Novel identi ﬁ ed aluminum hydroxide-induced pathways prove monocyte activation and pro-in ﬂ ammatory preparedness ☆

Aluminum-based adjuvants are the most widely used adjuvants in human vaccines. A comprehensive under- standing of the mechanism of action of aluminum adjuvants at the molecular level, however, is still elusive. Here, we unravel the e ﬀ ects of aluminum hydroxide Al(OH) 3 by a systems-wide analysis of the Al(OH) 3 -induced monocyte response. Cell response analysis by cytokine release was combined with (targeted) transcriptome and full proteome analysis. Results from this comprehensive study revealed two novel pathways to become activated upon monocyte stimulation with Al(OH) 3 : the ﬁ rst pathway was IFN β signaling possibly induced by DAMP sensing pathways like TLR or NOD1 activation, and second the HLA class I antigen processing and presentation pathway. Furthermore, known mechanisms of the adjuvant activity of Al(OH) 3 were elucidated in more detail such as in ﬂ ammasome and complement activation, homeostasis and HLA-class II upregulation, possibly related to increased IFN γ gene expression. Altogether, our study revealed which immunological pathways are activated upon stimulation of monocytes with Al(OH) 3 , re ﬁ ning our knowledge on the adjuvant e ﬀ ect of Al(OH) 3 in primary monocytes. Signi ﬁ cance: Aluminum salts are the most used adjuvants in human vaccines but a comprehensive understanding of the working mechanism of alum adjuvants at the molecular level is still elusive. Our Systems Vaccinology approach, combining complementary molecular biological, immunological and mass spectrometry-based tech- niques gave a detailed insight in the molecular mechanisms and pathways induced by Al(OH) 3 in primary monocytes. Several novel immunological relevant cellular pathways were identi ﬁ ed: type I interferon secretion potentially induced by TLR and/or NOD like signaling, the activation of the in ﬂ ammasome and the HLA Class-I and Class-II antigen presenting pathways induced by IFN γ . This study highlights the mechanisms of the most commonly used adjuvant in human vaccines by combing proteomics, transcriptomics and cytokine analysis revealing new potential mechanisms of action for Al(OH) 3 .


Introduction
Since 1926, colloidal aluminum salts are known for their adjuvant features, when Glenny et al. discovered that diphtheria toxoid adsorbed to aluminum salts showed a significantly higher antibody titer against the toxoid compared to antigen alone [1]. Since then, aluminum salts have been widely used as vaccine adjuvants, even though the mechanism of action has remained largely unknown. Shortly after Glenny et al. discovered its potential, it was observed that the clearance of aluminum salt-adsorbed toxoids was delayed compared to non-adsorbed toxoids [2]. This observation was the basis for the theory that antigen-aluminum salt aggregates form a depot at the site of injection, leading to a slow release of the bound antigen. However, when the aluminum salt deposit was removed at various time points, the immune response was not compromised, suggesting that this is not the only mechanism by which aluminum salts increase the immune response to antigens [3]. It was shown that antigens desorb rapidly from AlPO 4 as well as Al(OH) 3 after administration, explaining that even when the aluminum salt-deposit is removed, an antibody response is still created [4]. It was demonstrated by flow cytometric analysis that Al(OH) 3 causes the differentiation of monocytes to antigen-presenting dendritic cells, which then can elicit a specific immune response in the presence of an antigen [5][6][7]. Al(OH) 3 has been shown to steer the CD4 + T cell response towards a T helper type 2 (Th2) response and induces the recruitment of immune cells (e.g. dendritic cells, inflammatory monocytes, eosinophils, neutrophils natural killer cells and CD11 + cells) to the site of injection [7][8][9][10][11][12].
Additional mechanisms for the adjuvant effect of Al(OH) 3 have been described: for example, the activation of the NLRP3 inflammasome, as identified by both in vivo and in vitro studies [12][13][14][15]. Inflammasome activation results in the secretion of IL-1β, a pro-inflammatory cytokine and potent inducer of the adaptive immune response and Th2 polarization [13,15,16]. The inflammasome, in vitro, appeared to be critical for Al(OH) 3 dependent IL-1β secretion and antibody responses. In vivo however, there is conflicting data about the necessity of the inflammasome to induce a humoral response [13,15,17]. Complement activation is another mechanism involved in the adjuvant activity of Al (OH) 3 [12,18]. The release of Danger Associated Molecular Patterns (DAMPs) like uric acid and host DNA, is induced by Al(OH) 3 [13,19,20]. Uric acid is involved in T cell priming and humoral responses [13,15]. Besides this, uric acid can also activate the inflammasome in vitro [15,21]. However, in vivo IL-1β production did not depend on the presence of uric acid [14,15]. Uric acid is involved in the influx of inflammatory monocytes to the site of injection as is MyD88, since MyD88-deficient mice showed a significant reduction of the influx of inflammatory monocytes [13]. Even though MyD88 appears to be dispensable for antibody production there is a role in the adjuvant activity of Al(OH) 3 [13]. DNA is also involved in the adjuvant activity of Al (OH) 3 [19] . DNA can result in the secretion of IL-1β or in the secretion of IFNβ and enhances MHC class II antigen presentation [19,20]. In conclusion, there are many potential mechanism described for the adjuvant activity of Al(OH) 3 . A comprehensive study compiling transcriptome data with proteome data would make it possible to follow pathways in more detail and create an overview of which pathways are actually activated by the Al(OH) 3 adjuvant.
In this study, we investigated the Al(OH) 3 -induced innate cell response at the transcriptome and proteome level, aiming to further complete the overview of molecular pathways and cellular processes involved in the adjuvant effect of Al(OH) 3 . Primary monocytes were chosen as a model, since these prominent mononuclear phagocytes play an important role in tissues when activated, bridging the innate and adaptive immune responses [22,23]. Besides there known differentiation into monocyte derived dendritic cells (MDDCs), monocytes can also enforce their antigen presenting role to T cells in response to various stimuli as recently reviewed [24]. In the current study the effect of Al(OH) 3 on monocytes will be investigated. The combination of transcriptome/proteome analyses and cytokine measurements enabled the comprehensive identification of molecular pathways in Al(OH) 3 stimulated monocytes, further clarifying the mechanism by which Al (OH) 3 works.

Ethics statement
This study was conducted using blood donations, provided by the National Institute for Public Health and the Environment (Bilthoven, The Netherlands), for primary cell isolation. The blood donations for this research goal were specifically approved by the accredited Medical Research Ethics Committee (MREC), METC, Noord-Holland in The Netherlands. The study was conducted according to the principles expressed in the Declaration of Helsinki and written informed consent was obtained from all blood donors before collection and use of their samples. Blood samples were processed anonymously. All human primary cells described in this study were obtained by the rules of this ethical statement.

Reagents used in cell stimulation
Aluminum hydroxide (Alhydrogel 2%; Brenntag; Frederikssund; Denmark) is referred to as Al(OH) 3 . Lipopolysaccharide (LPS) from E. coli K12 (Invivogen; San Diego; California; USA) was used as positive control and is referred to as LPS.

Isolation and stimulation of monocytes
Peripheral blood derived from 7 healthy adult donors and obtained as described in the ethical statement, was used for monocyte isolation. First, peripheral blood mononuclear cells (PBMCs) were obtained by gradient centrifugation at 1,000 ×g for 30 min on Lymphoprep (Nycomed; Zurich; Switzerland). Second, monocytes were isolated from the PBMC fraction using MACS in combination with anti-CD14 MACS beads (Miltenyi Biotech; Bergisch Gladbach; Germany). Purity check by flow cytometric analysis of CD14 cell surface expression was performed and only if the purity of the monocyte population was ≥95% the cells were used for proteome and transcriptome analysis.
After incubation, culture supernatants were collected and stored at −80°C for cytokine analysis. Small aliquots were taken from cell suspensions from each donor per time point for flow cytometry analysis.
For proteome analysis, the cells of at least one well per condition per time point per donor (three individuals) were washed with PBS, before adding 500 μl lysis buffer (4 M Guanidine·HCl in phosphate buffer pH 7.5) (Sigma Aldrich; St Louis; Missouri; United States).
For transcriptome analysis, the cells of at least one well per condition per time point per donor (5 individuals) were placed in 350 μl of RLT buffer (Qiagen; Venlo; The Netherlands). The samples for proteome and transcriptome analysis were stored at −80°C.

Generation and stimulation of primary Monocyte-Derived Dendritic Cells (MDDCs)
Monocytes from the peripheral blood of 6 donors, obtained as described in the ethics statement, were cultured in a 24-well culture plate (0.4 × 10 6 cells/ml, 1 ml/well) in DC culture medium, i.e. IMDM (Gibco) containing 1% FCS, 100 units/ml penicillin, 100 units/ml streptomycin, 2.92 mg/ml L-glutamine (Gibco), 500 units/ml GM-CSF (Prepotech; Rocky Hill; New Jersey; USA) and 800 units/ml IL-4 (Sanquin; Amsterdam; The Netherlands) for 6 days. On day 6, the immature MDDCs (iMDDCs) were either left unstimulated, or were stimulated for 48 h with 250 μl of LPS or Al(OH) 3 in DC culture medium containing 250 units/ml GM-CSF, with a final concentration of the stimulus of 0.1 μg/ml or 10 μg/ml. After 48 h, supernatant were stored at −80°for cytokine analysis. The stimulated MDDCs were harvested and suspended in FACS buffer. Aliquots were taken from cell suspensions for flow cytometric analysis which was used as a quality control for the cell culture procedure. Flow cytometric analysis of monocytes was used as a quality control for the cell culture procedure. LPS served as a positive control and performed as expected, by inducing CD80 expression (Supplementary S1). Samples were only used for further analysis if the monocytes were at least 95% pure.

Culture and stimulation of THP-1 cells
The human monocytic cell line THP-1 (ATCC; Teddington; Middlesex; U.K.) was used for verification experiments of leads identified in primary monocytes. THP-1 cells were cultured according to the supplier's protocol with the addition of 100 units/ml penicillin, 100 μg/ ml streptomycin and 300 μg/ml L-glutamine to the medium, hereafter referred to as THP-1 culture medium, in culture flasks.
Cells were primed with 300 ng/ml phorbol 12-myristate 13-acetate (PMA) (Sigma Aldrich) for 24 h, after which the medium was discarded and new medium without PMA was added. After 24 h in the refreshed medium, the following stimulations were performed: mock, 50 or 100 μg/ml Al(OH) 3 (based on literature [5,14] and preliminary experiments). Each of these conditions was performed in the presence and absence of 25 μg/ml glybenclamide (Invivogen). An additional cell batch was left completely unstimulated and unprimed. Twenty-four and 48 h after stimulation, supernatants of cultures were harvested and used for an IL-1β ELISA.

Flow cytometric analysis
For purity analysis, isolated monocytes were stained with PE-conjugated anti CD14 clone (H5E2). Cultured monocyte and MDDC cell suspensions were stained with PE-conjugated anti-CD40 (BD Biosciences; clone 5C3), APC-conjugated anti-CD80 (BioLegend; clone 2D10), FITC-conjugated anti-CD83 (BD Biosciences; clone HB15e), Pacific Blue-conjugated anti-CD86 (BioLegend; clone IT2.2), HLA-DR FITC (Sanquin; clone L243) and finally with LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Invitrogen) for 30 min at 4°C. The cells were washed in FACS buffer and fixed in FACS buffer containing 1% paraformaldehyde (PFA). Data were acquired on a FACS Canto II (BD Biosciences) and analyzed using FlowJo software (Tree Star) by comparing the MFI of the stained control with the MFI of the stained stimulus. The data were corrected for the percentage autofluorescence determined in verification experiments by comparing the unstained stimulated with the stained stimulated cells for each stimulus.

mRNA expression analysis
mRNA was isolated from monocyte samples using the RNeasy mini kit (Qiagen) according to the animal cell spin protocol provided. Isolated RNA concentrations and purities were determined by the spectrophotometric analysis of the 260-nm and 280-nm absorbance on the Nanodrop 2000 (Thermo Fischer). Subsequently, cDNA synthesis was performed with the RT cDNA synthesis kit (Qiagen) and the RT preAMP pathway primer mix, innate and adaptive immunity comprising primers for 89 functional RNA species and controls, on 12 ng of RNA per condition (Qiagen). cDNA was subsequently frozen at −20°C.
qPCR was performed by using the Innate and Adaptive Immune Responses RT 2 Profiler PCR Arrays (PAHS 052ZC) (Qiagen) (Supplementary S2) on a qPCR apparatus (ABI step one plus; Applied Biosystems; Foster City; California; USA) with a melt curve determination as a quality control, included in the measurement.
qPCR data were analyzed with the qPCR statistical web software from Qiagen. Gene expression was normalized to HPRT1 as housekeeping gene, after which the fold change was calculated according to the manufacturer's protocol. Fold change-values greater than one indicate an upregulation, in which the fold regulation is equal to the fold change. Fold change-values less than one indicate a downregulation in which the fold regulation is the negative inverse of the fold change. For donor F (with three technical replicates), p-values were calculated based on a Student's t-test of the replicate values for each gene in the control group and treatment group. Genes were considered regulated if p-value < 0.05 and a two-fold change in expression occurred. We validated technical replicate reproducibility with the first donor. The Coefficient of Variation (CV) between technical replicates was 3.5% for all transcriptome data and all the genes that met the two fold difference criterion also had a p-value < 0.05. Therefore, for the other four donors we used one technical replicate and considered genes regulated if they met the two-fold difference criterion. Overall, genes were considered regulated by Al(OH) 3 if they showed a two-fold up-or down-regulation in at least three out of five donors, or were regulated in two donors with a clear trend in at least one other.
LPS served as a control for the time point and for potential artifacts [25] of the cell isolation with CD14 beads, by determining if LPS-related genes were indeed increased. It was confirmed that LPS-related genes were indeed increased after 24 h (Supplementary S3), indicating that the time point can indeed be used and that artifacts (no induction of LPS-related genes) might not play a role here.

Protein isolation
Monocytes were lysed by adding 500 μl 4 M Guanidine·HCl in 100 mM phosphate buffer (pH = 7.5) to the culture plate and incubated for 2 h at 4°C. During these 2 h, the cells were subjected to a freezethaw step. The protein concentrations were determined with a BCA protein assay (Pierce biotechnology; Waltham Massachusetts; USA) according to the manufacturer's protocol. Lysed cells were stored at −80°C in a culture plate.

Protein digestion
Protein samples from monocytes were diluted 4 times with 100 mM phosphate buffer pH 7.5 to reduce the Guanidine·HCl content to 1 M and adjust the pH to 7.5. Proteins were digested at 37°C with Lys-C (Roche; Basel; Switzerland) in an enzyme-to-substrate ratio of 1:10 (w/ w). After 4 h, fresh Lys-C was added in a 1:10 (w/w) ratio for an overnight incubation.

Protein labeling in monocyte samples
Following the protein digestion, aliquots from the six monocytes samples per donor (three donors) were taken containing equal protein amounts and per condition labeled using tandem mass tag labeling 6plex (TMT(6), Thermo Fisher). For donor A 25 μg per condition per time point was labeled, for donor B 50 μg protein per condition per time point was labeled and for donor C 25 μg protein and 10 μg per condition was labeled at 24 and 48 h, respectively. Labeling was performed on Solid Phase Extraction (SPE) columns (Waters; Milford; MA; USA). The SPE columns were equilibrated as described by the manufacturer and washed with 100 mM phosphate buffer pH 7.5. The digested protein samples were loaded onto 6 separate SPE columns using a vacuum manifold (Waters) and washed with 100 mM phosphate buffer pH 7.5. The TMT-label (0.8 mg per TMT-label) was reconstituted in 41 μl acetonitrile (AcN) according to the supplier's protocol. The AcN concentration was reduced to a maximum of 2.5% (v/v) with 100 mM phosphate buffer pH 7.5. The individual TMT labels were loaded onto the 6 SPE columns, leaving 0.5 ml reagent on top of the column for a 30 min incubation, fresh label was added for another 30 min of incubation, after which the columns were washed with water containing 0.5% formic acid (FA). The six samples were eluted with 90% AcN containing 0.5% FA, pooled, dried by centrifugal evaporation and reconstituted in water containing 0.1% trifluoroacetic acid (TFA).

LC-MS/MS analysis
Peptide separation was performed on a Proxeon Easy-nLC 1000 system (Thermo Scientific). Peptides were trapped on a double-fritted trapping column Reprosil C18 (Dr. Maisch; Ammerbuch; Germany; df = 3 μm, 2 cm length × 100 μm I.D.) and separated at a column temperature of 40°C on an analytical column Poroshell 120 EC-C18 (Agilent; df = 2.7 μm, 50 cm length × 50 μm I.D.) both packed in house. Both columns contained a 1-mm length KASIL frit prepared as described by Meiring et al. [27], to retain the packing bed. The second KASIL frit, upstream in the trapping column, was made the same way after flushing the packing bed with helium. Solvent A was 0.1% FA in MilliQ water and solvent B was 0.1% FA in AcN (Biosolve). The peptides were separated in 143 min (10 min at 2% B for peptide loading onto the trapping column, followed by a 118-minute gradient from 2% to 30% B and 5 min at 70% B) in a non-linear optimized gradient [28]. After 128 min, the system was kept at 5% B for 15 min to equilibrate the column. The column effluent was electro-sprayed directly into the MS using a gold-coated fused silica tip with a 5-μm TipID and a spray voltage of 1.8 kV.
Mass spectrometric data were acquired on a Tribrid-Orbitrap Fusion (Thermo Fisher Scientific), where the full scan (MS 1 ) spectra were acquired with a scan range of m/z 350-1500 Da at 120 K resolution (FWHM) with an Orbitrap readout. The Automatic Gain Control (AGC) for the MS 1 was set to 200,000 and the maximum injection time to 50 ms in top speed mode with a duration of 3 s, where precursor ions with an intensity of > 5000 were selected for fragmentation (MS 2 ). Charge states between 1 and 7 were selected. MS 2 was performed using Collision-Induced Dissociation (CID) in the linear ion trap with a normalized collision energy of 35%. The AGC for the MS 2 was set to 10,000 and the maximum injection time to 35 ms. Synchronous-Precursor-Selection was enabled to include up to 10 MS 2 fragment ions. These ions were further fragmented by Higher energy Collision Dissociation (HCD) with a normalized collision energy of 50%. TMT reporter ions were analyzed in the Orbitrap analyzer, with the AGC set to 100,000 and a maximum injection time of 120 ms.
Proteomics data were analyzed with Proteome Discoverer 2.1 (PD 2.1) (Thermo Fisher Scientific) using default settings unless stated otherwise. Precursor mass tolerance was set to 5 ppm, MS/MS scans were searched against the human Uniprot database (Nov 2014), containing 23,048 entries, using the Sequest HT search engine with full enzyme specificity for Lys-C, with b and y type ions enabled for CID and HCD data with a fragment mass tolerance of 0.5 Da. The data was searched with Aspargine deamidation and Methionine oxidation as dynamic modifications. TMT(6) was set as a static modification on the Lysine residues and the peptide N-termini. For relative quantitation, the quantification node was used with TMT(6) as defined quantification method and an integration tolerance of 0.2 Da. A decoy database defined in the Percolator node was used to validate and filter the peptideto-spectrum matches with a False Discovery Rate (FDR) of < 5%. Only medium (FDR < 5%) and High (FDR < 1%) confident identified proteins were used in the further data analysis.
The protein data from the SCX fractions of an individual donor were combined by PD 2.1 in a consensus report, resulting in one output table per donor/technical replicate. If multiple entries occurred for the same protein, based on Uniprot and NCBI data, the ratios given by Proteome Discoverer were Log2-transformed and averaged for further analysis. Next, a median correction normalization was performed. For donor C, three technical replicates of the medium control and Al(OH) 3 -stimulated culture conditions were measured (their median CV was 10.5% for the full proteomics data set) and these were combined into a single average value for this donor to make sure each donor is weighed equally in further analysis. For donors A and B, one sample per condition was analyzed. Data of three biological replicates (donors) were compared and proteins that were upregulated or downregulated by 1.5 fold or more in at least two out of three biological replicates were considered substantially regulated. These regulated proteins were imported in STRING (string.embl.de) [29], to identify enriched pathways in these regulated protein sets (FDR < 0.05), within Gene Ontology (GO) biological processes.
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [30] partner repository with the dataset identifier PXD008452. ELISA data were analyzed with GraphPad Prism®. Significance of difference was determined using a 2-way ANOVA and a Tukey's multiple comparison test with an Alpha of 0.05.

Al(OH) 3 induces changes in cell surface markers on monocytes but not on Immature monocyte derived dendritic cells
When immature monocyte-derived dendritic cells (iMDDCs) were stimulated with Al(OH) 3 , no significant changes in cell surface marker expression were observed, as determined by flow cytometry, whereas LPS stimulation resulted in the upregulation of CD40, CD80 and CD83 (Supplementary S4). Based on the limited effect of Al(OH) 3 on iMDDCs, further detailed analyses were only performed using monocytes.
The effect of incubation of monocytes with Al(OH) 3 on their cell surface marker expression was investigated at the level of gene expression (with targeted transcriptomics) and protein identification and quantification. Al(OH) 3 stimulation upregulated gene expression of, CD80, and CD8A in monocytes compared to control cells after 24 h of stimulation. Gene expression of CD14, a typical monocyte marker that is lost upon differentiation into iMDDCs and linked to LPS stimulations [31], was downregulated in Al(OH) 3 -stimulated monocytes after 24 h (Fig. 1), while gene expression was upregulated in LPS-stimulated monocytes (Supplementary S3, S5). Moreover, at the protein level (analyzed with quantitative mass spectrometry) CD14 was downregulated after 48 h of Al(OH) 3 stimulation (Supplementary S6, S7). Al (OH) 3 stimulation of monocytes resulted in an increased protein expression of activation markers CD9 and CD71 (TRFC) after 48 h (Table 2 and Supplementary S6, S7) (please note: these markers were not analyzed using targeted transcriptomics). These transcriptome and proteome results confirmed that in an Al(OH) 3 -stimulated monocyte population cells became activated and start changing their cell surface marker phenotype including the loss of CD14 expression.

Al(OH) 3 induces changes in immunogenic genes upon monocyte stimulation
Targeted transcriptome analysis of monocytes was performed on an array of 96 genes, comprising innate and adaptive immune system genes, housekeeping genes and control genes (Supplementary S2). In total, 89 functional gene-transcripts could be measured, of which 40 genes were found to be upregulated and 5 genes were downregulated (Table 1 and Fig. 1). Some donor variation was observed since genes were not always differentially regulated in all donors. Fig. 1 is most representative for all donors used.

Quantitative proteomics reveals up and downregulation of homeostatic and immunogenic processes after Al(OH) 3 stimulation of monocytes
Quantitative proteome analysis of Al(OH) 3 -stimulated monocytes resulted in the identification of over 4000 unique proteins, of which 3000 proteins could be relatively quantified. About 200 proteins were upregulated and about 190 proteins were downregulated in Al(OH) 3stimulated monocytes compared to unstimulated monocytes (Supplementary S6-S10). Some of these proteins could be matched to the transcriptome data, e.g. HLA-A, HLA-E and CD14 (Table 2). Cytokines were not identified in the lysates. However, several cytokine inducible proteins were detected in the lysates (e.g. Mx1, Mx2 and IFIT3). These proteins are induced by IFNβ [32][33][34].
Over-representation analysis of GO terms, revealed the involvement of various biological processes (Figs. 2 and 3). A comprehensive overview of all differentially regulated proteins and enriched GO terms is represented in Supplementary S7 and S8-10. Immune processes, like antigen processing and presentation and innate immune response were overrepresented as well as processes involved in homeostasis like metabolic processes, localization and transport (Fig. 2). Specific pathways that were induced after 24 h were related to localization and copper homeostasis: specified proteins related to this process were e.g. copper-transporting ATPase 2 (ATP7B) and Amyloid-beta A4 (APP). After 48 h of stimulation, additional GO terms were enriched in the upregulated protein sets as compared to 24 h of stimulation including: intracellular transport processes (e.g. protein transport, protein localization) and response to stress (Fig. 3B). Processes related to cell death were also increased after Al (OH) 3 stimulation (programmed cell death). A complete overview can be found in Supplementary S8-S9.
Within the set of downregulated proteins after 24 h of Al(OH) 3 stimulation, no GO terms were over-represented. However, in the data set representing downregulated proteins after 48 h, over-representation of various GO terms was identified i.e. blood coagulation, inflammatory response, regulation of immune system process and response to stress (Supplementary S10). These data show that Al(OH) 3 regulated both processes involved in homeostasis as well as in the immune response.

IL-1β production as a consequence of Al(OH) 3 stimulation is partially inflammasome dependent
Inflammasome-related gene expression was increased in transcriptome analysis after 24 h of Al(OH) 3 stimulation, i.e. NOD1 and IL-1R1, while CASP1 and MyD88 were upregulated in two donors (Fig. 1). IL-1β ELISA analysis in the culture supernatant of two donors showed a trend towards increased IL-1β secretion (Supplementary S11). Cytokine analysis with an IL-β ELISA in THP-1 cells confirmed the upregulation of IL-1β after 48 h (Fig. 4 and Supplementary S12).
To investigate the involvement of the inflammasome in IL-1β production upon Al(OH) 3 stimulation in more detail, THP-1 cells (a human monocytic cell line) were differentiated into macrophages by addition of PMA and stimulated with 50 μg/ml of Al(OH) 3 in the presence and absence of the inflammasome blocker glybenclamide. Measurement of IL-1β in cell culture supernatants of THP-1 cells revealed that IL-1β secretion upon Al(OH) 3 stimulation decreased in the presence of glybenclamide ( Fig. 4 and Supplementary S12). A similar trend was observed when tested in primary monocytes: one donor with blocking (donor A) and one donor without blocking experiment (donor B) (Supplementary S11). However, this response was less pronounced if compared to the response in PMA-primed THP-1 cells. The loss of IL-1β production upon glybenclamide addition provided evidence that the IL-1β production of Al(OH) 3 -stimulated monocytic cells is indeed partially dependent on the activation of the inflammasome.

Al(OH) 3 activates the complement system
Our proteomics data of Al(OH) 3 -stimulated primary monocytes identified the altered expression of proteins from all complement pathways: C4 (lectin pathway), C8 (alternative pathway), and C5aR (classical pathway) were significantly upregulated after 24 h of stimulation (Table 2 and Supplementary S7). In addition, gene expression of C3 was upregulated and CASP1 was upregulated in two donors (Table 1).

Stimulation with Al(OH) 3 increases the chemotactic capacity of the monocytes
Transcriptome data showed that Al(OH) 3 -stimulated monocytes regulated mechanisms to attract other immune cells. The gene expression of CCL2 was upregulated in Al(OH) 3 -stimulated monocytes ( Table 1); this gene encodes for a molecule involved in monocyte trafficking. Al(OH) 3 stimulation also induced the expression of chemokine receptors CCR6 and CCR8 (Table 1).

Al(OH) 3 induces upregulation of components of the HLA processing and presentation pathways in monocytes
The upregulation of HLA class II pathways in monocytes induced by Al(OH) 3 was observed at the level of the proteome. The pathway antigen processing and presentation by HLA class II was enriched in the upregulated protein data set of Al(OH) 3 -stimulated monocytes. HLA class II-related upregulated proteins included Legumain [37][38][39] and Cathepsin D, S and L ( Table 2 and Supplementary S7) [40]. mRNA expression of IFNγ was upregulated as was the protein expression of IL-3R, both synergizing in increasing HLA class II expression [41,42].
In addition to HLA class II pathway upregulation, induction of HLA class I-related genes and proteins was observed. Al(OH) 3 stimulation of monocytes increased gene and protein expression of HLA-A and HLA-E, as well as the gene expression of β2M compared to controls (Table 1 and Fig. 1). HLA class I-related proteins were also upregulated upon Al (OH) 3 stimulation i.e. heat shock protein 90 (HSP90) and Minor Histocompatibility Complex HM13 (Table 2).

Al(OH) 3 stimulation of monocytes induces IFNβ secretion potentially by upregulating TLR or NOD like signaling
Al(OH) 3 stimulation of monocytes resulted in a significantly increased IFNβ gene expression (p < 0.01). Protein expression of proteins downstream of IFNβ: i.e. Mx1, Mx2, and IFIT3 was increased in Al (OH) 3 -stimulated monocytes (Supplementary S7), as was the gene expression of Mx1. This provides evidence that IFNβ indeed has been secreted, since these proteins would otherwise not have been increased. These proteins could be induced by IFNβ via STAT1. Indeed, gene expression of STAT1 was upregulated in Al(OH) 3 -stimulated monocytes ( Fig. 1 and Supplementary S5). These data confirm the secretion and consumption of IFNβ.
Indications of IFNβ transcription were also found since the gene expression of IRF7 a key transcriptional regulator was found to be increased (Fig. 1). IRF3 gene expression was increased in two donors. Genes involved in the pathways via which IRF7 can mediate the induction of type I interferons were increased, i.e. TICAM1 and IRAK1 while MyD88, and TRAF6 clearly trended towards upregulation (Table 1 and Supplementary S5) [43]. The protein TBK1 in the IRF7 signaling pathway was identified at the level of the proteome but was not increased (Supplementary S6). Upstream of both MyD88/IRAK1 and TICAM/TRAF6, TLR and NOD signaling could be involved, although Al(OH) 3 is not a ligand for both TLRs and NOD1. However, Al (OH) 3 induced cell stress and cell death (Supplementary S8 and S9) [13,16,19], resulting in the release of DAMPs/endogenous ligands. These Fig. 1. Gene expression profile of Al(OH) 3 -stimulated monocytes after 24 h. Gene expression was normalized to a housekeeping gene. This figure depicts the LOG10(normalized) gene expression profile after 24 h of Al(OH) 3 stimulation (10 μg/ml) compared to the control, of two representative biological replicas (for an overview of all the altered genes in three out of five donors see Table 1). The genes above the upper dotted line are significantly upregulated by at least a factor of 2 as a result of the Al(OH) 3 stimulation, while the genes below the lower dotted lines are downregulated by at least a factor of 2 due to the Al(OH) 3 stimulation . Italic genes: genes that are upregulated/downregulated in 2 donors and a trend towards upregulation in at least one additional donor. 1 Annotated genes that are downregulated by at least a factor of 2 in at least 3 donors. 2 Annotated genes that are upregulated by at least factor of 2 in at least 3 donors. endogenous ligands can bind various receptors leading to the secretion of type I interferons. Gene expression of a subset of these DAMP sensors was identified in this data set, specifically DDX58, (no further pathway evidence identified) TLR2 (upregulated in two donors but downregulated in one), TLR3 (upregulated in four donors) and NOD1 (in two donors tending towards upregulation in at least one more) (Fig. 1). TLR3 can be activated by endogenous mRNA [44][45][46] resulting in the secretion of type I interferons, via the induction of TICAM1, TRAF6, TBK1 and IRF7 [44,45,47]. TLR2 (upregulated in two donors and downregulated in one donor) can be induced by endogenous ligands and induce the secretion of type I interferons via the pathway described by Dietrich et al. [48]. NOD1 can also be induced by endogenous ligands and results in the formation of type I interferons [49,50]. These data indicate the TLR signaling and NOD1 signaling could be involved in the induction of type I interferons by DAMP sensing. Although involvement of other DNA sensors cannot be excluded, hypothetical Al(OH) 3 -induced pathways resulting in the release of type I interferons, based on the data described and the literature are depicted in Fig. 5.

Discussion
A comprehensive analysis of the effects of Al(OH) 3 on monocytes revealed two novel cellular processes to be involved in the adjuvant action of Al(OH) 3 : first, IFNβ secretion via NOD1 or TLR signaling and second, the upregulation of antigen processing and presentation via HLA class I-related proteins, potentially leading to more antigen processing and presentation when an antigen is present. Together with a deeper elucidation of known Al(OH) 3 -affected mechanisms, these novel processes emphasize that Al(OH) 3 -stimulated monocytes prepare for a pro-inflammatory function and antigen presentation in the immune response [22][23][24].
Known mechanisms of Al(OH) 3 (as reviewed by He et al. [12]) were elucidated in more detail: inflammasome activation, complement activation, cytokine induction, monocyte activation and differentiation and induced HLA-class II activation possibly due to increased IFNγ gene expression. It is noteworthy that IFNγ gene expression is induced as a result of Al(OH) 3 stimulation. As the gene targets were measured at mRNA level and the proteins induced by these measured genes were determined to be increased by proteomics-enabled mass spectrometry, it can be stated that the genes actually have been translated into proteins and subsequently induced downstream pathways; e.g. IFNβ induction was measured at the gene expression level and Mx1, Mx2 and IFIT3 were measured at protein level (Table 2 and Supplementary S7), these proteins are specifically induced by type I interferons [32,33]. By combining the transcriptome and proteome analysis, pathways can be identified in more detail, despite the sometimes relative low coherence between protein and gene expression with respect to the same Uniprot entry. This might be related to the limited set of genes investigated, due to differences in kinetics or due to the fact that secreted proteins probably have not been identified in the cell lysates. By combining these techniques it is possible to analyze activated pathways, based on the identification of selected cytokines and downstream proteins. One of the novel processes identified is the induction of IFNβ secretion resulting from Al(OH) 3 stimulation, most likely due to TLR2, TLR3 NOD like or other DAMP signaling. IFNβ gene expression was increased in Al(OH) 3 -stimulated monocytes. Downstream of IFNβ, various proteins were induced: e.g. Mx1, Mx2 and IFIT3 were increased in Al(OH) 3 -stimulated monocytes. These proteins can be induced by IFNβ via STAT1. Indeed, gene expression of STAT1 was upregulated in Al(OH) 3 -stimulated monocytes (Fig. 1). Upstream of IFNβ, IRF7 gene expression was increased as was the expression of IRF3 in two biological replicates. These inducers of type I interferon secretion can be activated via various pathways. Multiple of these inducers were identified and upregulated at the level of gene expression: TICAM1, TRAF6, MyD88 and IRAK1. Upstream of these molecules, TLR signaling can be involved. In this data set, TLR3 was upregulated at the level of gene expression. TLR2 was upregulated in two donors (Table 1). Al(OH) 3 is not a direct ligand for either of these pathways, however Al(OH) 3 causes the release of endogenous ligands by inducing cell death [13,16,19]. Indeed, markers of cell stress and cell death were identified in the protein dataset in this study (Fig. 3A, B, Supplementary S6-S9). These endogenous ligands can activate TLR3 and downstream signaling [44][45][46] and TLR2 directly or indirectly via TLR3 [51][52][53][54]. The activation of TLR3 results in recruitment of TICAM1 followed by the activation of IRF3/IRF7 leading to type I Interferon transcription. The activation of TLR2 can result in the formation of a complex with MyD88 and the internalization of this complex towards the endolysosome where this complex binds IRAK1. Subsequently, IRAK1 phosphorylates IRF3/IRF7, resulting in the transcription and secretion of type I interferons [48]. Many of the genes involved in these pathways were found to be upregulated as can be seen in Fig. 5. Since TLR2 was upregulated in two donors but downregulated in one donor, the pathway via TLR3 is Fig. 2. Heatmap of all Al(OH) 3 -regulated proteins in monocytes. The heatmap represents the median of all upregulated (red) and downregulated (green) proteins from 3 biological replicates on a LOG2 scale after 24 and 48 h of Al(OH) 3 stimulation (10 μg/ml). Regulated proteins were at least altered by a factor 1.5 compared to the control in at least two out of three biological replicates. Enrichment analysis of GO biological processes was performed on the regulated proteins; depicted processes have FDRs < 0.05. The proteins were clustered by GO term as described in Supplementary S7. The GO terms are biological processes that are significantly overexpressed (p-values < 0.1) in the upregulated protein data set after 24 h of Al(OH) 3 stimulation (10 μg/ml) (3A) and 48 h (3B) from three biological replicates. The bars depict the percentage of upregulated proteins out of the total proteins in this GO term. The numbers display the ratios between upregulated proteins and total proteins in this GO term and the numbers between brackets are referring to the specified GO term. more likely to be involved in type I Interferon induction than the pathway via TLR2.
Detection of cytosolic DNA or other DAMPs is another possible cause of IFNβ induction [16]. Two of those DAMP sensors are NOD1 and DDX58 [19,49,50]. NOD1 gene expression was found to be increased. Activation of NOD1 results in a similar pathway as the TLR2/ TLR3 signaling pathway (Fig. 7). DDX58 gene expression was upregulated in this data set, however, no further evidence of the activation of this pathway was identified. This is in agreement with the data from Marichal et al. describing that the DDX58-STING pathway might not be involved in the adjuvant effect of Al(OH) 3 [12,19]. However, based on this data this pathway cannot be excluded. TLR8 and TLR9 are other potential DAMP sensors. Protein expression of TLR8 was downregulated and the gene expression of TLR8 and TLR9 was not increased in Al(OH) 3 -stimulated monocytes, probably excluding TLR8 and TLR9 from a role in the adjuvant effect of Al(OH) 3. These results show that Al (OH) 3 -induced DAMPS could be involved in the formation of IFNβ, most likely by activation TLR2 and/or TLR3 and by NOD1 signaling pathways, since many genes and proteins were upregulated in Al(OH) 3stimulated cells that belong to these pathways. Other potential sensors for DNA in the cytoplasm signaling via the Caspase recruitment and activation domain (CARD) were not identified in the proteomics data, but cannot be excluded from being involved in Interferon secretion. It has been described that TLR signaling is not involved in the Ab response in Al(OH) 3 -stimulated DCs [12,15], however, a potential role for TLR signaling in the induction of interferons has been described earlier [16], as is the interaction with the inflammasome [55][56][57]. This data indicate that TLR signaling can be involved in the adjuvant effect of Al(OH) 3 .
The induction and secretion of type I interferons can be related to processes described above, amongst others: differentiation away from a monocytic cell type, chemotaxis via the induction of CCL2, the increase of HLA class I and class II, activate complement and induce or inhibit the secretion of various cytokines as described previously [58][59][60]. This implies that type I interferons are crucial mediators of the innate immune response induced by Al(OH) 3 . In addition, secretion of type I interferons (IFNβ) after Al(OH) 3 stimulation, might have several effects if the effects induce innate memory. Innate memory is the phenomenon of innate immune cells, like NK cells and macrophages, that are permanently changed after stimulation. Interferons play a role in inducing this innate memory [61]. First, type I interferons induced via TLR signaling are essential in the host response to virus. Second, they prime NK cells and macrophages [62,63], of major importance for an early clearance of infections and the maturation of Th1 cells, which can be needed to induce memory [64,65]. Third, type I interferons are essential for the survival of both CD4 + and CD8 + T cells [63], but most relevant for the immune response upon vaccination is the stimulation of local B cells by type I Interferons [66]. Therefore, IFNβ secretion can be important in the adjuvant effect of Al(OH) 3 .
The other novel mechanism identified in this study is the increased activation of the antigen presentation pathway via HLA class I. It was recently described that activated monocytes have a fundamental role in antigen presentation to T cells in the lymph nodes and that antigen presenting monocytes can in fact cross present to HLA class I [24]. This study showed an upregulation of HLA class I molecules HLA-A and HLA-E, on both the level of gene expression and protein expression. Moreover, an increased protein expression of HM13 and an increased gene expression of β2M and IFNγ was observed, all related to antigen presentation via HLA class I [41,67]. These results show that Al(OH) 3 increased the expression of HLA class I molecules on monocytes which leads to enhanced antigen presentation. Although Al(OH) 3 is capable of activating CD8 + T cells, these T cells will likely not develop into mature cytotoxic T cells but rather into CD8 + memory T cells [8,13,[68][69][70]. Nevertheless, Al(OH) 3 is mainly considered an inducer of CD4 + T cells [5,13,71]. HLA-E plays a role in the stimulation of NK cells. NK cells specifically recognize the HLA-E-peptide complex, which can inhibit and excite the response, depending if the peptide is self or foreign, such as a viral peptide [72,73]. This is a useful process in inducing the immune response or preventing autoimmunity.
Besides increased expression of HLA class I molecules, an increased expression of components of the HLA class II pathway after Al(OH) 3 stimulation was observed, like Legumain, Cathepsins S, L, B and D. There is conflicting literature regarding the effect of Al(OH) 3 on HLA class II expression [5,6,74]. The increased expression of HLA class II molecules by Al(OH) 3 might be due to IFNγ signaling [41]. Indeed, in our study mRNA of IFNγ was increased. Moreover, protein expression of IFNγR1, and IFI30, a downstream molecule of the IFNγ signaling pathway, was increased. IFNγ-induced increase of HLA class II can be synergized by IL-3 and the IL-3R receptor, of which protein expression was increased [42]. The detection of several Cathepsins (S, L, D) and IFI30, which is required for antigen presentation via HLA class II, supports the hypothesis that Al(OH) 3 increases HLA class II antigen presentation by monocytes and that the increase of HLA class II molecules in monocytes might be due to the induction of IFNγ and subsequent signaling.  5. Hypothetical pathways resulting in IFNβ secretion, potentially via TLR2/TLR3 and NOD1 pathways that become activated upon Al (OH) 3 stimulation of monocytes. TLR2 and TLR3 can be activated by DAMPS, after which two pathways can be activated: one leading to the formation of NFκB and the second pathway leading to the formation of IFNβ and downstream proteins. NOD1 activation leads to the secretion of IFNβ. Genes are annotated in circles, proteins are annotated in boxes.
As described previously, Al(OH) 3 induced the downregulation of CD14 [6]. This finding indicates that Al(OH) 3 -stimulated monocytes are starting to differentiate away from a monocytic cell type and that the cells become activated, which was confirmed by the increase of activation marker CD71 for monocytes. For the response to Al(OH) 3 it was expected that the response would be well measureable after 24 h, since the responses towards LPS could still be detected after 24 h; this time point was also based on the study of Ulanova et al. [75][76][77].
Induction of cytokines by an adjuvant is important in steering the immune response. Al(OH) 3 is described as an adjuvant that primes for a Th2 response by inducing IL-4 [5]. Indeed, our results show production of the Th2 polarizing cytokine IL-4 by monocytes upon Al(OH) 3 stimulation. Presumably, the IL-4 response is dominant and may inhibit Th1 polarization In response to Al(OH) 3 . However, innate IFNγ is associated with a Th1 response [36,69]. A more Th1-related response to Al(OH) 3 was observed in multiple studies related to Al(OH) 3 [69]. IFNγ was secreted in IL-4 knockout mice [9] resulting in a more Th1-related response. IFNγ has more functions then solely being a Th1-associated cytokine. Thus, the polarization towards a Th2 type response occurs upon Al(OH) 3 stimulation despite the IFNγ produced.
One of the known mechanisms in the adjuvant activity of Al(OH) 3 besides preparing the innate immune system for antigen presentation and T cell activation, is the activation of the inflammasome leading to the secretion of IL-1β. This was confirmed in primary monocytes, by the increased gene expression of NOD1, IL1R1 and a trend in CASP1 and MyD88. In addition, a trend towards increased IL-1β secretion was observed in primary monocytes (2 donors) as was the reduction of IL-1β secretion when the inflammasome was blocked (1 donor). In PMAprimed THP-1 cells, an inflammasome inhibition experiment was performed that resulted in a reduced IL-1β secretion. PMA-primed THP-1 cells were used to further elucidate the inflammasome dependent origin of IL-β. THP-1 cells were used because, the levels of IL-1β detected in primary monocytes were low and THP-1 cells are a useful model since PMA priming causes the production of pro IL-1β and subsequent exposure to Al(OH) 3 can increase actual IL-1β, mimicking the two step activation required for inflammasome induction [78][79][80]. THP-1 cells are immortalized cells that differ in sensitivity form primary monocytes in response to stimuli such as LPS, due to differences in the cell surface marker expression of [81,82]. However, the confirmation of the role of the inflammasome pathway in the IL-1β secretion in THP-1 cells suggests a similar involvement of the inflammasome in primary monocytes, since leads were also found in the primary monocyte data.
Complement activation is another mechanism described to be induced by Al(OH) 3 [18], either by Al(OH) 3 directly or by dying cells; genes and proteins related to the classical, alternative and lectin pathways were upregulated in this dataset [18]. In our study, transcriptome data revealed the upregulation of C3 showing the initiation of the complement cascade, while at the level of the proteome C4, C5A and C8 were upregulated, representing the different complement pathways. These results show that combined use of transcriptomics and proteomics is valuable. At the transcriptome level, the initiation of the complement cascade was observed while with the proteome data the specific pathways were determined. The activation of complement pathways can be useful for bacterial lysis, and clearing apoptotic cells, which could prevent autoimmunity [83].
This study demonstrates that the response to Al(OH) 3 is very complex (Fig. 6), but also indicates that there might be considerable differences between in vivo and in vitro models. This might be caused by the use of a single cell type in vitro versus the availability of an intact, coherent immune system in vivo. Also, the species difference between cells from humans and mice might play a role. In mouse models, it was observed that Al(OH) 3 skews to a Th2 response [84]. In this study we indeed observed a polarization of the immune response towards a Th2 response, but in the presence of the Th1-associated gene IFNγ.

Conclusion
The immune response is a complex interplay of many molecules, residing in different tissues and cellular compartments. Hence, our model system, utilizing primary monocytes, does not include the coherent adaptive immune response upon Al(OH) 3 stimulus. Moreover, the present study did not include a particular antigen adsorbed to Al (OH) 3 which might also contribute to the innate immune response induced by Al(OH) 3 .
Our systems biology-based approach revealed a detailed overview of the molecular mechanisms of Al(OH) 3 as an adjuvant for monocyte responsiveness (Fig. 6). Most important, the monocyte response to Al (OH) 3 is broad as illustrated by the induction of Th2-polarizing factors, the induction of type I and type II interferons and the upregulation of both HLA class I and HLA class II pathways, even in the absence of antigens. In this study, two new modes of action of Al(OH) 3 were elucidated, i.e. the upregulation of IFNγ in relation to HLA regulation and antigen presentation and the IFNβ secretion possibly as a result of TLR2, TLR3 or NOD1 activation. Finally, by combining complementary techniques, e.g. functional assays, transcriptome data and proteome data, it was possible to create a detailed overview of cellular processes in response to Al(OH) 3 , which are important in the adjuvant effect of Al (OH) 3 .

Transparency document
The Transparency document associated with this article can be found, in online version.