ABSTRACT
We have analyzed the gene expression profile of monocytes in response to a highly purified cell wall fraction of Mycobacterium bovis BCG, a clinically approved adjuvant known as BCG cell wall skeleton (BCG-CWS). It is composed of mycolic acid, arabinogalactan, and peptidoglycan and confers Toll-like receptor 2 (TLR2)- and TLR4-dependent signaling that induces monocytes to differentiate into antigen-presenting cells (APCs). Here we report differential gene expression analysis with BCG-CWS-stimulated versus nonstimulated monocytes. BCG-CWS exerted massive induction of genes regulated by TLR signaling. Marked gene regulatory characteristics in BCG-CWS-stimulated cells compared to lipopolysaccharide (LPS)-stimulated cells follow. (i) Spliced mRNAs encoding soluble forms of TREM-1 and TREM-2 (recently discovered inflammatory-signal-amplifying receptors) were regulated by BCG-CWS, resulting in their differential expression. (ii) The genes for zinc-iron transporter protein (ZIP)-like family proteins HKE-1 and LIV-1 were induced exclusively by BCG-CWS. (iii) Interleukin-23 (IL-23), rather than IL-12p70, was induced by BCG-CWS, while interferon-inducible genes were induced only by LPS. By Northern and reverse transcription-PCR analyses, we confirmed the differential expression of more than 30 BCG-CWS regulatory genes, and their expression was compared with that of LPS and other known TLR ligands. A battery of genes responded rapidly and for a short time to LPS but for a long time to BCG-CWS. Structural analysis of the identified novel or hypothetical proteins revealed that some are potential candidates as signaling mediators or transcriptional regulators. Hence, BCG-CWS may profoundly modulate APC responses in a way distinct from that of LPS, leading to possible advantages for its adjuvant-active therapeutic potential.
Microbial components can exert immunostimulatory and toxic effects in vivo. The immunomodulatory properties of mycobacterial constituents, including the cell wall and CpG DNA, have long been known as adjuvants and successfully applied in cancer immunotherapy and efficient vaccination (4, 23, 46). However, the mechanisms by which microbial adjuvants are recognized by the immune system had remained obscure until identification of the Toll-like receptor (TLR) family (35) and rapid advances in the field of innate immunity. In recent years, a series of microbial constituents have been identified as specific TLR ligands and designated pathogen-associated molecular patterns (PAMPs) (53). Gram-negative bacterial lipopolysaccharide (LPS) is a TLR4 ligand, and peptidoglycan (PGN) and lipopeptides of gram-positive bacteria are ligands for TLR2. Specific molecular patterns are crucial for the TLR-ligand function since the CpG motif of nonmethylated DNA, poly(I-C) (representing viral double-stranded RNA), lipid-containing peptides, and synthetic adjuvants like imidazoquinoline activate TLRs (35, 53).
Our previous study demonstrated that the induction of tumor necrosis factor alpha (TNF-α) production by Mycobacterium bovis BCG cell wall skeleton (BCG-CWS) was impaired in macrophages derived from both TLR2 and TLR4 knockout mice, which indicated the involvement of TLR2 and -4 pathways in the effects of BCG-CWS (50). All TLR receptors bear an interleukin-1 (IL-1) receptor-like cytoplasmic signaling domain, Toll-interleukin receptor homology domain (TIR), but their signaling domains are not identical and recruit different sets of adapter molecules (26, 41, 55). Their intracellular signaling cascades and cross talk are just beginning to be elucidated. Furthermore, TLRs work in a combinatorial manner, thereby recognizing diverse ligands and probably also eliciting complex intracellular signaling (2, 28, 53). Thus, the analysis of immune cells' transcriptional responses toward structurally defined microbial ligands is currently of growing interest to clarify adjuvant activity and its signaling pathways (17, 27, 48, 54). Since BCG-CWS is responsible for part of the BCG adjuvant activity and lacks LPS-like toxic potential in a dose effectively inducing immunostimulatory activity (4, 23), it has been used postoperatively for adjuvant immunotherapy to treat patients with cancer. It also offered us an opportunity to evaluate the properties of a useful compound, PAMP.
Mechanistic studies of BCG-CWS have suggested that the PGN portion is recognized by human TLR2 and TLR4 and is the main portion responsible for cytokine induction (52), whereas the carbohydrate moiety at the linker region presumably interacts with an as-yet-unidentified lectin receptor that mediates BCG-CWS internalization (51). Moreover, the presence of the mycolic acid portion in its structure makes it a likely candidate for presentation by CD1 molecules during their cell surface sampling. In addition, various PGN recognition proteins and binding proteins including NOD proteins were recently reported (18, 30). Thus, BCG-CWS binds multiple receptors and exerts some effects that are overlapping and others that are distinct from those of LPS on antigen-presenting cells (APCs).
In order to view the effect at the gene expression level, we conducted differential screening in a BCG-CWS-monocyte interaction model based on subtraction technology (suppression-subtractive hybridization [SSH]), cDNA array analysis, and mRNA differential display (DDRT)-PCR. These techniques enabled us to identify rare and novel transcripts (6, 37) rather than to estimate the global gene expression profile. We also evaluated the gene products specifically regulated by BCG-CWS, because they might represent the as-yet-undefined effector components downstream of a PAMP-TLR axis, which may help us to understand therapeutic adjuvant activity and the vital link between innate and adaptive immunity (2, 4, 35, 45, 53).
MATERIALS AND METHODS
Cells and reagents.The following materials were obtained as indicated: fetal calf serum (FCS) from BioWhittaker (Walkersville, Md.), Dulbecco's modified Eagle medium and RPMI 1640 medium from Gibco BRL (Rockville, Md.), granulocyte-macrophage colony-stimulating factor and IL-4 from Pepro Tech EC (London, United Kingdom), and LPS (Escherichia coli O127:B8) from Difco Laboratories (Detroit, Mich.). Macrophage-activating lipopeptide 2 (MALP-2) and oligodeoxynucleotides (ODN) were made as reported previously (39, 52). Lipoteichoic acid, PGN, and zymosan were purchased from Sigma Chemical Co. (St. Louis, Mo.), and their functional purity was tested as described previously (39).
BCG-CWS was prepared as described previously (50). This reagent contains virtually no trace of LPS and is specific for TLR2 and -4, and the purity of this lot was described elsewhere (52). Since BCG-CWS is insoluble in water and organic solvents, oil-in-water emulsion forms of BCG-CWS particles were used throughout this study. Dried powder of BCG-CWS was suspended at a concentration of 1 mg/ml in emulsion buffer (phosphate-buffered saline containing 1% Drakeol and 1% Tween 80) with a Potter homogenizer and then sterilized by heating for 30 min at 60°C (50).
All of the human materials used in this study were approved by the committee of our institute. Peripheral blood mononuclear cells were prepared from 400 ml of citrate-phosphate-dextrose-supplemented human blood by methylcellulose sedimentation and density gradient centrifugation with Ficoll-Paque (Amersham Pharmacia Biotech AB, Piscataway, N.J.) (50). Monocytes were isolated from peripheral blood mononuclear cells with a magnetic cell sorting system with anti-CD14 antibody-coated microbeads (MiltenyiBiotec, Gladbach, Germany) (50). Cells were used after 1 day of culture in RPMI 1640 medium containing 10% FCS. Immature dendritic cells (iDCs) were generated from monocytes (5 × 105/ml) by culturing for 6 days in RPMI 1640 medium (10 ml) supplemented with 10% heat-inactivated FCS in the presence of 500 IU of human granulocyte-macrophage colony-stimulating factor per ml and 100 IU of human IL-4 per ml (50). iDCs were prepared by the criteria of surface markers, CD14− CD40+ CD83− CD80low CD86low CD1a+. In contrast, BCG-CWS-treated mature DCs had a CD40+ CD83+ CD80high CD86high profile.
Where indicated, iDCs (105/ml) were exposed to 100 ng of LPS per ml, 10 μg of lipoteichoic acid per ml, 15 μg of PGN per ml, 3 × 106 zymosans per ml, 100 nM MALP-2, 50 μg of poly(I-C) per ml, 2 μM ODN (CpG or GpC), or 15 μg of BCG-CWS per ml in emulsion buffer (1% Drakeol, 1% Tween 80). These amounts of PAMPs are functionally equivalent in terms of TNF-α induction in human DCs (39, 52).
SSH.SSH was performed with a PCR-select subtraction kit (Clontech) in accordance with the manufacturer's instructions (16). Two batches of 2 × 108 monocytes were prepared by elutriation, and half of the cells in each batch were stimulated with BCG-CWS for either 4 or 8 h. The other half of each batch was treated with an equal volume of emulsion buffer (50) as a control. Following stimulation, the total RNA was isolated with the TRIzol Reagent (Life Technologies, Inc.) RNA extraction procedure, and poly(A)+ RNA was purified with an oligo(dT) cellulose column (Pharmacia). Complementary DNAs were synthesized from 2 μg of poly(A)+ RNA from BCG-CWS-stimulated [tester or BCG-CWS(+) cDNA] and nonstimulated [driver or BCG-CWS(−) cDNA] monocyte RNAs, respectively. This tester-versus-driver combination was used for forward subtraction to obtain BCG-CWS-induced genes and vice versa for reverse subtraction. Following RsaI digestion of the cDNAs, two adapters, A1 and 2R, were ligated to the tester cDNA. Adapter-ligated tester cDNA was hybridized with excess driver cDNA for 6 h, and then differential cDNAs were selectively amplified by suppression PCR. PCR products of forward and reverse subtractions were ligated into the pCRII vector (TA cloning kit; Invitrogen). Two forward-subtracted libraries (BCG-CWS stimulation for 4 and 8 h) and one reverse-subtracted (nonstimulated) library were constructed. More than 100 clones from each forward-subtracted library were randomly picked up and subjected to differential hybridization (with a PCR-select differential screening kit [Clontech]) with subtracted and nonsubtracted probes. Positive clones were sequenced from both ends with the respective adapter primers.
Gene array hybridization and analysis.Atlas human cancer cDNA expression arrays were purchased from Clontech. The complete gene list with the GenBank accession numbers for the array can be obtained at the Clontech website. Random-primed probes were synthesized from the forward- and reverse-subtracted cDNAs following SSH (8-h BCG-CWS induction set). Equal amounts of column-purified, [α-32P]dCTP-labeled probes (106 cpm/ml) were applied to individual blots prehybridized with Express Hyb buffer (Clontech) for 1 h at 68°C. Hybridization was carried out overnight at 68°C, and after a high-stringency wash, membranes were subjected to autoradiography. The hybridization dot intensity was measured with the National Institutes of Health Image Analyzer. The results for the BCG-up-regulated genes were confirmed by GeneChip analysis of BCG-CWS-stimulated versus nonstimulated monocytes prepared from four independent individuals. At least one of the four samples was up-regulated in response to BCG-CWS in terms of >80% genes collected by the array method (data not shown).
DDRT-PCR.BCG-CWS-stimulated and -nonstimulated total RNAs were prepared as described above, and DDRT-PCR was performed basically as previously described (7, 29). Briefly, 2-μg samples of DNase I-treated total RNA from control and stimulated monocytes, respectively, were reverse transcribed with Superscript-II (Life Technologies, Inc.) with either an arbitrary primer (for random-primed DDRT-PCR) or a degenerate-anchored oligo(dT) primer (T12MG, T12MA, T12MT, or T12MC, where M represents dA, dC, or dG). The reverse-transcribed product (2 μl) was amplified by PCR either in the presence of an arbitrary primer only or in the presence of the anchored primer and one arbitrary primer. Arbitrary oligomer primers were either selected from RNAmap kits I and II (Gene Hunter Corporation, Nashville, Tenn.) or randomly chosen from various 10-mer oligonucleotides available in our laboratory. PCR amplification was performed for 40 cycles (94°C for 30 s, 40°C for 2 min, 72°C for 30 s) with Takara PCR kit components (Takara, Tokyo, Japan) and [α-32P]dCTP (1 μCi) as described in the instructions for the RNAmap kit. In order to perform secondary screening, gel-extracted, differentially expressed fragments were amplified by PCR and spotted onto nylon membranes at a fixed concentration. Duplicate membranes were prepared and hybridized with cDNA probes prepared from BCG-CWS-stimulated and -nonstimulated RNAs. Positive PCR fragments obtained from this secondary screening were sequenced directly or following TA cloning. We initially isolated ∼300 bands, and in a secondary screening, 20% of these genes showed reproducible results, although 30% of these genes appeared to encode hypothetical proteins or showed no homology with any known genes.
Database search and bioinformatics.Sequence similarity database searches were performed with the gapped BLAST2 program (http://www.ncbi.nlm.nih.gov/BLAST ) against the nonredundant sequence database nr. The cDNA sequences that did not show matches with nr database entries were further analyzed against an expressed sequence tag database. Informative gene fragments were then subjected to virtual contig generation, blastx, and PSI blast analyses, respectively. For signal sequence determination and transmembrane (TM) domain predictions, multiple programs from the Expasy site were used. Protein domain searching was performed with the nprotein domain database ProDom 99.1 (http://protein.toulouse.inra.fr/prodom.html ).
Northern blotting and RT-PCR analyses.Northern blot analysis was performed as described previously (7, 37). In the reverse transcription (RT)-PCR analysis, all PCRs were performed with a Takara Taq polymerase PCR kit and a Perkin-Elmer PCR machine (model 9600). DNase I-treated total RNA (2 μg) was used for the synthesis of cDNA with a first-strand cDNA synthesis kit (Life Technologies-BRL) with oligo(dT) as the primer. The cDNA reaction product was diluted 1:25 for PCR with various primer sets, and 10-μl aliquots of the product were analyzed by electrophoresis on gels. Multiple sets of monocyte RNAs were prepared at different times following PAMP stimulation, and induction of BCG-CWS or LPS was verified by standard cytokine profiling. Three sets of such RNA samples were then chosen for subsequent analysis of other genes. The primer pairs used for this study are shown in Table 1.
Primer sets used for this study
Real-time PCR analysis.RNA was isolated with an RNeasy kit (Qiagen, Hilden, Germany). A cDNA template for quantitative real-time PCR analysis was then synthesized, and PCR was performed with an iCycler thermocycler (Bio-Rad, Hercules, Calif.) (17). Small integral membrane protein of lysosome/late endosome (SIMPLE), IL-12p40, IL-12p35, IL-23p19, and IP-10 primers were made as follows: SIMPLE forward, TTCCAGGACCTTACCAGGCG; SIMPLE reverse, TCACAAGCCCCGTAGTTGC; IL-12p40 forward, AAGGAGGCGAGGTTCTAAGC; IL-12p40 reverse, CCAGCAGGTGAAACGTCCAG; IL-12p35 forward, GCTCCAGAAGGCCAGACAAACTC; IL-12p35 reverse, AGGCCAGGCAACTCCCATTAG; IL-23p19 forward, GTTCCCCATATCCAGTGTGG; IL-23p19 reverse, GACTGAGGCTTGGAATCTGC; IP-10 forward, CCACGTGTTGAGATCATTGC; IP-10 reverse, TCCCCTCTGGTTTTAAGGAG.
ELISA.Enzyme-linked immunosorbent assay (ELISA) plates for determination of human IL-1β (R&D Systems, St. Paul, Minn.), IP-10 (R&D Systems), IL-12p40 (TECHNE Corporation, St. Paul, Minn.), TNF-α (Amersham Bioscience, Amersham, United Kingdom), beta interferon (IFN-β; TFB, Tokyo, Japan), and IL-6 (Pierce, Chicago, Ill.) were purchased. BCG-CWS-stimulated and nonstimulated monocytes were prepared as described above and allowed to stand for 24 h in medium. Cells (106) were then transferred to a 24-well plate and cultured for 24 h at 37°C in RPMI 1640 medium-10% FCS. Protein levels of the cytokines in the supernatants were determined by ELISA as described previously (50).
RESULTS
Strategy used to identify BCG-CWS regulatory genes in human monocytes.Two screening strategies, a PCR-based subtraction in combination with solution hybridization technique (SSH) (16) and DDRT (29), and cDNA array blot analysis were used to analyze the gene expression profiles of BCG-CWS-stimulated versus unstimulated monocytes. In the first two, the levels of mRNA from both of the sources were compared side by side and the differentially expressed samples were further analyzed. Both forward and reverse subtractions were conducted to obtain fractions enriched with BCG-CWS-inducible and -suppressed mRNAs, respectively. We designated the forward-subtracted mRNA species BCG-CWS(+) and the down-regulated or reverse-subtracted mRNA species BCG-CWS(−) and subsequently used them for probe preparation and library construction because our main aim was to identify novel inducible genes. We analyzed only those genes that became positive in both of the two rounds of screening following SSH-DDRT-PCR, and their differential expression profiles were confirmed by Northern or RT-PCR analysis. About 100 BCG-CWS regulatory genes were collected, 22 of which were unidentified or hypothetical genes in the database. We tested whether these genes respond to LPS, a representative toxic PAMP, to discriminate BCG-CWS-specific regulatory genes.
Atlas cDNA array analysis with subtracted probes.A pair of Atlas human cDNA arrays containing a set of 588 test genes were hybridized with BCG-CWS(+)- and BCG-CWS(−)-subtracted probes, and the hybridization results of each blot were presented as six successive blocks in two parallel rows (Fig. 1A). Each block contains 98 genes, and they can be classified into six broad functional groups (Fig. 1A). The signal levels obtained with housekeeping gene sets for control are summarized in an inset table in Fig. 1. Except for phospholipase A2, the positive and negative control gene sets included showed equivalent signals with both sets of probes, confirming the use of equal amounts of probes and the equally efficient subtraction of those genes.
Identification of BCG-CWS-regulated genes in monocytes with Atlas cDNA arrays. (A) Hybridization of two identical gene array membranes with radiolabeled BCG-CWS(−) and BCG-CWS(+) cDNA probes. The probes were the BCG-CWS-induced and -suppressed cDNA fractions, respectively, and were obtained by SSH. The autoradiographic image of each membrane is presented as six blocks (numbered 1 to 6), and the type of probe used for each is shown on the left. Numbers below the blot indicate the genes whose signals are pointed out by arrows or arrowheads. These numbers correspond to the gene numbers in the bar graph. PLA2, phospholipase A2; MHC, major histocompatibility complex; Ribo.prot., ribosomal protein. (B) Quantitative representation of the gene array analysis. The genes that were exclusively detected with the plus (+) or minus (−) probe are shown first (numbered 1 to 49), and the genes that were detected with both probes are shown next (numbered 50 to 65). The intensities of the signals determined with the plus and minus probes were plotted and are indicated by filled and open bars, respectively. The genes for vimentin (no. 52) and IFN-γ receptor accessory factor (no. 65) are marked by arrowheads in blocks 1 and 6, respectively, and the two highly intense signals in block 2 are for the caspase 10 gene. Four exclusively inducible genes, those for CDK inhibitor Waf/p21 (no. 2), STAT1 (no. 10), TMP21 (no. 17), and EB1 (no. 30), are indicated by diagonal arrows on the plus probe blot (lower part of panel A, blocks 1, 2, 3, and 5, respectively), and the corresponding regions with no signals are also indicated in the minus probe blot (upper part of panel A). Expression of signaling intermediates EPS8 (no. 19) and zyxin (no. 26) was detected only with the BCG-CWS(−) probe and is shown by upward-pointing arrows in blocks 3 and 4, respectively.
The relative expression levels of 65 genes (Fig. 1B) show the presence of clearly differentially expressed genes. As we used fractions of BCG-CWS-induced or -suppressed mRNAs as probes, it was theoretically expected that the genes that hybridized with the forward-subtracted probes would not be detected with the reverse-subtracted probes. We succeeded in demonstrating such a result for 75% of the total genes with signals detected on the membrane. On this basis, we split the plot into two halves, the first of which shows the expression of the 49 genes that were exclusively induced (signal only with plus probe) or suppressed (signal only with minus probe), and they are represented by filled and empty bars, respectively. Finally, we confirmed the BCG-CWS regulatory profiles of these genes by GeneChip analysis and found that in at least one of the four samples, >80% of these genes are regulated similarly by the two methods (data not shown). The other half of the plot shows the relative expression levels of 16 genes for which we detected signals on both of the blots, indicating that those mRNA species were not efficiently removed during the subtraction procedure or that their basal expression level was relatively high. Direct visual comparison of the upper and lower panels reveals that the genes in categories 1, 2, 5, and 6 were the most markedly affected by the BCG-CWS stimulation, and these genes encode mainly cell cycle regulators, apoptosis-related proteins, regulators of cell-cell interaction or cell invasion, growth factors, and cytokines, respectively.
Rescreening of differentially expressed genes in subtracted libraries.Figure 2A shows a typical result of the postsubtraction differential screening of a batch of 90 inducible clones (out of 300 clones in the forward-subtracted library) with minus and plus probes. As we expected, most of the clones were positively regulated and showed stronger signals with the BCG-CWS(+) probe, whereas the same set of clones in blot I failed to hybridize with the reverse-subtracted (minus) probe enriched with down-regulated mRNA species. We picked up 120 clones from this screening, and on the basis of the sequencing of these clones, we divided them into 40 different genes. Despite the abundance of IL-1β- or IL-8-like molecules, a number of detected fragments were not included in the commercial cDNA array analyzed, and some of them appeared to be completely novel fragments. Next, we examined several genes by a series of Northern blot assays to confirm the differential expression, including that of the fragments of unknown genes. The results of the Northern hybridization of 15 selected genes from our combined subtraction and differential display work are shown in Fig. 2B. We confirmed their up-regulation in a time course study after treatment with BCG-CWS. Although the data are not shown, the genes for stress-inducible gene hop (STIP), (25), EB1 (44), GRIM-19 (3), TMP21 (8), ZFM (12), MIP-1 (13), Id2 (47), IDO (36), and TTP (31), which are known to be important for induction of an innate immune response, were also up-regulated in monocytes and DCs in response to BCG-CWS. Notably, however, most of the BCG-CWS regulatory genes were induced not only by BCG-CWS but also by LPS (Fig. 2C). We next tried to pick out BCG-CWS-specific properties from this gene collection.
Differential screening and confirmation of differential expression. (A) A forward-subtracted library with putative BCG-CWS-inducible clones was initially constructed. Clones were selected randomly from the library, and the inserts were amplified by PCR; 90 PCR fragments were spotted per membrane in duplicate (blots I and II) and further hybridized with the subtracted cDNA probes. A typical result of this differential screening is shown in which most of the candidate inducible clones were positive with the plus probe. (B) Northern blot analysis to confirm the differential expression of 15 genes in response to BCG-CWS treatment. Total RNA was isolated from freshly prepared monocytes stimulated with BCG-CWS for 8 h. Lanes containing induced and noninduced RNAs are indicated by plus and minus signs, respectively, above the blots. Gene names are shown below the blot; F-38 and F-73 are as-yet-unidentified genes. EF-1β, elongation factor 1β; Trxp, thioredoxin peroxidase; Ref-1, redox factor 1. (C) RT-PCR analysis of vimentin, WAF1, and tristetraprolin mRNAs in monocytes induced by BCG-CWS or LPS. Total RNA was isolated at the indicated times and subjected to RT-PCR along with appropriate controls. Amplification of the actin housekeeping gene from the same samples confirms the equivalence of the amounts of cDNA.
DAP12-associated receptors are transcriptional targets of BCG-CWS.We found that BCG-CWS-inducible genes include some expressed sequence tag-matched fragments encoding parts of lectin and immunoglobulin (Ig) superfamily proteins such as TREM-1 (NKp44-like myeloid Ig superfamily receptor) (39), MDL (myeloid DAP12-associated type C lectin receptor) (5), and CRACC (CD2 Ig family receptor) (10). TREM and MDL receptors (5, 9) are expressed specifically on myeloid cells and associated with DAP12, an immune receptor tyrosine-based activation motif-bearing signaling molecule. Comparative RT-PCR analyses were performed with BCG-CWS- and LPS-stimulated monocytes (Fig. 3). In this analysis, we measured the levels of an LPS-inducible lectin, MINCLE (32), and those of DAP12, which is required for the cell surface expression of TREM and MDL. MINCLE was induced only in LPS-stimulated cells, and DAP12 was not induced by either of the cell wall components, although it was abundantly expressed in monocytes. In contrast, the receptors whose cell surface expression was dependent on DAP12 were up-regulated in BCG-CWS- or LPS-activated cells.
Differential regulation of Ig superfamily genes and lectin receptors by bacterial cell wall. RT-PCR analysis was performed with mRNA isolated from monocytes stimulated with BCG-CWS or LPS for the indicated times. TREM-1, together with a related variant form (shown by an arrow) was up-regulated gradually, while TREM-2 was down-regulated. Induction of the type C lectin MDL by BCG-CWS was prominent during the late period, and moderate induction was detected in the case of LPS by 2 h. DAP12 with no induction can be considered an internal control. CRACC gene expression was time dependent in response to the PAMP. The experiments were performed four times with different time frames. Representative results obtained with two individual samples (panels A and B) are shown.
TREM-1 and MDL were more robustly induced by BCG-CWS than by LPS. The LPS-mediated induction was rapidly increased and then soon down-regulated, while the BCG-CWS-mediated induction was slower but was sustained longer. In contrast, CRACC, an NK cell-expressing cytotoxic receptor (10), was induced in monocytes with similar kinetics in response to both of these cell wall components.
Identification of TREM-1 and -2 variants by differential display.TREM-1 induction by BCG-CWS was accompanied with an additional lower band. The two TREM-1 bands were observed in BCG-CWS- but not LPS-stimulated monocytes (Fig. 3A and B). Sequence analysis suggested the presence of a BCG-inducible variant transcript for the TREM-1 receptor, most likely representing the naturally occurring soluble form. This variant was generated in BCG-CWS-activated macrophages and DCs (Fig. 4).
Identification of non-TM isoforms of TREM. (A) Portion of a differential display showing a weak band (*) present only in the BCG-CWS-untreated monocyte RNA lane minus. (B) Schematic representation showing the approximate position of the identified DDRT-PCR fragment (*) compared to that of the reported TREM-2 transcript (GenBank accession no. NM_018965 ). The positions of the two primers used for the RT-PCR, P1 and P2, are also indicated above the map. The black bar shows the 411-bp product derived from TREM-2, and the gray bar indicates the 217-bp product (identical in sequence to the DDRT-PCR [DD-RT] fragment). orf, open reading frame. In lanes 1, 2, 3, and 4, RNAs from monocytes, BCG-CWS-treated monocytes, iDCs, and DCs, respectively, were used for RT-PCR. Lane M, molecular size markers. (C) Hydrophobicity plots of TREM-2 and TRM-2V proteins showing the absence of the TM region in TREM-2V. The bars below the plot indicate their identical and unique regions. aa, amino acids. (D) Comparison of the exon-intron organization of TREM-2 and that of TREM-2V. The spliced-out region in TREM-2V is indicated on the genomic structure of TREM-2. (E) Identification of sTREM-1, a non-TM variant of TREM-1. The RT-PCR results, hydrophobicity plots, and difference in exon-intron organization between TREM-1 and sTREM-1 are presented. Black and gray arrows denote the TREM-1 and sTREM-1 PCR products, respectively.
By DDRT-PCR analysis, we initially found that TREM-2 was down-regulated in BCG-CWS-stimulated monocytes (Fig. 4A), which is consistent with the results shown in Fig. 3. TREM-2 also gave rise to a putative soluble variant, named TREM-2V (GenBank accession no. AB062787 ), during maturation to DCs (Fig. 4B and C). Thus, regulation of the level of TREM-2 depends on the differentiation pathways selected by myeloid precursor cells. The properties of TREM variants and an explanation of the origins of these transcripts are summarized in Fig. 4.
To further validate these findings, we analyzed these transcripts compared to the genomic context (Fig. 4D and E) and thereby confirmed that the transcripts originated via differential splicing. Splicing out of exon 3 in the TREM-1 gene resulted in generation of the soluble variant. The TREM-1 variant encoded a non-TM version (Fig. 4E). We termed this fragment sTREM-1 (GenBank accession no. AY074783 ) and showed that it consisted mainly of the N-terminal 136 amino acids of TREM-1, followed by a novel 14-residue C terminus owing to the absence of exon 3 and a frameshift that caused early termination and deletion of the TM region.
Splicing out of exon 4 in the TREM-2 gene led to deletion of the TM domain, producing the putative soluble variant form TREM-2V. This differential splicing conferred a frameshift at the junction of exons 3 and 5, generating a new 58-residue C-terminal stretch connecting to the 161 N-terminal amino acids of TREM-2. TREM-2V and sTREM-1 were differentially expressed in monocytes and DCs, resulting in inverse regulation of these soluble forms by BCG-CWS. LPS induced TREM-2V but not sTREM-1 in DCs and down-regulated TREM-2 and -2V in monocytes.
Expression analysis of novel membrane proteins.In Fig. 5A, we introduce a class of novel TM proteins and their expression patterns in monocytes following BCG-CWS or LPS stimulation. The structures of the SIMPLE (37) and BIGM103 (6) proteins, identified as BCG-CWS-inducible products in our laboratory, are shown diagrammatically below the residues of the PCR analysis (Fig. 5B). SIMPLE is an LPS- or BCG-inducible gene product and is expressed on endosomal membrane (37). BIGM103 and others belong to a novel class of multi-TM proteins that possess a C-terminal domain similar to those of zinc-iron transporter family (ZIP) proteins (19, 22).
Novel membrane proteins and classic inflammatory mediators are both equally targeted by PAMP induction. (A) Expression analysis of five novel membrane proteins (SIMPLE, BIGM103, KIA0062, HKE4, and LIV-1) together with two zinc transporters, hZIP1 and hZIP2. RT-PCR was performed with total RNAs prepared at different times after BCG-CWS and LPS treatment of monocytes as described in Materials and Methods. Results from two different individual samples are shown (right and left sides). (B) Schematic representation showing the structural features of the two BCG-CWS-inducible novel membrane proteins, SIMPLE and BIGM103 (BIGMo-103). SIMPLE consists of an N-terminal proline-rich (Prol. rich) region with a Nedd4-interacting motif and a C terminus with a TM and a unique region (37). The region homologous with the ZIP transporter family protein is indicated above. Lyso, lysosome; endo. and Endo., endosome. (C) Expression profiles of a set of known and commonly induced monocyte activation-related genes. RT-PCR was done with the same reverse-transcribed cDNA samples that were used for the membrane protein analysis mentioned above (panel A). LPS-stimulated monocytes were separately examined with a different sample (left side), and similar results were obtained. Only the results for the indicated time points are shown. (D) Quantitative PCR analysis of BCG-CWS regulatory genes. Total RNA was isolated and converted to cDNA for quantitative real-time PCR analysis with primers specific for SIMPLE, IL-12p40, IL-12p35, IL-23p19, and IP-10. Experiments were repeated with three samples, and a representative one is presented in copy numbers with reference to the threshold analyses (see Fig. 6). cont, control.
In Fig. 5A, we present the expression pattern of BIGM103, a putative transporter structurally resembling hZIP2, and three related family members, KIAA0062, LIV-1, and HKE4 (6). As references, we included SIMPLE and two known ZIP family proteins, hZIP1 and hZIP2. HKE4 and LIV-1 were exclusively induced in monocytes by BCG-CWS stimulation. In contrast, KIAA0062 was induced only by LPS. The ZIP transporters are expected to be constitutively expressed, but they were not detected in monocytes and DCs under the standard conditions (25-cycle PCR amplification) or stimulated by the activating agents. Therefore, it remains to be elucidated whether ZIP-like gene products could complement the ZIP-like function in APCs.
Cytokine profiles induced by BCG-CWS in monocytes.The expression profiles of a set of cytokines and chemokines were also determined (Fig. 5C). In principle, IL-23p19 (40) was induced by BCG-CWS and IP-10 (20) was induced by LPS. This issue was confirmed by a quantitative PCR (Fig. 5D) in which predominant expression of the IP-10 gene by LPS and long-term up-regulation of the IL-23p19 message by BCG-CWS were observed. Under the conditions used, reasonable induction of the SIMPLE and IL-12p40 mRNAs and minimal up-regulation of the IL-12p35 mRNA by either LPS or BCG-CWS occurred (Fig. 5D). The up-regulation of IL-23p19 and IP-10 by BCG-CWS and LPS, respectively, was further confirmed by dose-response experiments (Fig. 6). Both TNF-α and IL-6 were induced by BCG-CWS and LPS, but the TNF-α induction and its down-regulation by LPS were rapid compared to the dynamics induced by BCG-CWS (Fig. 5C). Similarly, IL-6 continued to be induced for up to 8 h by BCG-CWS, but its expression level dropped dramatically 2 h after LPS stimulation. IL-1β and IL-8 were also strongly induced by BCG-CWS, which was sustained for a longer period compared to those induced by LPS. The levels of the IL-1β, IL-6, IFN-β, and IP-10 proteins, in addition to those of TNF-α and IL-12p40 (52), were determined by ELISA (Table 2). The results principally supported those of the quantitative PCR. An IFN-inducible gene, the IP-10 gene, was not induced by BCG-CWS but was mildly induced by LPS in monocytes (Table 2), although the levels of IFN-β protein were below the detection limit in either group of stimulated monocytes (data not shown). Most of the NF-κB-inducible genes were found to respond to BCG and LPS, which may reflect their having a TLR-signaling pathway in common.
Determination of cytokine levels by real-time PCR. Dose-dependent increases in the copy numbers of mRNAs in macrophages in response to LPS or BCG-CWS were measured. (Top) PCR samples were extracted from agarose gels and used as a template to evaluate the threshold cycles in each sample. mRNA copy numbers were estimated from the graphs. (Bottom) Cells were stimulated with the indicated doses of LPS or BCG-CWS for 8 h (IL-23p19) or 4 h (IP-10), and RNA samples were analyzed as described in the legend to Fig. 5D. Data are representative of three independent experiments.
Cytokines induced by BCG-CWS in monocytesa
Comparison of differential expression of distinct TLR ligands.How nine BCG-CWS regulatory genes respond to specific ligands of TLR2, TLR3, TLR4, or TLR9 was examined (Fig. 7A), and the responses of 25 genes to LPS, BCG-CWS, and poly(I-C) are summarized in Fig. 7B. The top part of Fig. 7A includes three genes, IDO, TTP, and TREM-2; IDO exerts T-cell-suppressive effects (36), TTP is known to bind the 3′ untranslated region of TNF-α mRNA (31), and the function of TREM-2 in monocytes is unknown. The CpG ODN, the TLR9 ligand, had no effect on the expression of any of the genes, including TREM-1. The TLR2 agonists MALP-2, lipoteichoic acid (LTA), and zymosan seemed to be inducers of IDO and TTP, while the TLR4 agonist LPS down-regulated TREM-2 and -2V although TLR2 agonists had no such effect. All of the tested TLR2 agonists of bacterial components strongly induced the TREM-1 and sTREM-1 transcripts in monocytes, while they had no significant effect on the expression of DAP12. During our comparative analyses, we also noticed that CCR-7, a chemokine receptor induced via DC maturation (11), was elevated in monocytes by LPS and BCG-CWS stimulation but not by poly(I-C), the TLR3 agonist. MxA GTPase (24) and IP-10 were specifically induced in both monocytes and DCs by LPS and poly(I-C) but not by BCG-CWS. Marked differential responsiveness of IL-12p35 and IL23p19 was also found for BCG-CWS. These results suggest that each TLR family member has its own unique gene regulatory profile besides the common cellular routes of gene expression.
Comparison of differential gene expression in response to TLR-specific PAMPs. (A) RT-PCR analysis of 10 genes, including that for actin, performed with nonstimulated (NS) and PAMP-stimulated monocyte RNAs (stimulated for 2, 6, or 8 h). For TREM-1 and DAP12, only the expression pattern at 6 h of stimulation is shown. CCR7, IP-10, IL-23p19, and MxA expression was analyzed in a different preparation of monocytes following 8 h of stimulation. The concentrations of ligands for TLR4 (LPS), TLR2 (PGN, MALP-2 [MALP], lipoteichoic acid [LTA], zymosan [ZYM]), TLR3 [poly(I-C) (PIC)], TLR2, TLR4, PGN derived from BCG (BPGN) (52), and TLR9 (GpC and CpG ODN) are described in Materials and Methods. (B) Summary of the genes for which we detected differential expression in response to at least one of the PAMPs [BCG-CWS, LPS, and poly(I-C)] tested. The plus and minus signs indicate whether a particular gene responded to the respective TLR ligand or not. Transcriptional up- or down-regulation is shown by up or dn next to a plus sign, and nd indicates that the status was not determined. MCP1, macrophage chemotactic protein 1.
DISCUSSION
BCG-CWS differentiates monocytes into macrophages and immature DCs into an activated form (50). BCG-CWS has been used as a therapeutic adjuvant without severe side effects. The BCG-CWS-regulated genes fluctuated similarly in BCG-CWS-treated mature DCs (data not shown). Through this work, we aimed to address the adjuvant-active potential of PAMPs at the gene expression level and select the specific markers for BCG-CWS-activated macrophages and DCs. These markers will be useful tools in predicting the prognoses of cancer patients with adjuvant immunotherapy.
Several characteristics of BCG-CWS-induced genes compared to genes induced by LPS (an adjuvant with a high toxicity potential) were identified in this study. (i) The splice variants were generated in TREM receptors concomitant with long-term up-regulation of DAP12-binding receptors. (ii) ZIP-like family proteins were specifically induced. (iii) Of the cytokines tested, BCG-CWS induced IL-23 but not IRF-3- or IFN-β-induced effector genes. BCG-CWS harbors its own unique gene regulatory profile, partly acting as a TLR2 or TLR4 agonist, in addition to unknown routes of gene expression. We have drawn these conclusions after comparing genes induced in monocytes by BCG-CWS with those induced by LPS as described below.
We first mentioned the identification of non-TM forms of TREM-1 and TREM-2 receptors; these receptors have been characterized recently (9, 11). TREM-1 induces massive TNF-α production upon antibody ligation in the presence of LPS, and recombinant TREM-1, consisting of the extracellular domain of TREM-1, effectively blocks septic shock. Therefore, the sTREM-1 identified here may act as a natural TREM-1 competitor for protection against enhanced inflammatory responses. Following BCG-CWS induction, there was a gradual increase in sTREM-1 expression with time, which may support a feedback or counteractive mechanism. It will be of interest to know how a TLR2 agonist can elicit such novel splicing, leading to an increase in the accumulation of the soluble protein form. On the other hand, TREM-2 was found to be preferentially expressed in DCs (11) and subjected to down-regulation by BCG-CWS or LPS, via either TLR2 or TLR4. Antibody ligation to TREM-2 produces a unique mature DC population with a surface expression profile (11) different from that of LPS-induced DCs. Therefore, a possible interpretation is that by down-regulating the expression of TREM-2 and -2V, individual PAMPs interfere with the natural DC maturation pathway, giving rise to pathogen-specific mature DCs. This issue will be functionally analyzed if the ligands for the receptors TREM, MDL, and CRACC are identified. sTREM-1 induction may be related to part of the less toxic properties of PAMPs.
We have described the properties of BIGM103, a novel member of the ZIP-like family of proteins (6). This protein is predicted to possess zinc transporter and metalloprotease activities. The endosomal expression of this protein (6), and presumably other proteins in this family, may be closely associated with phagocytosis-mediated elimination of microbial components in macrophages and DCs. Although there is little evidence that the proteins in this family are involved in the innate immune response, including differentiation or activation of APCs, the PAMP-specific induction profiles of the ZIP-like family of proteins may reflect the contribution of these proteins to the antigen-presenting role of APCs.
IL-12p70 (the p40-p35 complex) is expressed in monocyte-derived DCs (APCs) and acts as a crucial factor on lymphocytes for IFN-γ production (49), and our differential gene expression analysis showed that IL-12p40, but not IL12p35, transcripts were markedly induced by BCG-CWS. Virtually no production of the IL-12p70 protein was confirmed by ELISA (38). This is consistent with our recent finding (33) and a previous report (43) that BCG-CWS induced the release of large amounts of IL12p40 in vitro and IL-23p19 is coexpressed with IL-12p40 by BCG-CWS and other TLR2 agonists. In guinea pig studies, we confirmed that a large amount of IL-23, the p19-p40 complex, is released in a BCG-CWS-rich environment (I. Shiratori and T. Seya, unpublished data). IL-23 may participate in a distinct mode of protective immunity regulating autoimmune responses (15). IL-12 can induce chemokine receptors CCR5 and CXCR3 in lymphocytes and encourage the effector lymphocytes to kill target cells expressing MIP-1 and IP-10 (21). Although there are many DC subsets that differentially respond to PAMPs to reveal distinct cytokine profiles (34), maturation of myeloid or monocyte-derived DCs would be crucial for APC function. DC maturation, in concert with a functioning cytokine network, will be important in establishing efficient antitumor and infectious immunity. The balance of IL-12 versus IL-23 (or IL-27) induced by different PAMPs may critically affect the adjuvant function for vaccines.
BCG-CWS serves as a ligand for TLR2 and TLR4, while LPS serves as a ligand for TLR4. TLR2 and TLR4 partly share their signaling pathways since their adapter molecules are MyD88 and TIRAP (also designated Mal), leading to activation of NF-κB (26, 55). TLR4 additionally recruits TICAM-2 and TICAM-1, which activate IRF-3 potentially and NF-κB weakly (42). Thus, we could not see an all-or-none difference in the NF-κB-dependent genes expressed by the two TLR ligands. That is, cytokines such as TNF-α, IL-6, IL-8, and IL-1β or chemokines like MIP and MCP are NF-κB dependent and are commonly induced by BCG-CWS and LPS. Thus, there are minimal differences (in terms of gene expression) in NF-κB-dependent gene expression between BCG-CWS and LPS. However, the point is that there are several genes specifically expressed by BCG-CWS. IL-23p19, LIV-1, and HKE-4 are examples of such genes and are suggestive of the presence of additional signaling pathways for the BCG-CWS response. LPS, but not the TLR2 agonists, activates IRF-3 and induces expression of IFN-inducible genes in a MyD88-independent fashion, and BCG-CWS-mediated TLR4 activation barely induces activation of IRF-3 and expression of IFN-inducible genes. LPS-TLR4 signaling may be toxic because of simultaneous activation of MyD88 and TICAM-1. In light of our results, LPS-dependent TLR4 signaling somewhat resembles TLR3 signaling (34) and is distinct from BCG-CWS-mediated TLR4 signaling. These differences mean that the BCG-CWS-mediated gene regulatory profile common to TLR2 and TLR4 is important for therapeutic potential in adjuvant therapy. Indeed, MyD88 is an adapter common to TLR2 and TLR4 and is essential for effective cytotoxic T lymphocyte induction with MyD88-deficient mice (1).
Intracutaneous application is the usual mode of clinical BCG-CWS application, and as a result, mast cells, Langerhans cells, and iDCs underneath the skin are probably the immune cells that first encounter the administered BCG-CWS. In vitro BCG-CWS-monocyte interaction studies do not always mimic the way BCG-CWS acts in a local environment, where an APC encounters a pathogen component directly and humoral and cell-mediated immunity takes place through a complex interplay between cells of the adaptive and innate immune systems. This complex interplay involves not only monocytes but also cytotoxic T lymphocyte and NK activation in vivo.
A previous in vitro study demonstrated that BCG-CWS was efficiently taken up by iDCs and promoted DC maturation (50). BCG-CWS also induces the expression of various costimulatory molecules and elicits an extensive mixed-lymphocyte reaction (50), yet we lack sufficient knowledge about the BCG-CWS-induced appearance of neoreceptors on DCs (51, 52). Therefore, a rational next step would be to focus on analyzing such cell surface molecules, which may add new information on the signaling and cell-cell interaction abilities of BCG-CWS-primed APCs. It would be more informative to analyze gene expression in BCG-CWS-induced DCs for estimation of their potential for immune responses. In fact, in BCG-CWS therapy, a good prognosis is often correlated with elevated serum IFN-γ levels in postoperative patients, although the source of the elevated IFN-γ levels remains unclear (4, 23).
It is becoming clear that the mode of the DC response is critical in determining whether NK or T cells are preferentially activated via cytokines and adhesion molecules (14). We hold that the differential DC activation should be sustained by its gene regulation. The possible contributions of this study to the resolution of unsettled issues are summarized as follows. First, the gene expression pattern could help us to understand how BCG-CWS differentiates immature myeloid cells into APCs. Second, as BCG-CWS interacts with TLR2 and TLR4 but its immunomodulatory profile and toxic threshold are different from those of LPS, we can see differential signaling and markers for the therapeutic potential of PAMPs at the gene expression level. Third, we may be able to identify as-yet-undiscovered molecules or pathways downstream of TLRs that are potentially linked to the innate responses.
ACKNOWLEDGMENTS
This work was supported in part by CREST, JST (Japan Science and Technology Corporation), by Grants-in-Aid from the Ministry of Education, Science, and Culture (Specified Project for Advanced Research) and the Ministry of Health and Welfare (Organization for Pharmaceutical Safety and Research) of Japan, and by the Takamatsunomiya Princess Memorial Foundation. N.A.B. and M.T. are research fellows in the Cancer Research Institute Japan.
FOOTNOTES
- Received 31 March 2003.
- Returned for modification 26 May 2003.
- Accepted 15 October 2003.
- Copyright © 2004 American Society for Microbiology
