Identification of Compounds that Prolong Type I Interferon Signaling as Potential Vaccine Adjuvants

Abstract

Vaccines are reliant on adjuvants to enhance the immune stimulus and type I interferons (IFNs) have been shown to be beneficial in augmenting this response. We were interested to identify compounds that would sustain activation of an endogenous type I IFN response as a co-adjuvant. We began with generation of a human monocytic THP-1 cell-line with an IFN-stimulated response element (ISRE)-beta-lactamase reporter construct for high-throughput screening. Pilot studies were performed to optimize the parameters and conditions for this cell-based Förster resonance energy transfer (FRET) reporter assay for sustaining an IFN-α induced ISRE activation signal. These conditions were confirmed in an initial pilot screen, followed by the main screen for evaluating prolongation of an IFN-α induced ISRE activation signal at 16 h. Hit compounds were identified using a structure enrichment strategy based on chemoinformatic clustering and a naïve ‘Top X’ approach. A select list of confirmed hits was then evaluated for toxicity and the ability to sustain IFN activity by gene and protein expression. Finally, for proof-of-concept, a panel of compounds was used to immunize mice as co-adjuvant with a model antigen and an interferon inducing Toll-like receptor 4 agonist, lipopolysaccharide as an adjuvant. Selected compounds significantly augmented antigen-specific immunoglobulin responses.

Introduction:

Much of vaccine development has focused on the mixture of antigens necessary to evoke a protective immune response. Equally important to the antigen components, however, is the composition of the accompanying adjuvant required to augment the strength of the response without harmful toxicity. The use of appropriate adjuvants is particularly important for at risk populations with reduced immune responses such as the elderly. In 2015, the U.S. Food and Drug Administration (FDA) approved Fluad™, the first seasonal influenza vaccine containing an adjuvant marketed in the United States to improve efficacy in the elderly population.^{1, 2} Fluad™ is formulated with the adjuvant MF59, an oil-in-water emulsion of squalene oil which has been shown to act by differentiation of dendritic cells (DCs) and enhancement of antigen uptake.^{3, 4} Additional agents to enhance and prolong immune responses in conjunction with the antigen preserving components may result in further reductions in mortality from infections, including influenza.

Type I IFN has been shown to be a potent adjuvant when co-administered with a human influenza vaccine.^{5, 6} Hence we aimed to discover well-characterized small molecules that would prolong the activation signal from endogenously produced type I IFN. When a foreign antigen is introduced, it is engulfed by antigen presenting cells (APCs) and trafficked into the draining lymph node, where an adaptive immune response is primed in contrast to the primary effect of an adjuvant at the local site of injection.⁷ We hypothesized that prolonging the IFN response would enhance the antigen presenting activity after migration of APCs to the lymph node beyond the usual window of signal decay.

Although there are multiple isoforms of type I interferon alpha (IFN-α) and one isoform of IFN beta (IFN-β), they bind a common receptor on the cell surface.⁸ The type I IFN receptor (IFNAR) is composed of two subunits, IFNAR1 and IFNAR2, which are associated with the Janus activated kinases (JAKs) tyrosine kinase 2 (TYK2) and JAK1, respectively.^{9, 10} Activation of the JAKs that are associated with the type I IFN receptor results in tyrosine phosphorylation of STAT1 (signal transducer and activator of transcription 1) and STAT2; this leads to the formation of IFN-stimulated gene factor 3 (ISGF3) complexes comprised of STAT1–STAT2–IRF9 multimers.¹¹ These complexes translocate to the nucleus and bind IFN-stimulated response elements (ISREs) to initiate gene transcription.^{12, 13} An increase in expression of genes associated with this IFN gene signature including LY6E, MX1, OAS3, IFI44L, IFI6 and IFITM3 has been linked to the early phases after vaccination.¹⁴ The duration of the interferon response is tightly regulated at the transcriptional and protein levels.

Prior high throughput screens (HTS) sought to discover compounds that directly induced IFN up to 8 hours as novel antiviral agents.^15–17 However, here we aimed to identify small molecules that would prolong an existing IFN signal in an APC during its transition to the draining lymph node to enhance adaptive immune responses. We previously identified pyrimidoindoles and 4-aminoquinazolines as potent adjuvants through an HTS using a Förster resonance energy transfer (FRET) based NF-κB reporter assay in the human APC-like monocytic cell line, THP-1.^{18, 19} Using this same reporter cell line, we recently demonstrated the overall feasibility of HTS for identifying co-adjuvants that prolonged NF-κB signaling.¹⁹ Based on this success, we developed a THP-1 cell line with a robust ISRE reporter construct, which was then used in a HTS, where IFN-α was added concurrently with the screening compounds and ISRE activity was assayed after 16 h. This strategy revealed a variety of chemotypes that improved vaccine efficacy as co-adjuvants in a murine model using lipopolysaccharide (LPS, an adjuvant) and ovalbumin (OVA, a test antigen).

Material and Methods:

ISRE-bla THP-1 cells

The ISRE-beta lactamase (bla) reporter human monocytic THP-1 cell line was developed with Thermo Fisher Scientific (formerly Life Technologies, Carlsbad, CA). An existing ISRE-bla reporter lentiviral construct was transduced into THP-1 cells (ISRE-bla THP-1). A stable pool of cells expressing the construct was generated following blasticidin-selection. Clones were selected by pulsing the cells with LiveBLAzer™ FRET-B/G and sorting by flow cytometry. The ISRE-bla THP-1 cells were maintained in culture in complete RPMI 1640 growth medium supplemented with 10% fetal bovine serum (FBS, Cat# 26400–036, Thermo Fisher Scientific), penicillin-streptomycin (100 Units/mL and 100 μg/mL respectively), 1% Minimum essential medium with non-essential amino acid (MEM-NEAA, Thermo Fisher Scientific) and blasticidin (5 μg/mL) at 37°C in 5% CO₂. All the assay validations were carried out in assay medium OptiMEM® I Reduced Serum Medium (Cat# 31985–070, Thermo Fisher Scientific) in 384-well plates (Cat# 3712, Corning). Serial dilutions of human IFN-α (Cat #11101–1, PBL Assay Science) were carried out in assay medium without FBS.

Assays for interferon signaling in ISRE-bla THP-1 cells

For over-expression of USP18, human USP18 was cloned into the MSCV-IRES-Puro (MIP) retroviral vector.²⁰ For stable knockdown of USP18, control short hairpin RNA (shCTRL) or USP18-specific shRNA were cloned into the pSuper-retro-puro retroviral vector.²¹ Transfection was conducted by using PEI (polyethylenimine).²² For retrovirus production, 293T cells were co-transfected with plasmids encoding viral vectors and packaging vectors pCL-10A1. Viral particles were collected 48 h after transfection, filtered with 0.45 μm sterile filter. For the retrovirus infection, spin infection [2000×g, 3 h, 30°C; Allegra X12R (Beckman Coulter)] in the presence of polybrene (8 μg/mL) was performed. Single colonies were isolated by culturing cells in a 96-well plate and tested for the knockdown of USP18. ISRE-bla THP-1 transduced cells were treated with 1000 Units/mL of human IFN-α, lysed in radioimmuno-precipitation assay (RIPA) buffer composed of 25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM EDTA, and 0.5% Nonidet P-40 with proteinase and phosphatase inhibitors (Roche). The cell lysates were denatured in 1x sample buffer [50 mM Tris-HCl (pH 6.8), 2% SDS, 2-mercaptethanol, 10% glycerol, and 1% bromophenol blue] for 5 min at 100 °C. The proteins were separated by SDS-PAGE, blotted onto methyl cellulose membrane, and probed with the indicated antibodies:anti-phospho-STAT1 (Tyr 701) (Cell Signaling Technology), anti-STAT1 (Cell Signaling Technology), anti-USP18,²³ and anti-tubulin (Sigma Aldrich). For the quantification of protein blots, signals were detected with the LI-COR Odyssey system.

Compound library for HTS and purchased compounds

A library of diverse compounds was acquired from the UCSF Small Molecule Discovery Center consisting of 172,145 chemical entities from ten suppliers (Supplementary Figure S1). Hit compounds were purchased from the vendors listed in Supplementary Table S2.

High throughput screen

HTS was performed by Thermo Fisher Scientific at their commercial robotics facility (Madison, WI). Frozen ISRE-bla THP-1 cells were first thawed and incubated in assay media containing 5% FBS for 24 h. Then cells were transferred to the 384-well assay plates at a density of 40,000 cells/well, except for the cell-free background control wells. Human type I IFN-α was added to the “IFN 16h” control wells and to the test wells. At the same time, test compounds at final concentration of 5 μM were added to the test wells. After 10 h incubation, IFN-α was added to the “IFN 6h” control wells. After an additional 6 h the beta lactamase substrate was added to all wells, giving 16 h total incubation for the “IFN 16h” wells and 6 h incubation for the “IFN 6h” activation control wells. The fluorescence emission at 535 nm and 465 nm was assayed after 3 h further incubation with LiveBLAzer™ FRET-B/G (CCF4-AM) beta-lactamase substrate.

Four types of control wells were included: cell free wells for background estimation, “No IFN+ cells” wells to estimate the readout of unstimulated cells, “IFN 16h” wells to estimate the readout of the ISRE reporter gene after 16 h of incubation, during which time the signal will decay and an additional control “IFN 6h” wells for ensuring activation of ISRE-bla THP-1 cells. There were 46 plates in the pilot screen and 572 plates screened over 8 assay runs in the main screen.

Measurement of ISRE activation in ISRE-bla THP-1 cells

Fluorescence was measured at an excitation wavelength of 405 nm, and emission wavelengths of 465 nm and 535 nm. In each assay plate, negative control wells were included that were cell-free and cells with vehicle i.e. 0.5% DMSO (No IFN+cells). The background fluorescence values were subtracted from the raw fluorescence values (cell free wells at the same fluorescence wavelength) and emission ratios were calculated by dividing background-subtracted values at emission wavelength of 465 nm to emission wavelength of 535 nm. The response ratios were calculated as follows: [(emission ratio of a test well)/(average emission ratio of “No IFN+cells” wells (0.5% DMSO))]. The percent activation was calculated using “IFN 6h” as reference. The activation values were not standardized to “IFN 16h” due to significant variability of the “IFN 16h” values.

For dose titration experiments, cells were plated in 96 well plates (10⁵ cells/well) and treated with serially diluted compounds in the presence of vehicle or 3 different doses of IFN-α (20 U/mL, 100 U/mL, and 500 U/mL) for 16h. Substrate was added and incubated for 3 h. Fluorescence was measured, and emission ratios calculated as described above.

Cell viability assay

THP-1 cells were dispensed in 96-well plates (10⁵ cells/well) and treated with 5 μM of each compound. After 18 h of incubation, a solution of 3-[4,5-dimethylthiazol-2-yl]-2,5-dipheyl tetrazolium bromide (MTT, Thermo Fisher Scientific) in assay media (0.5 mg/ml) was added to each well and further incubated for 4–6 hours, followed by addition of cell lysis buffer (15% w/v SDS and 0.12% v/v 12N HCl aqueous solution), incubated overnight and then absorbance measured at 570 nm using 650 nm as reference with a plate reader.

Quantigene assay and immunoblots

Murine bone marrow derived dendritic cells prepared from wild type (WT) C57BL/6 mice were plated at a density of 2 ×10⁵ cells/well in a 96-well plate. The cells were then treated with 5 μM final concentration of each compound. IFN-α (100 Units/mL) was then added and the plates were incubated for 10 h. The cells were then lysed and analyzed by Quantigene assay (Thermo Fisher Scientific) for expression of Isg56, Ifi2712a, Ifi16, Isg15, and Irf7 in addition to Rps20, Hprt, and Rpl19 genes used as reference. Gene expression in human THP-1 cells was performed using lower concentrations of select compounds (1 μM of 3, 5, 6, 24) in presence of human IFN-α (1000 Units/mL) at an incubation time of 8 hours. ISG15 and ISG56 gene expression was quantified using q-PCR technique using GAPDH as reference gene. Immunoblots were performed as mentioned above.

In vivo adjuvant activity study

Seven to nine-week-old WT C57BL/6mice were purchased from The Jackson Laboratories (Bar Harbor, MA). All animal experiments received prior approval from the UCSD Institutional Animal Care and Use Committee. The mice (n=5 per group) were immunized in the gastronemius muscle with ovalbumin (20 μg/animal, Worthington Biochemical Co. Lakewood, NJ) mixed with LPS (3 μg/animal, L2654, Sigma Aldrich) and one of the compounds 1–12, or 14–15, or 24–25 (100 nmol/animal), or DMSO as negative control on days 0 and 7. On day 17, immunized mice were bled and antigen specific IgG1 and IgG2c were measured by ELISA as previously described.²⁴

Statistical analysis

Procedures to identification of hits from the HTS using a structural enrichment method are detailed in the results section under Pilot, main and confirmation screen. Data for in vitro studies are represented as mean ± standard deviation (SD). Prism 6 (GraphPad Software, San Diego, CA) statistical software was used to obtain p-values for comparison between groups (p<0.05 was considered significant) for in vivo study. One-way ANOVA followed by Dunnett’s post hoc test was used to compare multiple groups against the vehicle group.²⁵

Results:

Overall HTS strategy and design.

We devised a cell-based HTS assay to broadly cover cellular pathways that might prolong an IFN-induced signal. We first developed a cell line with an ISRE reporter gene and optimized assay conditions to accommodate longer cell incubation times needed to assess prolonged activation of IFN-α. Extensive preliminary experiments were then performed to optimize this newly created THP-1 ISRE-bla cell line for HTS assays. This was followed by an initial pilot screen with 14,597 compounds. This assay was then used in a HTS main screen with 172,145 compounds, and potential hits were rescreened in duplicate in a confirmation screen of 2,026 compounds. Additional validation screens for confirmed hits included toxicity studies and gene and protein analysis. This approach led to the discovery of several chemical scaffolds that were able to sustain a type I IFN response. The overall value of this screening strategy was confirmed by the ability of selected compounds to augment a vaccination response to a test antigen in mice in the presence of a known interferon-inducing compound (Fig. 1).

Figure 1.

HTS workflow strategy. The cell-based assay using THP-1 ISRE-bla reporter cells was optimized for identification of compounds that prolonged ISRE activation when stimulated with IFN-α for 16 h. First, the ISRE-bla THP-1 cells were generated and a feasibility test was performed to validate the transduction of ISRE-bla gene responsive to activation with IFN-α. The assay conditions were then optimized for HTS and a pilot screen of 14,597 compounds was run to validate the approach. In the main library screen, 172,145 compounds were screened in the presence of IFN-α for ISRE activation at 16 h. Cluster based statistical analysis yielded 2,026 compounds, which were retested in a confirmation screen at 16 h with IFN-α in duplicate. 265 compounds that had confirmed activity at 16 h were structurally clustered using similarity cutoff index of 0.5. Several large families of chemotypes were identified as hits. The selected compounds from these chemotypes were subjected to cell-based toxicity studies, in vitro gene and protein analysis as well as in vivo co-adjuvanticity studies in mice, which led to identification of key lead molecules. The number in parentheses corresponds to number of compounds.

Generation of ISRE-bla THP-1 cells and feasibility studies.

We selected the human THP-1 cell line because it had previously been utilized for robotic HTS and is a monocytic cell line with features characteristic of antigen presenting cells (APCs).^{18, 19, 26} THP-1 cells were transduced with an existing validated ISRE beta-lactamase reporter construct by Thermo Fisher Scientific. The stable pool of THP-1 cells expressing ISRE-bla was further expanded and sorted by FACS to obtain single cell clones. For the initial screen, a total of 36 clones were screened against a single concentration of IFN-α (10 nM). Eight top responding clones were selected and further screened against graded doses of IFN-α. EC₅₀ and response ratios of stimulation were determined for each clone. As all eight clones showed similar EC₅₀ for IFN-α stimulation, the clone, which yielded the highest response ratio, was selected for further feasibility assays. These cells have maintained FRET activity with exogenous IFN up to nine passages.

The THP-1 ISRE-bla cell line generated from the selected clone was propagated and first tested for the ability to regulate IFN responses. Ubiquitin specific peptidase 18 (USP18) is a protease that removes ISG15 from substrate proteins and also functions as a potent inhibitor of type I IFN signaling.^{20, 27, 28} This mechanism was tested in the THP-1 ISRE-bla cell line by overexpression and knock down approaches. Overexpression of USP18 using retroviral transduction (MIP-USP18) reduced the level of STAT1 phosphorylation upon IFN-α treatment (Fig. 2A). Conversely, USP18 silencing by introducing a short hairpin RNA (shRNA) prolonged STAT1 phosphorylation (Fig. 2B). Hence the THP-1 ISRE-bla cell line biologically responded to IFN-α and had potential to detect compounds that altered IFN signal regulation (demonstrated here by alterations in USP18 levels).

Figure 2.

Cell feasibility tests. The expression of ISRE in THP-1 cells was confirmed using (A) over-expression and (B) knockdown studies of USP18, by examining the level of STAT1 phosphorylation. Response of ISRE-bla THP-1 cells was tested for the following conditions (C) 40,000; 20,000; 10,000; or 5,000 cells/well (384-well plate) at serum concentration of 2.5% FBS, 5 h incubation time, and 3 h incubation with the substrate; (D) serum concentration of 1%, 2.5%, 5%, or 10% with cell density of 40,000 cells/well, 5 h incubation time, and 3 h incubation with the substrate; (E) substrate incubation time of 1 h, 1.5 h, 2 h, or 3 h with cell density of 40,000 cells/well, 5% FBS and 5 h incubation time; and (F) incubation time of 6 h, 12 h, 20 h, or 30 h with cell density of 40,000 cells/well, 5% FBS, and 3 h incubation time with the substrate.

To assess the feasibility of HTS using the newly generated THP-1 ISRE-bla cell line, cells were tested for responses over graded doses of human IFN-α2A subject to the following variables: 1) varying cell counts including 5,000, 10,000, 20,000 and 40,000 cells/well; 2) FBS concentration at 1.0%, 2.5%, 5.0% and 10%; 3) substrate incubation time of 60, 90, 120 and 180 minutes; and 4) total stimulation time with IFN-α treatment of 6, 12, 20 and 30 h (assay stimulation time) (Fig. 2C–2F). The signal increased with increasing cell counts (Fig. 2C) and increasing substrate incubation time (Fig. 2E) suggesting a direct relationship between activation, cell density and exposure to substrate. The serum concentration did not show any appreciable effect on activation (Fig. 2D). With regard to assay incubation time, the signal peaked at 6 h and began to fall over time, as expected (Fig. 2F). These feasibility studies performed with both the fresh and cryo-preserved cells showed that the activation signal was sufficiently robust to further evaluate this THP-1 ISRE-bla cell line for assay optimization to be used for the HTS.

Assay Optimization for HTS:

In order to avoid variability in assay performed over different days, the HTS experiments required the use of frozen cells (from the same passage cycle), which were thawed and plated prior to the screens using the robotic system. We conducted an additional series of experiments with the thawed cells to optimize the assay conditions for the HTS. We first examined the following parameters: 1) assay incubation time post IFN-α addition (6 h, 12 h, 30 h); 2) cell counts (20,000 vs. 40,000 cells/well); 3) serum concentration in assay media (5% vs. 2.5%); and 4) the optimal dose of IFN-α for both the test wells and the full stimulation control wells (Fig 3A). The substrate incubation time was kept fixed at 3 hours. Within each of these experimental conditions there were dose response curves against serial dilutions beginning with a concentration of 100 nM IFN-α and an additional control “IFN-α 6h” for 100 nM IFN-α added 6 hours prior to the total assay incubation time.

Figure 3.

Assay optimization for HTS. The assay conditions were optimized for cell density, serum concentration and incubation time. (A) Broad scanning of assay conditions shows peak activation at 12 h and complete loss of activation at 30 h for IFN-α dose response curve with starting concentration of 100 nM. The cell density of 40,000 cells/well showed marginally better response than 20,000 cells/well for both the IFN-α dose response and “IFN 6h” control. (B) Fine tuning for incubation time at 50 nM IFN-α concentration showed 16 h to be ideal for both decayed IFN-α and “IFN 6h” response. (C) The schematic shows the optimum assay parameters and the assay protocol used for the HTS. Briefly, the cells were plated at a density of 40,000 cells/well in 5% FBS 24h prior to addition of 50 nM IFN-α. An additional control (IFN 6h) was added 6 h prior to the addition of FRET substrate at 16 h and the plates were read at 19 h.

Initial experiments, revealed that 6 h and 12 h of IFN-α stimulation were too short to see signal decay. By 30 hours the signal had completely decayed, suggesting that an incubation time of 30 h was too long to be reliable for the assay (Fig 3A). 40,000 cells/well and 5% FBS appeared to have better response than other conditions at the 12 h time point. Dose response curves indicated that good stimulation was achieved at concentrations lower than 100 nM IFN-α as well, so a second set of focused validation experiments was run using 50 nM IFN-α. The goal was to maximize the difference between the “IFN-α 6h” activation controls for 50 nM IFN-α and the decayed IFN-α signal at 16, 18 and 24 h. This difference was greatest at 16 h incubation time (Fig. 3B).

The incubation time before the cell stimulation did not significantly affect the signal. In addition, the cells tolerated up to 1% DMSO without significant change in emission ratios. An optimal assay protocol using the above standardized conditions was deemed to be following: the cells are first incubated in assay media containing 5% FBS for 24 h after thawing and then are distributed in the 384-well assay plates at a density of 40,000 cells/well (except for the cell-free control wells). IFN-α is added to the “IFN 16h” control wells and to the test wells. After 10 h incubation, IFN-α is added to the “IFN 6h” control wells. After an additional 6 h the beta-lactamase substrate is added to all wells, giving 16 h total incubation for the “IFN 16h” wells and 6 h incubation for the “IFN 6h” full activation wells. The fluorescence emissions are measured after 3 hours of further incubation with the substrate (Fig. 3C).

Pilot, main and confirmation screens:

We selected compounds representative of the chemical diversity in the compound library at UCSF SMDC. A total of 46 plates comprising 14,597 compounds were included in the pilot screen (Fig. 4A). As shown, the assay exhibited a substantial difference in “% Activation” between the two control conditions “IFN 16h” (purple circles)” exhibiting normal decay in IFN-α signaling over 16 h and “IFN 6h” (blue triangles), demonstrating maximum IFN-α signaling (Fig. 4A). This difference in control conditions indicates that the assay had adequate power to detect compounds that prolong IFN-α signaling. Encouraged by this result, the main screen which included 172,145 test compounds on 572 plates was performed over 8 assay runs (Supplementary Figure S3). All 14,597 compounds from the pilot screen were repeated for statistical analysis. After background correction, outliers were trimmed for control wells and “% Activation” for the test compounds was computed using the “IFN 6h” condition as the 100% activation control and “No IFN+cells” wells as the 0% value. Z-prime for the main HTS ranged between 0.50 and 0.90 with a median of 0.71.

Figure 4.

Compound screening. Percent activation values for the controls and test compounds from (A) the pilot screen and (B) the confirmation screen. The histogram plot to the right for each screen shows the intra-assay statistics on the percent activation values for “No IFN+cells” (0.5% DMSO, negative control), “IFN 16h” (50 nM IFN-α control added at time 0h), and “IFN 6h” (50 nM IFN-α added 6 h before the assay read out). The scatter plot shows percent activation values for all of the compounds and controls used in each screen. Each black circle represents a test compound, while controls are colored as in the histogram plot. (C) Cytotoxicity and relative expression of IFN-α stimulated genes; Left: MTT viability assay of THP-1 cells after overnight treatment with 5μM compound relative to vehicle control (Veh, 0.5% DMSO) treated cells. Relative viability was calculated as a percentage of OD at 630 nm for compound treated cells compared to vehicle treated. The OD at 630 nm for vehicle treatment was 1.63 ± 0.035 (mean ± SD); Right: Relative gene expression for Isg15 (black squares), Isg56 (yellow diamonds), Irf7 (red circles), Ifi16 (green upper triangles), and Ifi2712a (blue lower triangles) in murine BMDCs stimulated for five hours with the indicated compounds at 5 μM in the presence of 50 nM concentration of IFN-α compared to IFN-α alone control (100%). Gene expression was measured by Quantigene assay and also normalized to reference genes (Rps20, Hprt, Rpl19, box and whisker plot at the top). The chemotype information and group number for the compounds is shown on the extreme right. The y-axis corresponds to compound name/number.

We used a structural family enrichment method to identify hits.^{19, 26,29} Structural families were estimated using Murcko scaffold clusters³⁰ and Daylight fingerprint functional classes.³¹ Only Murcko clusters that had at least 5 test compounds, as members were included in the enrichment analysis. A total of 5,483 Murcko clusters were thus used and they contained 56% of all the test compounds. Functional class fingerprint clusters were obtained for all the test compounds by using k-medoids clustering with Euclidean distance, computed using the r function clara; k=300 was chosen based on estimated cluster sizes and similarity measures.³² Once compounds were clustered, all the test compounds were ranked according to their “% Activation” values. The top 2% of test compounds were identified as candidate hits. To avoid high activation merely caused by auto-fluorescence, compounds with activation values greater than 125% (i.e., mean + 4x standard deviation of all the “IFN 6h” full activation values) were not considered as candidate hits unless they were found to be included in a significant Murcko scaffold cluster. Next, each cluster (either Murcko class or functional class cluster) was scored for statistically significant enrichment based on comparing “% candidate hits” inside a cluster versus outside the cluster using Fisher’s exact test. The set of significant clusters was ranked by enrichment odds ratio and the ranked list of clusters was walked down until the desired number of hits was identified, or until clusters were judged no longer statistically significant. To attain the desired number of compounds and also to increase structural diversity, additional hits were added using a “Top X” approach, walking down the ranked list of test compounds until approximately 2,000 compounds were identified. Using this strategy, we identified 778 hits from 109 significantly enriched Murcko clusters, at a false discovery rate (FDR) of 0.25. We also identified 1,338 hits from 22 significantly enriched fingerprint clusters, at an FDR of 0.05. As 590 compounds were common by both the selection approaches, we filled out the remaining ~500 hits using the “Top X” approach to obtain a final list of 2,026 compounds for confirmation screening. To confirm the hits identified in the HTS main screen these compounds were rescreened in duplicates at 16 h with IFN-α (Fig. 4B). The confirmation screen as expected showed a higher number of hits above the decayed level of IFN-α at 16 h (“IFN 16h” controls shown in purple circles). Based on this screen, we identified 265 candidate hits using a data-driven method that fits a mixture of two linear models to the combined primary and confirmation screen “% Activation” values.^{19, 26} The confirmation rates by above methods of hit selection are provided in Supplementary Fig. S4.

Selection of compounds and bioactivity analysis of hits:

To prioritize compounds from the 265 confirmed hits for further biological analysis, we first classified compounds by sub-structure using the server based ChemMine tools (University of California, Irvine; http://chemmine.ucr.edu/tools/launch_job/Clustering/) and a binning clustering application with a similarity cutoff of 0.5.³³ This analysis allowed us to cluster these compounds into structural categories as shown in Supplementary Table S2. The number of compounds within a group (cluster size) indicates the relative scaffold distribution. Additional categories were generated from confirmed hits that did not have enough structural features for clustering (No Structural Features), and compounds that had been previously biologically characterized as known bioactives. We then chose to purchase representative confirmed hit compounds from each cluster that were available from commercial vendors (compounds 1–23). Of the 23 bioactive hit compounds, several bioactivity groups were identified including cardiac glycosides, histone deacetylase (HDAC) inhibitors, antifungals, antibacterials, and anticancer drugs such as azacytidine, topotecan, and a Src-kinase inhibitor, KX2–391.³⁴ As kinases could be attractive targets for compounds identified in this assay, we chose to purchase KX2–391 (compound 24) in addition to azacytidine (compound 25) and an HDAC inhibitor, belinostat, (compound 26) for further analyses. All of the purchased compounds were tested for purity by LC-MS techniques (Supplementary Material, Pages 10–34) and compounds with at least 95% purity were then numbered and grouped as shown in Supplementary Table S2. The physicochemical properties of these compounds as calculated using Chem3D (Version 15.1) are shown in Supplementary Table S5.

Dihydropyrazoles represented the largest group, but had been previously reported to have intrinsic fluorescence,³⁵ suggesting that these may have led to false positive readouts in the FRET based assay. To assess if the compounds in group I were fluorescent we first measured the absorbance spectra of several of these compounds at 405 nm, followed by measurement of fluorescence spectra at an excitation wavelength of 405 nm, which was used in the FRET assay. Indeed, most of the compounds showed variable fluorescence emission at 465 nm except compound 1 (Supplementary Figure S6). Thus, despite the potential for false positive discovery we selected compounds 1 (no fluorescence emission at 465 nm) and 2 (significant fluorescence emission at 465 nm) for further biological confirmation.

Initially the compounds were tested for toxicity in THP-1 cells using a MTT assay. Most of the compounds were found to be relatively non-toxic with a relative viability of more than 70% of the vehicle control (0.5% DMSO). However, both of the compounds from group XIV (triazolo[1,5-a]quinazolin-5-amines)and group XV (pyrazolo[4,3-c]quinolones) were found to decrease cell viability (<40%). Amongst the bioactive compounds, belinostat (26) was also found to be toxic (Fig. 4C, Left)

To assess the potential for the compounds to activate APCs in vivo, we evaluated their effect in primary murine bone marrow derived dendritic cells (BMDCs). Specifically, the panel of compounds was tested for the ability to augment an interferon induced gene signature pattern including Isg56, Isg15, Ifi2712a, Irf7, and Ifi16genes (Fig. 4C, Right). Gene expression was normalized to the geometric mean of three reference genes (Rps20, Hprt and Rpl19) that were not altered in the presence of IFN-α. None of the compounds showed any significant change in gene expression in the absence of IFN-α (Supplementary Figure S7). In addition, most of the non-toxic compounds also did not alter the gene expression compared to IFN-α alone. Compounds 3, 5 and 24 (KX2–391) significantly increased mRNA expression of all five genes tested (Fig. 4C, Right).

The relative potency of compounds 3, 5 and 24 in sustaining an IFN effect was tested at three doses of IFN-α (20 U/mL, 100 U/mL and 500 U/mL). Notably there was no increase in the FRET ratio even at higher concentrations of the compounds in the absence of exogenous interferon (Fig. 5A), suggesting that these compounds did not independently stimulate IFN production, concordant with the gene expression data. The relative EC₅₀s with 100 U/mL IFN were 270 nM (95% CI 146.1 to 498.6 nM), 616 nM (95% CI 137.4 to 2759 nM), 1643 nM (95% CI 1147 to 2354 nM) for compounds 3, 5 and 24 respectively (see Supplementary Table S8). The relative cytotoxicity for these compounds was verified in a dose response manner starting with a high dose of 50 μM. None of these three compounds showed any significant toxicity (% relative viability > 75%) in the absence of IFN-α (Fig. 5B). Only compound 5 showed dose dependent cytotoxicity in presence of 100 U/mL of IFN-α (CC₅₀ = 2954 nM, 95% CI 1713 to 5096 nM).

Figure 5.

Dose response curves and secondary screens. (A) Dose response ISRE activity curves for compounds 3, 5 and 24 with variable doses of IFN-α (0, 20, 100, 500 U/mL). (B) Dose response cytotoxicity curves for compounds 3, 5 and 24 with and without 100 U/mL of IFN-α. The IFN-α only controls are to the right of each plot. (C) Confirmation of ISG15 and ISG56 gene expression for 1 μM concentration of compounds 3, 5, 6, and 24 alone and in presence of IFN-α at 8 h incubation in human THP-1 cells. (D) Increased IFN-α induced protein expression of ISG15 in THP-1 cells after 24 h co-treatment with 1 μM compound 3, 5, and 24 was confirmed by immunoblot compared to 0.5% DMSO (Veh). Relative intensity compared to tubulin as the reference is shown below. (E) Sustained type I IFN induced STAT-1 phosphorylation in THP-1 cells pre-treated with compounds (final 1 μM) for 12 h and then IFN-α. Cells were lysed at the indicated times after IFN-α stimulation and analyzed for phosphorylated STAT1 and total STAT1. Relative density of phosphorylated STAT-1 to total STAT-1 is shown below as bar graph.

In order to understand the translation of gene to protein expression in human cells, we first confirmed the gene expression profile for select compounds (3, 5, and 24) in human THP-1 cells for the ISG15 and ISG56 genes. In addition, we also included compound 6 as an internal control, which was inactive in murine cells. All the active compounds caused approximately a 2-fold increase in gene expression for both ISG15 and ISG56 genes when compared to vehicle alone post 8 h incubation in the presence of 1000 U/mL of IFN-α. However, quantitative gene expression was relatively lower when the compounds alone were used for stimulation (Fig 5C). These compounds were further tested for IFN induced protein expression and signaling patterns. Human THP-1 cells were stimulated with compound alone or in combination with IFN-α and the cells were lysed after 24 h. Immunoblots demonstrated that compounds 3, 5 and 24 increased the levels of IFN-α induced ISG15, whereas the internal negative control compound 6 and vehicle treatment did not (Fig. 5D). Notably, compounds 3, 5 and 24 did not increase the amounts of ISG15 without the addition of IFN-α, suggesting that these compounds are sustaining the effect of IFN-α, but cannot inherently activate signaling (Fig. 5D). Sustaining STAT1 phosphorylation after IFN-α ligation of IFNAR1 and IFNAR2 would be one potential mechanism for this effect. The level of STAT1 phosphorylation was tested for a series of time points in the presence of IFN-α and compounds 3, 5 and 24. Cells that were treated with compounds 3, 5 and 24 had elevated levels of p-STAT1 after 30 minutes, but there was only a minimal residual effect seen with these compounds at 6 h post stimulation (Fig. 5E).

Thus, all the structurally clustered compounds that were relatively non-toxic (relative viability > 70%) and that did not show suppression of gene expression (such as compounds 13 and 16) were tested for co-adjuvant activity in murine vaccination studies. We chose lipopolysaccharide (LPS), a known TLR4 agonist and potent interferon inducer as adjuvant, and ovalbumin (OVA) as antigen for these vaccination studies. Examination of antigen-specific antibodies showed that co-immunization of LPS with some of the compounds as co-adjuvants induced statistically significant increases in antigen specific IgG2c and IgG1 titers when compared to mice immunized with LPS alone as adjuvant (Fig. 6). Based on the IgG2c titers, we could broadly classify compounds by statistical significance. Compounds that significantly enhanced IgG2c antibody responses above the vehicle control immunization (LPS + OVA) included compounds 1–3 (p<0.01) and 4–5 (p<0.05). A similar trend was observed with IgG1 titers and additionally compounds 6 and 9 also significantly boosted IgG1 antibody responses (p<0.05).

Figure 6.

Co-adjuvant activities of select compounds with LPS. Mice (n=10 per group) were immunized on days 0 and 7 with antigen ovalbumin (OVA, 20μg/animal), LPS (3μg/animal) and the indicated compound (100 nmol/animal) 1–12, or 14–15, or 24–25, or 0.5% DMSO as vehicle (Veh) control. The immunized mice were bled on day 17 and OVA-specific IgG1 (black) and IgG2c (cyan) titers were measured using ELISA. Note that the potent compounds 1-5 significantly augmented the production of ovalbumin-specific IgG2c by approximately 1.5 log fold when co-adjuvanted with LPS compared to LPS alone, and compounds 1–4, 6 and 9 significantly increased the IgG1 titers. **p<0.01 or *p<0.05 compared to the OVA+LPS group (Veh) by one-way ANOVA and Dunnett’s post hoc testing.

Discussion

There is an imperative need for potent adjuvants to enhance vaccine response. To address this unmet need we undertook a high-throughput screen based on our hypothesis that a small molecular weight compound that prolongs type I IFN signaling would function as a co-adjuvant in vaccines. To broaden our search, we utilized a cell-based approach to reflect the biologic and metabolic processes in an APC. Biological high-throughput screens suffer greater variability than targeted molecular inhibition; however our approach could provide a greater number of potential independent molecular targets. To test compounds for prolongation of type I IFN signaling in an APC we required a very robust assay system. Hence, we developed a monocytic cell line, THP-1, with an ISRE construct that was optimized for HTS. Other smaller HTS using established lung tumor or sarcoma cell lines have identified compounds that directly stimulated type I IFN production.^{15, 17} Our strategy was different in that the cells were assayed in the presence of IFN-α, as would be induced under normal vaccine conditions. Here we specifically examined the THP-1 ISRE-reporter cells after an extended incubation time in order to reveal compounds that could sustain IFN signaling.

To overcome the inherent variability flaws in cell based assays, we selected chemical scaffolds that were enriched as “hits”. This approach provided two advantages: large families suggest that there is replication in the activity, and negative data affords a structure activity relationship within the family to guide strategic structure activity designs. One advantage of a FRET based system in a prolonged HTS assay is that the baseline signal compensates for loss of cells in the individual wells. In other assay systems such as luciferase reporters a single read out cannot differentiate lack of activity from loss of cellular numbers or diminished cell function. Titrations of the test compounds to determine EC₅₀ or IC₅₀ can overcome this limitation, but the size of the screens makes such assays very cumbersome. One limitation of FRET based systems is the false positive rate due to compounds that are inherently fluorescent. One such large family of compounds, the dihydropyrazoles (Group I) were found to have this property in our screen. Rather than discarding such scaffolds, orthogonal assays for ISRE associated gene transcripts and eventual in vivo immunization confirmed potency for a few of these compounds.

The variety of molecular pathways that can enhance ISRE activity is suggested by the diverse chemical scaffolds that exerted this activity. The positive hits included some bioactive compounds, including cardiac glycosides and HDAC inhibitors. We selected belinostat, a Src kinase inhibitor (KX2–391), and an epigenetic modulator azacitidine.^36–38 Belinostat was relatively toxic in vitro, but, KX2–391 increased IFN induced gene expression without significant toxicity. In this regard, it has been reported that Src kinase inhibits TLR signaling.³⁹ Interferon has been reported to enhance the cytotoxicity of compounds that effect other signaling pathways.⁴⁰ In our study compound 5 was not intrinsically toxic, but exhibited toxicity in the micromolar range in the presence of exogenous IFN-α.

Thus, the assay successfully identified agents that enhance IFN production and the expression of IFN-induced genes. Several of the scaffolds identified do not yet have a known cellular target, which is a limitation of the cell-based approach. Two of the chemical groups (Group II and IV) increased the interferon gene transcription profile, prolonged interferon activation of STAT1, and were also good co-adjuvants in a murine immunization model. Several of the other scaffolds (Groups I, III, V and VII) were demonstrated to be effective co-adjuvants, suggesting that the cell-based assay detected agents with multiple regulatory mechanisms.

The key supporting evidence for this HTS in selecting potential co-adjuvants was the final assessment by co-immunization with an antigen and LPS in an in vivo test model. A single in vitro assay cannot replicate the complex system of multiple cell interactions and physical relocation that an antigen presenting cell must undergo to initiate an effective and lasting immune response. Not surprisingly in a complex biological system, the 16 h relative ratio values in the FRET assay did not always directly correlate with the immunoglobulin levels measured after co-immunization. Further work to identify the targets and the cellular mechanisms is warranted and ongoing. In summary, we discovered multiple chemical scaffolds that resulted in prolonged ISRE activation in conjunction with IFN-α stimulation.

ABBREVIATIONS

APC

antigen presenting cells

bla

beta-lactamase

DMSO

dimethyl sulfoxide

ELISA

enzyme-linked immunosorbent assay

FDA

Food and Drug Administration

FRET

Förster resonance energy transfer

HTS

high-throughput screening

Ig

immunoglobulin

IFN

interferon

ISRE

Interferon-stimulated response element

LPS

lipopolysaccharide

MTT

(3-[4,5-dimethylthiazol-2-yl]-2,5-dipheyl tetrazolium bromide)

NF-κB

nuclear factor kappa B

OD

optical density

OVA

ovalbumin

APC

antigen presenting cell

h

hour

SDS

sodium dodecyl sulfate

STAT

signal transducer and activator of transcription

ISGF3

Interferon-stimulated gene factor 3

U

Units

USP18

Ubiquitin specific peptidase 18

veh

vehicle

WT

wild type