. 3L). Treatment with GW4869, an exosome inhibitor, resulted in increased intracellular GSN level in low density, and more significantly, in high-density cells (Fig. 3M). Finally, we collected CM from high-density cells treated with GW4869 and added it to low-density cells. CM from high-density cells untreated with GW4869 caused increased PDCD4 expression as observed before, but CM from GW4869-treated cells failed to enhance PDCD4 expression (Fig. 3N). Heat-inactivated CM from both GW4869 treated and untreated cells failed to enhance PDCD4 expression.

pGSN is also reported to activate p38 MAPK signaling by facilitating the interaction between cell surface integrin α5β1 and b2 gp1 in a paracrine or autocrine manner (32, 33). We therefore also checked whether the upregulation of PDCD4 by CM from high-density cells was mediated by p38 MAPK signaling. Treatment of macrophage cells exposed to CM from high-density cells with a p38 MAPK inhibitor abrogated the increase in PDCD4 expression, whereas treatment with an ERK inhibitor did not cause reduction in PDCD4 expression (Supplemental Fig. 3A). Moreover, pretreatment of the low-density cells with a blocking Ab against integrin α5 subunit of integrin α5β1 receptor prevented the enhancement of PDCD4 expression on treatment with CM from high-density cells (Supplemental Fig. 3B). Together, these data showed that exosomally secreted GSN from high-density cells acts as a quorum sensor and upregulates PDCD4 expression in a cell density–dependent manner.

LPS has been shown to downregulate PDCD4 expression in human PBMCs and also in RAW264.7 mouse macrophages by inducing the miRNA miR-21 (27). As we observed that LPS downregulated the expression of PDCD4 in high-density mouse macrophages, we investigated the mechanism of this downregulation. PDCD4 mRNA level increased with increase in cell density over a 24-h period, but LPS treatment caused a significant downregulation of PDCD4 mRNA (Fig. 4A). LPS treatment also caused a significant enhancement of miR-21 level over this period, whereas increase in cell density did not cause a change in miR-21 expression (Fig. 4B). As miRNA-mediated repression caused mRNA degradation, we checked the stability of PDCD4 mRNA in presence and absence of LPS stimulation. In absence of LPS, PDCD4 mRNA was found to have a half-life over 8 h, whereas LPS treatment reduced the half-life to 7.3 h, signifying mRNA destabilization (Fig. 4C). PDCD4 has also been reported to be targeted for proteasomal degradation in RAW264.7 cells by LPS treatment. We investigated the stability of PDCD4 protein in high-density RAW264.7 cells by treatment with LPS and the protein synthesis inhibitor cycloheximide. LPS induced rapid degradation of PDCD4 protein within 4 h of treatment. However, treatment with the proteasome inhibitor MG132 strongly reduced the LPS-induced degradation of PDCD4, demonstrating the proteasome-mediated degradation of PDCD4 by LPS treatment of high-density macrophages (Fig. 4D).

FIGURE 4.

LPS causes downregulation of PDCD4 expression and reverses the high-density phenotype. (A) RAW264.7 cells were treated or not treated with 500 ng/ml LPS, and cells were harvested at indicated time points. Total RNA was extracted and PDCD4 mRNA levels were determined by quantitative RT-PCR (qRT-PCR). GAPDH mRNA level was used as normalization control. The data represent mean ± SD from three independent experiments. *p ≤ 0.05, ##p ≤ 0.01 from respective 0-h controls. (B) RAW264.7 cells treated as above. Total RNA was extracted and polyadenylated, and mature miR-21 levels were measured by qRT-PCR. U6 snRNA level was used as normalization control. The data represent mean ± SD from three independent experiments. #p ≤ 0.05, ##p ≤ 0.01, ###p ≤ 0.001 from 0-h control. (C) RAW264.7 cells were pretreated with or without LPS (500 ng/ml) for 2 h, and then actinomycin D was added to the cells. Cells were then harvested at the times indicated, and total RNA was extracted for the measurement of PDCD4 mRNA by qRT-PCR as described above. The data represent mean ± SD from three independent experiments. (D) RAW264.7 cells were pretreated or not treated with MG132 (5 μM) for 1 h and pretreated with cycloheximide (100 μg/ml) for 30 min and then treated with LPS (500 ng/ml). Cells were collected at the indicated time points and immunoblotted with PDCD4 and GAPDH Abs. Representative blot from three independent experiments. (E) RAW264.7 cells were treated or untreated with LPS (500 ng/ml) for 24 h. Then, cell viability was measured using MTT assay. The data represent mean ± SD from three independent experiments. ##p ≤ 0.01, difference between 24-h LPS+ and LPS cells. (F) RAW264.7 cells were treated or untreated with LPS (500 ng/ml) for 24 h. Then, apoptotic cells were estimated using caspase-3/7 assay. The data represent mean ± SD from three independent experiments. #p ≤ 0.05, difference between 24 h LPS+ and LPS cells. (G) RAW264.7 cells were treated or untreated with LPS (500 ng/ml) for 24 h. Then, cell cycle analysis was performed by flow cytometry by measuring DNA content using propidium iodide staining. The data represent mean ± SD from three independent experiments. #p ≤ 0.05, difference between 0- and 24-h LPS cells; *p ≤ 0.05, difference between 0- and 24-h LPS+ cells. (H) RAW264.7 cells were treated or untreated with LPS (500 ng/ml) for 24 h. Then, the cells were seeded in Transwell chambers, and Transwell migration assay was done as described earlier. Microscopic images were obtained at original magnification ×20. The data represent mean ± SD from three independent experiments. *p ≤ 0.05, difference between 24 h LPS+ and LPS cells.

FIGURE 4.

LPS causes downregulation of PDCD4 expression and reverses the high-density phenotype. (A) RAW264.7 cells were treated or not treated with 500 ng/ml LPS, and cells were harvested at indicated time points. Total RNA was extracted and PDCD4 mRNA levels were determined by quantitative RT-PCR (qRT-PCR). GAPDH mRNA level was used as normalization control. The data represent mean ± SD from three independent experiments. *p ≤ 0.05, ##p ≤ 0.01 from respective 0-h controls. (B) RAW264.7 cells treated as above. Total RNA was extracted and polyadenylated, and mature miR-21 levels were measured by qRT-PCR. U6 snRNA level was used as normalization control. The data represent mean ± SD from three independent experiments. #p ≤ 0.05, ##p ≤ 0.01, ###p ≤ 0.001 from 0-h control. (C) RAW264.7 cells were pretreated with or without LPS (500 ng/ml) for 2 h, and then actinomycin D was added to the cells. Cells were then harvested at the times indicated, and total RNA was extracted for the measurement of PDCD4 mRNA by qRT-PCR as described above. The data represent mean ± SD from three independent experiments. (D) RAW264.7 cells were pretreated or not treated with MG132 (5 μM) for 1 h and pretreated with cycloheximide (100 μg/ml) for 30 min and then treated with LPS (500 ng/ml). Cells were collected at the indicated time points and immunoblotted with PDCD4 and GAPDH Abs. Representative blot from three independent experiments. (E) RAW264.7 cells were treated or untreated with LPS (500 ng/ml) for 24 h. Then, cell viability was measured using MTT assay. The data represent mean ± SD from three independent experiments. ##p ≤ 0.01, difference between 24-h LPS+ and LPS cells. (F) RAW264.7 cells were treated or untreated with LPS (500 ng/ml) for 24 h. Then, apoptotic cells were estimated using caspase-3/7 assay. The data represent mean ± SD from three independent experiments. #p ≤ 0.05, difference between 24 h LPS+ and LPS cells. (G) RAW264.7 cells were treated or untreated with LPS (500 ng/ml) for 24 h. Then, cell cycle analysis was performed by flow cytometry by measuring DNA content using propidium iodide staining. The data represent mean ± SD from three independent experiments. #p ≤ 0.05, difference between 0- and 24-h LPS cells; *p ≤ 0.05, difference between 0- and 24-h LPS+ cells. (H) RAW264.7 cells were treated or untreated with LPS (500 ng/ml) for 24 h. Then, the cells were seeded in Transwell chambers, and Transwell migration assay was done as described earlier. Microscopic images were obtained at original magnification ×20. The data represent mean ± SD from three independent experiments. *p ≤ 0.05, difference between 24 h LPS+ and LPS cells.

Close modal

We then checked the effect of LPS treatment on the phenotype of high-density macrophages. MTT assay showed the presence of higher cell number in non–LPS-treated macrophages whereas LPS treatment resulted in significantly reduced cell number (Fig. 4E). LPS treatment also caused a significant increase in caspase-3/7 activity, suggesting higher number of apoptotic cells, a characteristic of activated macrophages (Fig. 4F). Cell cycle analysis showed a significantly higher number of cells in the G0/G1 phase in high-density macrophages, which was reduced by LPS treatment with a parallel increase in the number of sub-G0/G1 cells (Fig. 4G). Cells treated with LPS for 24 h showed significantly higher migration potential compared with high-density cells untreated with LPS (Fig. 4H). Together, these observations suggest that the high-density cells have a quiescent, nonmigratory phenotype that is reversed by LPS treatment into an activated, migratory, and apoptotic phenotype.

As LPS treatment induced miR-21 in mouse macrophages and caused downregulation of PDCD4 expression, we then investigated whether GSN itself was a target of miR-21–mediated downregulation. GSN has been reported to be downregulated by miR-21 in the cardiomyocytes of mice with diabetic cardiomyopathy; however, no miR-21 target site in the 3′ UTR was identified (34). Bioinformatic analysis showed the presence of a low affinity target site (nucleotides 192-214) for miR-21 in the 3′ UTR of GSN mRNA (Fig. 5A). LPS treatment of RAW264.7 cells transfected with a firefly luciferase reporter gene construct containing the 236 nt. GSN 3′ UTR downstream of the reporter gene showed a significant dose-dependent reduction in reporter gene activity (Fig. 5B). However, LPS treatment failed to inhibit the reporter gene activity from a reporter construct containing the GSN 3′ UTR in which the putative miR-21 target site has been mutated (Fig. 5B). The reporter gene activity was also rescued when the cells were cotransfected with an antagomiR against miR-21, suggesting that the LPS-induced downregulation of firefly luciferase-GSN 3′ UTR reporter activity was mediated by miR-21 (Fig. 5C). Overexpression of miR-21 from a miR-21 short hairpin RNA–expressing vector caused a dose-dependent reduction in GSN expression (Fig. 5D). The LPS-induced reduction in GSN protein expression in high-density cells was also reversed on antagomiR-21 transfection (Fig. 5E). The induction of miR-21 by LPS is reported to be mediated via the transcription factor NF-κB (27). Treatment of LPS-treated cells with an NF-κB inhibitor PDTC also resulted in the reversal of the LPS-induced downregulation of GSN protein and of PDCD4, which is known to be repressed by miR-21 (Fig. 5F). We confirmed that PDTC treatment reduced the LPS-induced upregulation of miR-21 expression (Fig. 5G) and reversed the LPS-induced downregulation of GSN mRNA level (Fig. 5H). Together, these results suggest that the activation of the LPS–NF-κB–miR-21 axis results in the downregulation of GSN expression on LPS treatment in parallel with the downregulation of PDCD4 expression, leading to the reversal of the high-density phenotype.

FIGURE 5.

LPS-induced upregulation of miR-21 inhibits GSN expression. (A) Schematic representation of Gsn mRNA 3′ UTR showing putative miR-21 target site between nucleotides 192–214 in the 3′ UTR. The underlined sequence shows the mutation in the putative miR-21 target site in the Gsn mRNA 3′ UTR. (B) RAW264.7 cells were transfected with either pCDNA3-Fluc or pCDNA3-Fluc-GSN 3′ UTR or pCDNA3-Fluc-GSN 3′ UTR miR-21 mut reporter gene constructs and after 24 h of transfection, cells were treated with increasing concentrations of LPS for another 24 h. pCMV-Rluc was cotransfected as transfection control. Cells were then harvested, and luciferase activity was measured. Data represent Fluc values normalized to Rluc values, shown as fold increase over non–LPS-treated cells, taken as 1. The data represent mean ± SD from three independent experiments. *p ≤ 0.05, **p ≤ 0.01, difference between LPS-treated and untreated cells transfected with pCDNA3-Fluc-GSN 3′ UTR; #p ≤ 0.05, ##p ≤ 0.01, difference between LPS-treated cells transfected with pCDNA3-Fluc-GSN 3′ UTR miR-21 mut. (C) RAW264.7 cells were transfected with either pCDNA3-Fluc or pCDNA3-Fluc-GSN 3′ UTR reporter gene constructs and cotransfected with or without two increasing concentrations of antagomir-21. After 24 h of transfection, cells were treated with or without LPS (500 ng/ml) for additional 24 h. Cells were then harvested, and luciferase activity was measured. Data represent Fluc values normalized to Rluc values, shown as fold increase over non–LPS-treated cells, taken as 1. The data represent mean ± SD from two independent experiments. *p ≤ 0.05, difference between antagomiR-21–transfected and –untransfected cells. (D) RAW264.7 cells, either mock transfected or transfected with two increasing concentrations of pSUP–miR-21 plasmid, were lysed after 36 h of transfection, and lysates were immunoblotted with GSN and GAPDH Abs (upper panel). Total RNA was extracted from the cell lysates and polyadenylated, and miRNA-21 levels were determined by quantitative RT-PCR (qRT-PCR). U6 snRNA was used as normalization control (lower panel). The data represent mean ± SD from two independent experiments. *p ≤ 0.05, difference between pSUP–miR-21–transfected and –untransfected cells. (E) RAW 264.7 cells were transfected with or without increasing concentration of antagomir-21 and treated with or without LPS (500 ng/ml) for 24 h. Cells were then lysed and immunoblotted using GSN and GAPDH Abs. (F) RAW264.7 cells were pretreated or not treated with 20 μM of PDTC for 30 min and then stimulated for 24 h with or without 500 ng/ml LPS. Cells were then harvested, and lysates immunoblotted with GSN, PDCD4, and GAPDH Abs. (D)–(F) are representative blots from three independent experiments. (G and H) RAW264.7 cells were pretreated or not pretreated with 20 μM of PDTC for 30 min and then stimulated for 24 h with or without 500 ng/ml LPS. Total RNA was extracted, and miR-21 and GSN mRNA levels were determined by qRT-PCR as described earlier. The data represent mean ± SD from three independent experiments. **p ≤ 0.001, ***p ≤ 0.001, #p ≤ 0.05, difference between LPS-treated and untreated cells.

FIGURE 5.

LPS-induced upregulation of miR-21 inhibits GSN expression. (A) Schematic representation of Gsn mRNA 3′ UTR showing putative miR-21 target site between nucleotides 192–214 in the 3′ UTR. The underlined sequence shows the mutation in the putative miR-21 target site in the Gsn mRNA 3′ UTR. (B) RAW264.7 cells were transfected with either pCDNA3-Fluc or pCDNA3-Fluc-GSN 3′ UTR or pCDNA3-Fluc-GSN 3′ UTR miR-21 mut reporter gene constructs and after 24 h of transfection, cells were treated with increasing concentrations of LPS for another 24 h. pCMV-Rluc was cotransfected as transfection control. Cells were then harvested, and luciferase activity was measured. Data represent Fluc values normalized to Rluc values, shown as fold increase over non–LPS-treated cells, taken as 1. The data represent mean ± SD from three independent experiments. *p ≤ 0.05, **p ≤ 0.01, difference between LPS-treated and untreated cells transfected with pCDNA3-Fluc-GSN 3′ UTR; #p ≤ 0.05, ##p ≤ 0.01, difference between LPS-treated cells transfected with pCDNA3-Fluc-GSN 3′ UTR miR-21 mut. (C) RAW264.7 cells were transfected with either pCDNA3-Fluc or pCDNA3-Fluc-GSN 3′ UTR reporter gene constructs and cotransfected with or without two increasing concentrations of antagomir-21. After 24 h of transfection, cells were treated with or without LPS (500 ng/ml) for additional 24 h. Cells were then harvested, and luciferase activity was measured. Data represent Fluc values normalized to Rluc values, shown as fold increase over non–LPS-treated cells, taken as 1. The data represent mean ± SD from two independent experiments. *p ≤ 0.05, difference between antagomiR-21–transfected and –untransfected cells. (D) RAW264.7 cells, either mock transfected or transfected with two increasing concentrations of pSUP–miR-21 plasmid, were lysed after 36 h of transfection, and lysates were immunoblotted with GSN and GAPDH Abs (upper panel). Total RNA was extracted from the cell lysates and polyadenylated, and miRNA-21 levels were determined by quantitative RT-PCR (qRT-PCR). U6 snRNA was used as normalization control (lower panel). The data represent mean ± SD from two independent experiments. *p ≤ 0.05, difference between pSUP–miR-21–transfected and –untransfected cells. (E) RAW 264.7 cells were transfected with or without increasing concentration of antagomir-21 and treated with or without LPS (500 ng/ml) for 24 h. Cells were then lysed and immunoblotted using GSN and GAPDH Abs. (F) RAW264.7 cells were pretreated or not treated with 20 μM of PDTC for 30 min and then stimulated for 24 h with or without 500 ng/ml LPS. Cells were then harvested, and lysates immunoblotted with GSN, PDCD4, and GAPDH Abs. (D)–(F) are representative blots from three independent experiments. (G and H) RAW264.7 cells were pretreated or not pretreated with 20 μM of PDTC for 30 min and then stimulated for 24 h with or without 500 ng/ml LPS. Total RNA was extracted, and miR-21 and GSN mRNA levels were determined by qRT-PCR as described earlier. The data represent mean ± SD from three independent experiments. **p ≤ 0.001, ***p ≤ 0.001, #p ≤ 0.05, difference between LPS-treated and untreated cells.

Close modal

Macrophages have been classified into M1 or M2 phenotypes based on their functional characteristics and gene expression programs in which M1 (classically activated) macrophages exhibit inflammatory functions, whereas M2 (alternatively activated) macrophages exhibit anti-inflammatory functions (6). To investigate the gene expression program underlying the high-density phenotype of macrophages, we performed a global transcriptomic analysis of high-density versus low-density cells and high-density versus LPS-treated high-density cells. The transcriptomes of RAW264.7 cells at low density and high density were sequenced and compared for differential gene expression. Also, the transcriptome of LPS-treated high-density cells was sequenced and compared with that of high-density cells for differential gene expression. A total of 801 genes were upregulated, and 454 genes were downregulated in high-density cells in comparison with low-density cells (Fig. 6A). Both GSN and PDCD4 expression was upregulated in high-density cells, with GSN showing a log2 fold change of 1.98 and PDCD4 showing a log2 fold change of 1.56. A number of other genes that are markers of the macrophage M2 phenotype, such as transglutaminase 2 (TGM2) and IL-4R, were found to be upregulated in high-density cells, whereas markers for the macrophage M1 phenotype, such as CXCL10, were found to be downregulated. GO analysis of the upregulated genes showed enrichment of genes involved in positive regulation of macroautophagy, negative regulation of cell proliferation, negative regulation of transcription, negative regulation of cell migration, and positive regulation of vascular endothelial growth factor signaling and angiogenesis (Fig. 6B). The downregulated genes were enriched in extrinsic apoptotic signaling pathway, cellular response to extracellular stimulus and innate immune response pathways. This supports the quiescent, nonmigratory phenotype of the high-density cells observed in our experiments and suggests an anti-inflammatory gene expression program resembling the M2 phenotype of macrophages.

FIGURE 6.

High-density cells show a differential gene expression program compared with low-density cells. (A) Volcano plot representing change in gene expression in high-density versus low-density cells. The data are represented as log10 fold change in statistical significance (p value) against log2 fold change in gene expression. Representative upregulated and downregulated genes, including PDCD4 and GSN, are indicated. (B) GO analysis of upregulated (upper) and downregulated (lower) genes in gene expression profiling by RNAseq of high-density versus low-density cells. (C) Volcano plot representing change in gene expression in high-density versus LPS-treated high-density cells. The data are represented as log10 fold change in statistical significance (p value) against log2 fold change in gene expression. Representative upregulated and downregulated genes, including PDCD4 and GSN, are indicated. (D) GO analysis of upregulated (upper) and downregulated (lower) genes in gene expression profiling by RNAseq of high-density cells treated with LPS versus high-density cells.

FIGURE 6.

High-density cells show a differential gene expression program compared with low-density cells. (A) Volcano plot representing change in gene expression in high-density versus low-density cells. The data are represented as log10 fold change in statistical significance (p value) against log2 fold change in gene expression. Representative upregulated and downregulated genes, including PDCD4 and GSN, are indicated. (B) GO analysis of upregulated (upper) and downregulated (lower) genes in gene expression profiling by RNAseq of high-density versus low-density cells. (C) Volcano plot representing change in gene expression in high-density versus LPS-treated high-density cells. The data are represented as log10 fold change in statistical significance (p value) against log2 fold change in gene expression. Representative upregulated and downregulated genes, including PDCD4 and GSN, are indicated. (D) GO analysis of upregulated (upper) and downregulated (lower) genes in gene expression profiling by RNAseq of high-density cells treated with LPS versus high-density cells.

Close modal

Differential gene expression analysis of the LPS-treated high-density cells compared with the high-density cells showed the upregulation of 1155 genes and the downregulation of 1122 genes (Fig. 6C). Both GSN and PDCD4 expression was downregulated, with GSN showing a log2 fold change of −2.06 and PDCD4 showing a log2 fold change of −1.72. A number of markers of the M1 macrophage phenotype such as CXCL10 (log2 fold change of 8.31) and CXCL11 (log2 fold change of 2.79) and other inflammatory markers such as Mmp9 and NOS2 were upregulated. GO analysis of the upregulated genes showed a high enrichment of genes involved in the inflammatory response and immune system processes and downregulation of genes involved in cell division and cell adhesion (Fig. 6D), signifying the switch of the high-density phenotype into an activated, inflammatory, and migratory phenotype.

As the gene expression program of the high-density cells showed changes in expression of a number of genes characteristic of the M2 phenotype, whereas LPS treatment altered it to resemble the gene expression program similar to the M1 phenotype, we therefore determined the changes in expression of a set of genes that are markers for the M1 and M2 phenotypes in these cells. Expression of the cytokine TNF-α and chemokine CXCL10, which are markers for the M1 phenotype, were significantly downregulated in the high-density cells, but were strongly upregulated on LPS treatment (Fig. 7A). CXCL11, another marker for the M1 phenotype, did not show appreciable change in expression in the high-density cells but was upregulated on LPS treatment. Interestingly, inducible NO synthase (iNOS), another characteristic marker for the M1 phenotype, showed increased mRNA expression in the high-density cells and a much higher expression on LPS treatment. However, estimation of relative NO level showed no increase in the high-density cells but around 8-fold increase in activity in high-density cells treated with LPS. This suggested posttranscriptional regulation of iNOS expression or activity in the high-density cells leading to no change in NO production (Fig. 7A).

FIGURE 7.

High density induces M2-like gene expression program in macrophages, which is reversed by LPS. (A) RAW264.7 cells were treated or not treated with LPS (500 ng/ml) for 0 and 24 h. Total RNA was extracted and levels of TNF-α, CXCL10, CXCL11, and iNOS mRNAs were determined by quantitative RT-PCR (qRT-PCR). GAPDH mRNA level was used as normalization control. NO level was measured by Griess assay. The data represent mean ± SD from three independent experiments. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, #p ≤ 0.05, ##p ≤ 0.01, ###p ≤ 0.001, $p ≤ 0.05, $$p ≤ 0.01, $$$p ≤ 0.001, difference between 24-h LPS-untreated and LPS-treated cells, between 24- and 0-h LPS-untreated cells, and between 24- and 0-h LPS-treated cells. (B) RAW264.7 cells were treated or not treated with LPS (500 ng/ml) for 0 and 24 h. Total RNA was extracted and levels of TGM2, MRC1, ARG1, IL-4R, and CD163 mRNAs were determined by qRT-PCR. GAPDH mRNA level was used as normalization control. The data represent mean ± SD from three independent experiments. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, #p ≤ 0.05, ##p ≤ 0.01, $p ≤ 0.05, $$p ≤ 0.01, $$$p ≤ 0.001, difference between 24-h LPS-untreated and LPS-treated cells, between 24- and 0-h LPS-untreated cells and between 24- and 0-h LPS-treated cells, respectively.

FIGURE 7.

High density induces M2-like gene expression program in macrophages, which is reversed by LPS. (A) RAW264.7 cells were treated or not treated with LPS (500 ng/ml) for 0 and 24 h. Total RNA was extracted and levels of TNF-α, CXCL10, CXCL11, and iNOS mRNAs were determined by quantitative RT-PCR (qRT-PCR). GAPDH mRNA level was used as normalization control. NO level was measured by Griess assay. The data represent mean ± SD from three independent experiments. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, #p ≤ 0.05, ##p ≤ 0.01, ###p ≤ 0.001, $p ≤ 0.05, $$p ≤ 0.01, $$$p ≤ 0.001, difference between 24-h LPS-untreated and LPS-treated cells, between 24- and 0-h LPS-untreated cells, and between 24- and 0-h LPS-treated cells. (B) RAW264.7 cells were treated or not treated with LPS (500 ng/ml) for 0 and 24 h. Total RNA was extracted and levels of TGM2, MRC1, ARG1, IL-4R, and CD163 mRNAs were determined by qRT-PCR. GAPDH mRNA level was used as normalization control. The data represent mean ± SD from three independent experiments. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, #p ≤ 0.05, ##p ≤ 0.01, $p ≤ 0.05, $$p ≤ 0.01, $$$p ≤ 0.001, difference between 24-h LPS-untreated and LPS-treated cells, between 24- and 0-h LPS-untreated cells and between 24- and 0-h LPS-treated cells, respectively.

Close modal

Conversely, mRNA levels of TGM2, mannose receptor C type 1 (MRC1), IL-4R, and the scavenger receptor CD163, markers for the M2 phenotype were upregulated in the high-density cells and strongly downregulated upon LPS stimulation (Fig. 7B). Interestingly, mRNA level of arginase 1 (Arg1), the enzyme which generates ornithine from arginine and is a marker for the M2 phenotype, was downregulated in the high-density cells and was further downregulated on LPS stimulation. Together, these results suggest that the high-density macrophage population has an M2-like gene expression program and activity, with some crucial differences especially in the NO metabolism pathway, whereas stimulation with LPS polarizes them into a decidedly M1-like phenotype. Therefore, macrophages at high density acquire a distinct phenotype in the M1–M2 phenotypic spectrum, characterized by PDCD4 expression and secretion of GSN as an autoinducer.

Recent studies have postulated several immune regulatory processes akin to quorum sensing in bacteria in the mammalian immune system (8, 9). These processes were shown to be elicited at sites of inflammation only when a sufficient cell density was achieved and led to alterations of cell behavior at the population level (16). Moreover, these processes involved diffusive molecules that might function as autoinducers in that they would be secreted in sufficiently low amount by single cells but would achieve a threshold level on secretion by a high number of cells such as to cause alterations in gene expression and cellular function (8). However, the identity of bona fide markers of cell density and secreted molecules that can function as autoinducers has mostly remained elusive. To our knowledge, in this study we identified a novel mechanism of sensing of macrophage cell density in vitro, with the inflammatory tumor suppressor gene PDCD4 as a marker of high macrophage cell density and secreted GSN as the quorum-sensing autoinducer. The high-density cells attain a quiescent, anti-inflammatory phenotype marked by reduced cell migration potential. Furthermore, we found that the inflammatory agonist, LPS, could inhibit the expression of these molecules in the high-density cells and thereby disrupt the quorum-sensing mechanism and revert the cells to an activated, inflammatory phenotype. These observations accord well with a mechanism that senses the accumulation of a sufficient number of macrophages as a signal to elicit the resolution of inflammation.

It is interesting that PDCD4 was found as a marker for high macrophage cell density. PDCD4 is a tumor suppressor gene that is known to regulate apoptosis in response to inflammatory stimuli (35). PDCD4 is induced in macrophages by inflammatory stimuli and functions by activating NF-κB and inhibiting IL-10 (27). Mice lacking PDCD4 were protected from LPS-induced death. PDCD4 can facilitate high-fat–induced atherosclerosis by negatively regulating the expression of anti-inflammatory cytokine IL-10 in macrophages in an ERK1/2- and p38-dependent manner (36). However, PDCD4 also plays a number of anti-inflammatory roles such as in inhibiting inflammation in acute liver injury and acute colitis mouse models (36). Overexpression of PDCD4 in rat alveolar macrophages induces the expression of markers of the anti-inflammatory M2 macrophage phenotype (37). Downregulation of PDCD4 in RAW264.7 macrophages also causes enhanced LPS-induced expression of inflammatory cytokines (38). Therefore, PDCD4 appears to play a crucial role in maintaining the balance between inflammation and its resolution, and mechanisms that regulate PDCD4 expression have inbuilt control systems that fine-tune its expression in presence and absence of inflammatory signals. Therefore, PDCD4 is appropriately poised to play a role in determining the activation status of inflammatory cells, and its regulation in response to macrophage cell density would contribute to its role in the mechanism of inflammation resolution. The function of PDCD4 in contributing to the quiescent anti-inflammatory phenotype might be mediated by its role as a translation repressor either by binding to the translation initiation factor eIF4a or by interacting with structured 5′ UTRs of specific mRNAs (39). LPS treatment has been shown to repress PDCD4 expression by inducing miR-21 and causing proteasomal degradation (27). We observed that LPS treatment of the high-density macrophages caused induction of miR-21 and reduction of PDCD4 mRNA stability, together with the proteasomal degradation of PDCD4. As the control of PDCD4 expression has been postulated as a key step in the negative regulation of the inflammatory response to LPS, the LPS-mediated downregulation of PDCD4 via the induction of miR-21 in the high-density macrophages, therefore, constitutes a mechanism for the initiation of a proinflammatory state. miR-21 is upregulated in many inflammatory conditions, and given its role in tumorigenesis, might be an important link between chronic inflammation and cancer (19).

The discovery of GSN as a quorum sensor and autoinducer in the high-density macrophages is remarkable, as it accords a novel function to this actin-capping protein and also ties in with its known anti-inflammatory roles. GSN is a multifunctional protein known for its actin filament severing, capping, and nucleating activity, resulting in the remodeling of cytoskeletal structure, and ultimately impacting cell shape and movement (40). Increasingly, attention has been drawn to the functions of its secreted isoform pGSN, which differs by a 25-aa N-terminal extension (29). pGSN was known to mainly function as a part of the extracellular actin scavenger system responsible for the removal of actin filaments released from dead cells into the bloodstream (41). However, an expanding role of pGSN has been found in apoptosis, signal transduction, transcriptional regulation, and modulation of inflammatory responses (40). pGSN modulates immune responses by preferential binding to bacterial cell wall–derived compounds such as LPS and lipoteichoic acid and preventing TLR activation (41). GSN limits the inflammatory process induced by LPS by reducing the levels of inflammatory cytokines IL-6, TNF-α, and NO released by macrophages and protects mice against LPS-induced death (42). Indeed, administration of pGSN in the form of recombinant human GSN has been shown to have beneficial effects in a variety of preclinical models of inflammation and injury, such as severe influenza and primary pneumococcal pneumonia (43, 44). Remarkably, investigation of immune based biomarkers in patients of COVID-19, the disease caused by SARS-CoV-2 infection, which is frequently marked by a hyperinflammatory response (45), has shown a correlation of decreased pGSN levels with disease severity and mortality (46). This has led to the proposition of recombinant human GSN as a potential therapeutic against COVID-19. To our knowledge, our discovery of pGSN as a quorum sensor in macrophages provides new insights into its mode of function as a crucial regulator of macrophage activity and inflammation.

It is not clear how GSN impacts the effect of LPS on macrophages beyond its direct role in preventing LPS binding with the TLR4 receptor. However, pGSN has also been shown to be a signaling molecule via its exosomal release and facilitating the interaction of β2 gp1 with the fibronectin receptor α5β1 integrin and activating the p38 MAPK pathway (32, 33). We also see that GSN is secreted exosomally by high-density macrophages and enhances PDCD4 expression via the p38 MAPK pathway. Therefore, the independent function of pGSN as the quorum sensor in high-density macrophages appears to be based on its function as a paracrine and autocrine signaling molecule. GSN expression is upregulated, together with that of other genes mainly associated with the anti-inflammatory M2 phenotype in high-density macrophages (47). This is reversed by LPS treatment, which triggers an M1-like phenotypic switch in these macrophages. Based on these observations we propose a model in which pGSN released from the high-density macrophages act as a quorum sensor and autoinducer, which upregulates PDCD4 expression, contributing to a quiescent anti-inflammatory phenotype (Fig. 8A). LPS treatment upregulated miR-21 by an NF-κB–dependent mechanism, downregulating both GSN and PDCD4, giving rise to a coherent feed-forward loop (C-FFL) (Fig. 8B) and reversing the anti-inflammatory phenotype to an activated proinflammatory phenotype. C-FFLs constitute sign-sensitive delay elements that act as persistence detectors, which cause changes in gene expression only in response to persistent signals and protect against brief input fluctuations (48). Hence, the C-FFL comprising miR-21, GSN, and PDCD4 is expected to allow inflammatory activation of macrophages only in presence of persistent inflammatory signals. The miR21-GSN-PDCD4 regulatory network, therefore, plays a crucial role in the macrophage quorum-sensing mechanism, allowing it to temporally control macrophage number and activity for the fine-tuning of the inflammatory response.

FIGURE 8.

The miR21-GSN-PDCD4 regulatory network plays a crucial role in the macrophage quorum-sensing mechanism. (A) Proposed model showing release of pGSN as a quorum sensor and autoinducer in high-density macrophages, which upregulates PDCD4 expression in a paracrine manner and results in an anti-inflammatory M2-like phenotype. LPS treatment induces miR-21, which represses expression of both GSN and PDCD4, and polarizes the macrophages into an inflammatory M1-like phenotype. (B) Network diagram showing C-FFL composed of miR-21, GSN, and PDCD4 with two input signals, LPS, and high density.

FIGURE 8.

The miR21-GSN-PDCD4 regulatory network plays a crucial role in the macrophage quorum-sensing mechanism. (A) Proposed model showing release of pGSN as a quorum sensor and autoinducer in high-density macrophages, which upregulates PDCD4 expression in a paracrine manner and results in an anti-inflammatory M2-like phenotype. LPS treatment induces miR-21, which represses expression of both GSN and PDCD4, and polarizes the macrophages into an inflammatory M1-like phenotype. (B) Network diagram showing C-FFL composed of miR-21, GSN, and PDCD4 with two input signals, LPS, and high density.

Close modal

We thank Prof. Paul L. Fox, The Lerner Research Institute, Cleveland Clinic, Ohio, and Prof. Peter H. King, UAB School of Medicine, Birmingham, Alabama, for providing reagents and allowing the use of laboratory facilities; Dr. Malancha Ta, Soma Jana, and Dr. Partha P. Datta, Indian Institute of Science Education and Research Kolkata, for providing reagents; and Uttam Das, Jawaharlal Nehru Memorial College and Hospital, Kalyani, India, for help with an assay. We thank Prof. Luke O’Neill and Dr. Frederick Sheedy, Trinity College, Dublin for helpful discussions.

This work was supported by The Wellcome Trust DBT India Alliance intermediate fellowship (WT500139/Z/09/Z) and a Science and Engineering Research Board extramural research grant (EMR/2016/003525 [to P.S.R.]), a U.G.C. senior research fellowship to R.K.S., a DST INSPIRE senior research fellowship to B.G., and a C.S.I.R. senior research fellowship to S.D.M. The Orbitrap Elite instrument (Proteomics and Metabolomics Core, LRI, Cleveland Clinic) was purchased via National Institutes of Health shared instrument grant 1S10RR031537-01.

Conceptualization, P.S.R. and R.K.S.; investigation, R.K.S., B.G., A.G., and B.W.; analysis, R.K.S., S.D.M., and P.S.R.; writing, P.S.R. and R.K.S.; supervision, P.S.R.; funding acquisition, P.S.R.

The mass spectrometry data presented in this article have been submitted to ProteomeXchange Consortium via the PRIDE partner repository (https://www.ebi.ac.uk/pride/archive/) under accession numbers PXD021978 and 10.6019/PXD021978 and the gene expression data have been submitted to National Center for Biotechnology Information’s Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) under accession number GSE159158.

The online version of this article contains supplemental material.

Abbreviations used in this article

     
  • AT-III

    antithrombin-III

  •  
  • BMDM

    bone marrow–derived macrophage

  •  
  • C-FFL

    coherent feed-forward loop

  •  
  • CM

    conditioned medium

  •  
  • ECM

    extracellular matrix

  •  
  • GO

    gene ontology

  •  
  • GSN

    gelsolin

  •  
  • iNOS

    inducible NO synthase

  •  
  • miRNA

    microRNA

  •  
  • PDCD4

    programmed cell death 4

  •  
  • PDTC

    pyrrolidine dithiocarbonate

  •  
  • pGSN

    plasma GSN

  •  
  • rGSN

    recombinant GSN

  •  
  • RNAseq

    RNA sequencing

  •  
  • siRNA

    small interfering RNA

  •  
  • snRNA

    small nuclear RNA

  •  
  • TGM2

    transglutaminase 2

  •  
  • UAB

    University of Alabama at Birmingham

  •  
  • UTR

    untranslated region

1.
Sugimoto
M. A.
,
L. P.
Sousa
,
V.
Pinho
,
M.
Perretti
,
M. M.
Teixeira
.
2016
.
Resolution of inflammation: what controls its onset?
Front. Immunol.
7
:
160
.
2.
Nathan
C.
2002
.
Points of control in inflammation.
Nature
420
:
846
852
.
3.
Ortega-Gómez
A.
,
M.
Perretti
,
O.
Soehnlein
.
2013
.
Resolution of inflammation: an integrated view.
EMBO Mol. Med.
5
:
661
674
.
4.
Nathan
C.
,
A.
Ding
.
2010
.
Nonresolving inflammation.
Cell
140
:
871
882
.
5.
Stables
M. J.
,
S.
Shah
,
E. B.
Camon
,
R. C.
Lovering
,
J.
Newson
,
J.
Bystrom
,
S.
Farrow
,
D. W.
Gilroy
.
2011
.
Transcriptomic analyses of murine resolution-phase macrophages.
Blood
118
:
e192
e208
.
6.
Shapouri-Moghaddam
A.
,
S.
Mohammadian
,
H.
Vazini
,
M.
Taghadosi
,
S. A.
Esmaeili
,
F.
Mardani
,
B.
Seifi
,
A.
Mohammadi
,
J. T.
Afshari
,
A.
Sahebkar
.
2018
.
Macrophage plasticity, polarization, and function in health and disease.
J. Cell. Physiol.
233
:
6425
6440
.
7.
Greten
F. R.
,
S. I.
Grivennikov
.
2019
.
Inflammation and cancer: triggers, mechanisms, and consequences.
Immunity
51
:
27
41
.
8.
Postat
J.
,
P.
Bousso
.
2019
.
Quorum sensing by monocyte-derived populations.
Front. Immunol.
10
:
2140
.
9.
Antonioli
L.
,
C.
Blandizzi
,
P.
Pacher
,
M.
Guilliams
,
G.
Haskó
.
2018
.
Quorum sensing in the immune system.
Nat. Rev. Immunol.
18
:
537
538
.
10.
Waters
C. M.
,
B. L.
Bassler
.
2005
.
Quorum sensing: cell-to-cell communication in bacteria.
Annu. Rev. Cell Dev. Biol.
21
:
319
346
.
11.
Antonioli
L.
,
C.
Blandizzi
,
P.
Pacher
,
M.
Guilliams
,
G.
Haskó
.
2019
.
Rethinking communication in the immune system: the quorum sensing concept.
Trends Immunol.
40
:
88
97
.
12.
Zhou
X.
,
R. A.
Franklin
,
M.
Adler
,
J. B.
Jacox
,
W.
Bailis
,
J. A.
Shyer
,
R. A.
Flavell
,
A.
Mayo
,
U.
Alon
,
R.
Medzhitov
.
2018
.
Circuit design features of a stable two-cell system.
Cell
172
:
744
757.e17
.
13.
Dai
X. M.
,
G. R.
Ryan
,
A. J.
Hapel
,
M. G.
Dominguez
,
R. G.
Russell
,
S.
Kapp
,
V.
Sylvestre
,
E. R.
Stanley
.
2002
.
Targeted disruption of the mouse colony-stimulating factor 1 receptor gene results in osteopetrosis, mononuclear phagocyte deficiency, increased primitive progenitor cell frequencies, and reproductive defects.
Blood
99
:
111
120
.
14.
Chen
C. C.
,
L.
Wang
,
M. V.
Plikus
,
T. X.
Jiang
,
P. J.
Murray
,
R.
Ramos
,
C. F.
Guerrero-Juarez
,
M. W.
Hughes
,
O. K.
Lee
,
S.
Shi
, et al
2015
.
Organ-level quorum sensing directs regeneration in hair stem cell populations.
Cell
161
:
277
290
.
15.
Muldoon
J. J.
,
Y.
Chuang
,
N.
Bagheri
,
J. N.
Leonard
.
2020
.
Macrophages employ quorum licensing to regulate collective activation.
Nat. Commun.
11
:
878
.
16.
Postat
J.
,
R.
Olekhnovitch
,
F.
Lemaître
,
P.
Bousso
.
2018
.
A metabolism-based quorum sensing mechanism contributes to termination of inflammatory responses.
Immunity
49
:
654
665.e5
.
17.
Jiang
Y.
,
Y.
Jia
,
L.
Zhang
.
2017
.
Role of programmed cell death 4 in diseases: a double-edged sword.
Cell. Mol. Immunol.
14
:
884
886
.
18.
Wang
T.
,
S.
Long
,
N.
Zhao
,
Y.
Wang
,
H.
Sun
,
Z.
Zou
,
J.
Wang
,
X.
Ran
,
Y.
Su
.
2015
.
Cell density-dependent upregulation of PDCD4 in keratinocytes and its implications for epidermal homeostasis and repair.
Int. J. Mol. Sci.
17
:
8
.
19.
Poria
D. K.
,
A.
Guha
,
I.
Nandi
,
P. S.
Ray
.
2016
.
RNA-binding protein HuR sequesters microRNA-21 to prevent translation repression of proinflammatory tumor suppressor gene programmed cell death 4.
Oncogene
35
:
1703
1715
.
20.
Aida
Y.
,
M. J.
Pabst
.
1990
.
Removal of endotoxin from protein solutions by phase separation using Triton X-114.
J. Immunol. Methods
132
:
191
195
.
21.
Wen
D.
,
K.
Corina
,
E. P.
Chow
,
S.
Miller
,
P. A.
Janmey
,
R. B.
Pepinsky
.
1996
.
The plasma and cytoplasmic forms of human gelsolin differ in disulfide structure.
Biochemistry
35
:
9700
9709
.
22.
Weischenfeldt
J.
,
B.
Porse
.
2008
.
Bone marrow-derived macrophages (BMM): Isolation and applications.
CSH Protoc
2008
:
pdb.prot5080
.
23.
Kacem
M.
,
G.
Simon
,
R.
Leschiera
,
L.
Misery
,
A.
ElFeki
,
N.
Lebonvallet
.
2015
.
Antioxidant and anti-inflammatory effects of Ruta chalepensis L. extracts on LPS-stimulated RAW 264.7 cells.
In Vitro Cell. Dev. Biol. Anim.
51
:
128
141
.
24.
Trapnell
C.
,
A.
Roberts
,
L.
Goff
,
G.
Pertea
,
D.
Kim
,
D. R.
Kelley
,
H.
Pimentel
,
S. L.
Salzberg
,
J. L.
Rinn
,
L.
Pachter
.
2012
.
Differential gene and transcript expression analysis of RNA-seq experiments with TopHat and Cufflinks. [Published erratum appears in 2014 Nat. Protoc. 9: 2513.]
Nat. Protoc.
7
:
562
578
.
25.
Trapnell
C.
,
D. G.
Hendrickson
,
M.
Sauvageau
,
L.
Goff
,
J. L.
Rinn
,
L.
Pachter
.
2013
.
Differential analysis of gene regulation at transcript resolution with RNA-seq.
Nat. Biotechnol.
31
:
46
53
.
26.
Huang
W.
,
B. T.
Sherman
,
R. A.
Lempicki
.
2009
.
Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources.
Nat. Protoc.
4
:
44
57
.
27.
Sheedy
F. J.
,
E.
Palsson-McDermott
,
E. J.
Hennessy
,
C.
Martin
,
J. J.
O’Leary
,
Q.
Ruan
,
D. S.
Johnson
,
Y.
Chen
,
L. A.
O’Neill
.
2010
.
Negative regulation of TLR4 via targeting of the proinflammatory tumor suppressor PDCD4 by the microRNA miR-21.
Nat. Immunol.
11
:
141
147
.
28.
Kwiatkowski
D. J.
,
T. P.
Stossel
,
S. H.
Orkin
,
J. E.
Mole
,
H. R.
Colten
,
H. L.
Yin
.
1986
.
Plasma and cytoplasmic gelsolins are encoded by a single gene and contain a duplicated actin-binding domain.
Nature
323
:
455
458
.
29.
Yin
H. L.
,
D. J.
Kwiatkowski
,
J. E.
Mole
,
F. S.
Cole
.
1984
.
Structure and biosynthesis of cytoplasmic and secreted variants of gelsolin.
J. Biol. Chem.
259
:
5271
5276
.
30.
Bucki
R.
,
I.
Levental
,
A.
Kulakowska
,
P. A.
Janmey
.
2008
.
Plasma gelsolin: function, prognostic value, and potential therapeutic use.
Curr. Protein Pept. Sci.
9
:
541
551
.
31.
Lee
P. S.
,
A. B.
Waxman
,
K. L.
Cotich
,
S. W.
Chung
,
M. A.
Perrella
,
T. P.
Stossel
.
2007
.
Plasma gelsolin is a marker and therapeutic agent in animal sepsis.
Crit. Care Med.
35
:
849
855
.
32.
Asare-Werehene
M.
,
K.
Nakka
,
A.
Reunov
,
C. T.
Chiu
,
W. T.
Lee
,
M. R.
Abedini
,
P. W.
Wang
,
D. B.
Shieh
,
F. J.
Dilworth
,
E.
Carmona
, et al
2020
.
The exosome-mediated autocrine and paracrine actions of plasma gelsolin in ovarian cancer chemoresistance.
Oncogene
39
:
1600
1616
.
33.
Bohgaki
M.
,
M.
Matsumoto
,
T.
Atsumi
,
T.
Kondo
,
S.
Yasuda
,
T.
Horita
,
K. I.
Nakayama
,
F.
Okumura
,
S.
Hatakeyama
,
T.
Koike
.
2011
.
Plasma gelsolin facilitates interaction between β2 glycoprotein I and α5β1 integrin.
J. Cell. Mol. Med.
15
:
141
151
.
34.
Dai
B.
,
H.
Li
,
J.
Fan
,
Y.
Zhao
,
Z.
Yin
,
X.
Nie
,
D. W.
Wang
,
C.
Chen
.
2018
.
MiR-21 protected against diabetic cardiomyopathy induced diastolic dysfunction by targeting gelsolin.
Cardiovasc. Diabetol.
17
:
123
.
35.
Lankat-Buttgereit
B.
,
R.
Göke
.
2009
.
The tumour suppressor Pdcd4: recent advances in the elucidation of function and regulation.
Biol. Cell
101
:
309
317
.
36.
Jiang
Y.
,
Q.
Gao
,
L.
Wang
,
C.
Guo
,
F.
Zhu
,
B.
Wang
,
Q.
Wang
,
F.
Gao
,
Y.
Chen
,
L.
Zhang
.
2016
.
Deficiency of programmed cell death 4 results in increased IL-10 expression by macrophages and thereby attenuates atherosclerosis in hyperlipidemic mice.
Cell. Mol. Immunol.
13
:
524
534
.
37.
Zhong
B.
,
X.
Yang
,
Q.
Sun
,
L.
Liu
,
X.
Lan
,
J.
Tian
,
Q.
He
,
W.
Hou
,
H.
Liu
,
C.
Jiang
, et al
2014
.
Pdcd4 modulates markers of macrophage alternative activation and airway remodeling in antigen-induced pulmonary inflammation.
J. Leukoc. Biol.
96
:
1065
1075
.
38.
Yasuda
M.
,
T.
Schmid
,
D.
Rübsamen
,
N. H.
Colburn
,
K.
Irie
,
A.
Murakami
.
2010
.
Downregulation of programmed cell death 4 by inflammatory conditions contributes to the generation of the tumor promoting microenvironment.
Mol. Carcinog.
49
:
837
848
.
39.
Biyanee
A.
,
J.
Ohnheiser
,
P.
Singh
,
K.-H.
Klempnauer
.
2015
.
A novel mechanism for the control of translation of specific mRNAs by tumor suppressor protein Pdcd4: inhibition of translation elongation.
Oncogene
34
:
1384
1392
.
40.
Piktel
E.
,
I.
Levental
,
B.
Durnaś
,
P. A.
Janmey
,
R.
Bucki
.
2018
.
Plasma gelsolin: indicator of inflammation and its potential as a diagnostic tool and therapeutic target.
Int. J. Mol. Sci.
19
:
2516
.
41.
Wang
H.
,
B.
Cheng
,
Q.
Chen
,
S.
Wu
,
C.
Lv
,
G.
Xie
,
Y.
Jin
,
X.
Fang
.
2008
.
Time course of plasma gelsolin concentrations during severe sepsis in critically ill surgical patients.
Crit. Care
12
:
R106
.
42.
Cheng
Y.
,
X.
Hu
,
C.
Liu
,
M.
Chen
,
J.
Wang
,
M.
Wang
,
F.
Gao
,
J.
Han
,
C.
Zhang
,
D.
Sun
,
R.
Min
.
2017
.
Gelsolin inhibits the inflammatory process induced by LPS.
Cell. Physiol. Biochem.
41
:
205
212
.
43.
Kobzik
L.
,
Z.
Yang
,
A.
Bedugnis
,
S.
Levinson
,
M.
DiNubile
,
T.
Stossel
,
Q.
Lu
.
2019
.
Delayed administration of recombinant plasma gelsolin improves survival in a murine model of severe influenza.
F1000 Res.
8
:
1860
.
44.
Yang
Z.
,
T. T. Y.
Chiou
,
T. P.
Stossel
,
L.
Kobzik
.
2015
.
Plasma gelsolin improves lung host defense against pneumonia by enhancing macrophage NOS3 function.
Am. J. Physiol. Lung Cell. Mol. Physiol.
309
:
L11
L16
.
45.
Ray
P. S.
,
B.
Goswami
.
2020
.
COVID-19 and hyperinflammatory syndrome.
Indian J. Biochem. Biophys.
57
:
662
669
.
46.
Abers
M. S.
,
O. M.
Delmonte
,
E. E.
Ricotta
,
J.
Fintzi
,
D. L.
Fink
,
A. A. A.
de Jesus
,
K. A.
Zarember
,
S.
Alehashemi
,
V.
Oikonomou
,
J. V.
Desai
, et al
NIAID COVID-19 Consortium
.
2021
.
An immune-based biomarker signature is associated with mortality in COVID-19 patients.
JCI Insight
6
:
e144455
.
47.
Sica
A.
,
A.
Mantovani
.
2012
.
Macrophage plasticity and polarization: in vivo veritas.
J. Clin. Invest.
122
:
787
795
.
48.
Mangan
S.
,
U.
Alon
.
2003
.
Structure and function of the feed-forward loop network motif.
Proc. Natl. Acad. Sci. USA
100
:
11980
11985
.
49.
Perez-Riverol
Y.
,
A.
Csordas
,
J.
Bai
,
M.
Bernal-Llinares
,
S.
Hewapathirana
,
D. J.
Kundu
,
A.
Inuganti
,
J.
Griss
,
G.
Mayer
,
M.
Eisenacher
, et al
2019
.
The PRIDE database and related tools and resources in 2019: improving support for quantification data.
Nucleic Acids Res.
47
(
D1
):
D442
D450
.
50.
Barrett
T.
,
S. E.
Wilhite
,
P.
Ledoux
,
C.
Evangelista
,
I. F.
Kim
,
M.
Tomashevsky
,
K. A.
Marshall
,
K. H.
Phillippy
,
P. M.
Sherman
,
M.
Holko
, et al
2013
.
NCBI GEO: archive for functional genomics data sets--update.
Nucleic Acids Res.
41
(
D1
):
D991
D995
.

The authors have no financial conflicts of interest.

Supplementary data