Engineering CREB-activated promoters for adenosine-induced gene expression

Abstract

Accumulation of the ribonucleoside, adenosine, triggers a CREB-mediated signaling pathway to suppress the function of immune cells in tumors. Here, we describe a collection of CREB-activated promoters that allow for strong and tunable adenosine-induced gene expression in human cells. By optimizing number of CREB transcription factor binding sites and altering the core promoter region of CREB-based hybrid promoters, we created synthetic constructs that drive gene expression to higher levels than strong, endogenous mammalian promoters in the presence of adenosine. These synthetic promoters are induced up to 47-fold by adenosine, with minimal expression in their ‘off’ state. We further determine that our CREB-based promoters are activated by other compounds that act as signaling analogs, and that combinatorial addition of adenosine and these compounds has a synergistic impact on gene expression. Surprisingly, we also detail how background adenosine degradation caused by the common cell culture media additive, FBS, confounds experiments designed to determine adenosine dose-responsiveness. We show that only after long-term heat deactivation of FBS can our synthetic promoters enable gene expression induction at physiologically relevant levels of adenosine. Finally, we demonstrate that the strength of a CREB-based promoter is enhanced by incorporating other transcription factor binding sites.

Graphical Abstract

The ribonucleoside adenosine is an immunosuppressive metabolite that accumulates within solid tumor environments. In this work, the authors engineer synthetic promoters that are activated by adenosine. When driving expression of a pro-inflammatory cytokine, the promoters can outperform constitutive expression of the cytokine interleukin 2, which can be toxic when delivered at high doses systemically. These promoters form the basis of a cell therapy engineering strategy for tumor-activated cell-mediated delivery.

Introduction

The ability to precisely activate protein expression in a tissue or environment-specific manner in human cells could greatly improve therapies for disease [1]. To this end, synthetic promoters that are activated by specific external states or exogenous inducers have recently been developed to control gene expression in human cells [2–4]. These inducible synthetic promoters typically exist in an “OFF” state – necessitating a specific input to activate and turn “ON” gene expression. For example, synthetic promoters constructed from tandem repeats of heat-shock transcription factor binding sites can be activated by near infrared radiation (NIR)-controllable increases in temperature. These heat-inducible promoters have been utilized to turn on the expression of chimeric antigen receptors and other T cell proteins with precise spatial control and specifically in solid tumors [5]. Synthetic promoters that have a native “ON” state and an inducible “OFF” can also be useful. In fact, induced “ON” and induced “OFF” synthetic promoters can be used in the same cell, for instance, to repress transcription with light and activate gene expression with doxycycline [6].

There are often limitations inherent to these synthetic promoters, as they typically require user-administered inducers (e.g., heat, light, or exogenous chemicals) [4,6]. The difficulty of delivering heat or light through skin to hard-to-reach physiological locations limits their practical and clinical application, and administration of exogenous chemical inducers can be hindered when the drug diffuses elsewhere in the body, compromising effective transcriptional control and potentially broaching safety concerns [7–10]. Synthetic promoters that can respond to local environmental cues to control gene expression in human cells may be more desirable, particularly if the local environmental cue were associated with a disease state.

In this vein, the ribonucleoside adenosine accumulates to levels approximately 100 to 1000-fold in solid tumors than it does in healthy tissue (i.e., nanomolar → micromolar levels), and tumor-accumulated adenosine has been shown to directly stimulate receptor-based signaling in several immune cells, particularly in T cells, inhibiting their inflammatory or cytotoxic capabilities [11]. When adenosine binds to the cell-surface adenosine receptors A_2AR or A_2BR, their coupled G_s proteins activate adenylyl cyclases that produce 3’5’-cyclic adenosine monophosphate (cAMP), the ligand for protein kinase A (PKA) [12]. Activated PKA phosphorylates the cAMP response element binding protein (CREB) at Serine 133 [13]. pCREB is a transcription factor, binding the palindromic cAMP response element motif (CRE), 5’-TGACGTCA-3’, and recruiting other co-activators to upstream enhancers/promoters to induce transcription [14,15]. In summation, through the cAMP/PKA signaling pathway, extracellular adenosine can control cellular gene expression (Figure 1A and 1B). Thus, by incorporating CREs within a synthetic promoter, it may be possible to engineer cells that exhibit tumor-activated gene expression through adenosine induction.

Figure 1: Creating synthetic CREB-responsive promoters.

(A) Depiction of adenosine signaling. Adenosine (red spheres) binds adenosine receptors A_2AR/A_2BR, which mobilizes the associated G protein (green) to activate adenylyl cyclases (orange receptor) and convert ATP into cAMP. Alternatively, forskolin (orange spheres) can activate adenylyl cyclases directly. cAMP binds PKA to phosphorylate CREB, which binds the palindromic DNA motif “TGACGTCA,” activating gene expression. (B) Promoter design and screening schematic. CREs (highlighted yellow) were cloned in 3x replicates, flanked with a guanine “G” (underlined) with six interspersed filler nucleotides (n). 3x CREs (gray squares) were then placed in 1–6 replicates upstream of a core promoter (blue arrow). Promoter activity was quantified with either Gaussian Luciferase (GLuc) or green fluorescent protein (eGFP). (C, D) HEK293T cells were reverse transfected with the indicated constructs (x-axis) in 96 well plates. 48hrs after transfection, the cell media was changed to media with vehicle (DMSO, light blue bars) or 20μM forskolin (Fsk, dark blue bars). Eight hours later, the media was sampled and tested for GLuc activity (by RLU). Bars represent average of n=3 experimental replicates and error bars represent standard error (SEM). ** Denotes P < 0.01 compared to all other samples by ANOVA Tukey Test. (E, F) Promoter inductions by flow cytometry. HEK293T cells were reverse transfected with the indicated constructs (x-axis) in 96 well plates. 48hrs after transfection, cell media was changed to untreated media (light blue bars), or media supplemented with 0.750mM adenosine (ADO, dark blue bars). Eight hours later, cells were trypsinized and resuspended in FACS buffer for flow cytometry. Y-axis represents the eGFP median fluorescent intensity of forward scatter (FSC) singlets. Bars represent the average of n=3 experimental replicates and error bars represent standard error (SEM). (G) Promoter dose responsiveness to adenosine. HEK293T cells were reverse transfected in 96 well plates with the constructs indicated in the legend and then cultured for 48 hours. Media was then changed to add different adenosine concentrations, sampled after 8 hours, and tested for GLuc activity (RLU). ** represents P<0.01 by ANOVA Tukey test of 12x-CRE_YB compared to all other samples at 1mM. Each point represents the average of n=3 experimental replicates and error bars are SEM.

PKA/pCREB signaling and transcriptional activation can also be mediated by other metabolites, specifically by prostaglandin E2 (PGE2), epinephrine, and norepinephrine (Norep) [11]. When PGE2 binds EP2 or EP4 receptors or epinephrine or Norep bind beta adrenergic receptors, the receptors’ coupled G_s proteins activate adenylyl cyclases to generate cAMP [16,17]. PGE2, epinephrine, and Norep have also been demonstrated to suppress immune cell antitumor immune function.

Here, we engineer and characterize synthetic promoters that exhibit CREB/adenosine-induced log-fold increases in gene expression. Using a small-molecule inducer of adenylate cyclase, we show that the optimal number of tandem-repeated CRE sequences in a synthetic promoter varies slightly by the core promoter region they induce (minCMV and YB). We then demonstrate that these promoters afford up to 32-fold induction of reporter gene expression in response to high levels of adenosine addition, with reporter protein levels surpassing those seen with strong, endogenous human and murine promoters. After deducing that common media formulations contain high levels of an adenosine degrading enzyme, we show that these promoters are fully induced by physiologically relevant adenosine levels. We further demonstrate that the promoters are activated both by other AC-stimulating compounds in a dose dependent manner with favorable on/off kinetics, and then show how addition of additional transcription factor binding sites, apart from CRE sequences, can further heighten adenosine-induced gene expression without severely impacting promoter off-state.

Materials and Methods

Cell Lines

HEK293T cells were maintained in GlutaMAX supplemented with 10% FBS from Gibco. Jurkat, MAD109, and 4T1 cells were cultured in RPMI-1640 (Gibco) supplemented with 10% FBS. All cell lines used in this study were tested for mycoplasma and verified negative.

Plasmid and promoter construction

GLuc-P2A-eGFP sequences were a generous gift from Dr. Gabe Kwong. Base, golden gate compatible pcDNA3 vectors containing the minCMV and YB core promoters with Esp3I cut sites and the GLuc-P2A-eGFP gene were constructed using Gibson Assembly. From there, either 6xCRE promoter repeats were ordered from Twist Biosciences as double stranded gene fragments bearing golden gate assembly cut sites/overhangs, or 3x CRE promoter repeats (and their respective reverse complements) bearing the proper golden gate assembly overhangs were ordered as single stranded DNA fragments from Eurofins Genomics and annealed prior to Golden Gate Assembly. A complete list of promoter DNA sequences can be found within Supplementary Table 1. The COX2 promoter was PCR amplified from HEK293T genomic DNA using primers jrc305/306 included within Supplementary Table 2. For stable cell line experiments, the promoter-GLuc-P2A-eGFP genes were PCR amplified and inserted into pLEGO-B backbones that were digested with XbaI and XhoI. The human interleukin 2 gene coding region was ordered from Twist Biosciences and inserted via Gibson Assembly.

Using standard electroporation, lentiviral backbones were transformed into NEB^® Stable E. coli and all pcDNA backbones were transformed into NEB^® 10β E. coli [18]. All plasmids were sequence confirmed via Sanger sequencing prior to use.

FBS adenosine degradation experiments

Heat deactivations of FBS were carried out in a water bath set to 56°C for the indicated time. For adenosine degradation analysis within growth media, media was supplemented with 1mM of adenosine and the indicated FBS and cultured in a 96 well U-bottom plate. Media was sampled at the indicated time points and read on a spectrophotometer for absorbance at 260nm. The percentage of adenosine remaining was calculated by normalizing to the initial reading.

Transient promoter induction experiments

For transient transfection promoter screening, HEK293T cells were transfected by reverse transfection with the indicated promoter construct (pcDNA3 backbones). Briefly, 400 ng of plasmid DNA was combined with 0.6μL of TransIT-LT1 (Mirus) and 18μL of OptiMEM (Gibco). After incubation for 10 minutes, 50,000 cells were added to each well of a 96 well TC plate. After cell attachment, the cells were treated as indicated (e.g., adenosine, forskolin, etc.) for 8 hours and supernatants were harvested for analyzed for GLuc expression by the Pierce Gaussian Luciferase kit (ThermoFisher) according to the manufacturer’s instructions or the cells were trypsinized for analysis by flow cytometry.

Lentivirus production

For lentiviral preparation, 4 × 10⁶ HEK293T cells were seeded in a 10cm dish. The next day, the cells were transfected with 5μg of transfer vector, 1μg of pMD2.G (Addgene #12259), and 4μg of psPAX2 (Addgene #12260) with TransIT-LT1 according to the manufacturer’s instructions. Two days later, supernatants were harvested, spun down at 1000xg for 5 minutes, passed through a 0.45μm PES filters (VWR), and then finally precipitated with PEG-it^™ viral precipitation solution (SBI, LV825A-1) according to the manufacturer’s instructions. A day later, precipitated virus was resuspended in ice cold PBS and stored at −80°C until use.

Promoter kinetics experiment

For experiments using stably engineered cell lines, cells were transduced with lentivirus as described and sorted for tagBFP+ expression prior to use on a BD FACS Melody. Promoter induction experiments and sampling with these sorted cell lines followed the same protocol as transient transfection experiments without the transfection step. Overall, cells were induced with the indicated compound and at the specified time point, harvested for analysis by flow cytometry to ascertain eGFP expression. For inductions containing adenosine, media was replenished every 12 hours to account for cellular consumption. To demonstrate promoter shut-off potential, 12 hours before analysis, media was replenished with 250μM of adenosine and 3μM of recombinant human adenosine deaminase (ADA1).

Enzyme Linked Immunosorbent Assay (ELISA) Procedure

Maxisorp plates (ThermoFisher) were coated with 5μg/mL anti-human IL-2 (R&D systems, #MAB202) antibody diluted in PBS −/− and incubated overnight at 4°C. The plate was then washed 4x with PBS −/− and blocked with 400μL of 3% BSA in PBS −/− for 1 hour at room temperature. After blocking, the plate was washed three times with PBST (PBS−/− with 0.05% Tween 20). 200 fold diluted supernatant samples were then applied to the plate and incubated for 1 hour at room temperature. After this incubation, the plate was washed three times with PBST and incubated at room temperature with 1000-fold diluted biotinylated polyclonal IL-2 antibody (ThermoFisher #13-7028-81). After washing three times with PBST, Streptavidin-HRP (R&D Systems, #DY998) was added to each well according to the manufacturer’s instruction. The plate was then washed five times with PBST, and 1-step TMB-ELISA Substrate Solution (ThermoFisher #34028) was applied to each well. After a 15-minute incubation, the reaction was terminated with 0.3M sulfuric acid and absorbance was read at 450nm with a Synergy HTX microplate reader.

Results

Synthetic CREB-promoter design and optimization of CRE sequence number for YB core promoter

To allow for tumor-specific activation, an ideal pCREB/adenosine-inducible promoter would have tight regulation (low expression when ‘OFF’) and mediate robust gene expression when activated (high strength when ‘ON’, resulting in several fold-induction). We used prior studies of endogenous human CREB-activated promoters, in addition to a preliminary experiment (Supplementary Figure 1), to develop design criteria for an optimized synthetic adenosine-induced promoter [14,15,19–22]. These design criteria focused on (1) CRE sequence element placement in relation to the gene of interest, (2) the optimal nucleotide sequence to be used for the CRE sequence element itself, and (3) the number of CRE sequence elements required to robustly induce gene expression.

For instance, limiting the total promoter size could be a key design criterion, as CRE sequences more than 200 base pairs from the transcriptional start site (TSS) may not function. Notably, two-thirds of native CREs are located within −50 to −150 base pairs of TSSs and exhibit maximal activity when placed at most 100bp from the TATA box [14,19]. In addition to TSS proximity, the specific CRE and adjacent nucleotide sequences may also impact synthetic promoter function. The CRE sequence is palindromic, and it has been shown that both half-CRE motifs and the full palindromic motif are functionally bound by CREB. However, the full palindromic motif is bound stronger and mediates superior activation than the half-site, suggesting that the full ‘TGACGTCA’ CRE sequence element should be included in synthetic promoters [20–22]. In certain promoters with multiple CRE sequences in tandem, including the somatostatin, phosphoenolpyruvate, and tyrosine aminotransferase promoters, there is a distinct nucleotide sequence (5’-GCCCC-3’) between neighboring CRE sequence elements [22]. Saturation mutagenesis has shown that the only nucleotide of this sequence that affects CREB responsiveness is the 5’guanine nucleotide, which is adjacent to the 3’ end of the CRE sequence element relative to the direction of transcription [19]. The impact of multiple CRE sequence elements in engineered promoters has yet to be thoroughly investigated. However, we performed initial experiments that showed that synthetic promoters containing less than three CRE sequence elements did not induce reporter gene expression in response to an adenylate cyclase-stimulating compound, forskolin, that is known to strongly activate CREB-responsive promoters (Supplementary Figure 1) [23].

Based on these design criteria, we constructed synthetic promoters harboring between three and eighteen full palindromic CRE elements, with six nucleotides between each element, and a G nucleotide placed at the 3’ end of each element (Figure 1B) [19]. We inserted replicates of 3X CRE tandem repeats upstream of the YB core promoter that was developed by Yaakov Benson and colleagues, which promotes minimal background transcription [3,5,24]. As a positive control for adenosine-mediated CREB activation, we used the native cyclooxygenase-2 (COX2) promoter [25–27]. To measure inducible gene expression, all synthetic promoters were inserted on a pcDNA3 vector backbone upstream of a Gaussian luciferase (GLuc)-P2A-eGFP construct that allows measurement of both bulk culture and individual cell gene expression levels. Promisingly, upon reverse transfection into HEK293T cells, all the synthetic promoter constructs enabled inducible GLuc expression after an eight-hour incubation with forskolin (Figure 1C). We found that the 12X-CRE_YB promoter mediated the highest expression level, and nearly the highest fold-induction (6.2x fold-induction for 12x-CRE_YB vs. 6.5x for 9x-CRE_YB), in response to forskolin, while the YB core promoter without any response elements did not induce gene expression, showing that the CRE sequence elements are mediating increased gene expression. Promisingly, the 12x-CRE_YB promoter afforded twice as much protein expression as the native COX2 promoter, which has previously been characterized as CREB-responsive, with much less background activity (Figure 1C) [25–27]. The COX2 promoter exhibited only a modest ~1.25X increase in expression upon treatment with forskolin. Therefore, engineered synthetic promoters appear to be better able to mediate controllable induction of gene expression in response to CREB-activation than certain native promoters. In sum, we designed functional CREB-activated promoters that outperform a native promoter in terms of strength, background expression, and fold-induction.

Tandem CREs activate gene expression with an alternative core promoter

The CMV promoter, originally characterized by Mark Stinski and colleagues and derived from human cytomegalovirus, uses upstream regulatory sequences to mediate strong gene expression [28]. The truncated minimal core CMV promoter (minCMV), slightly larger than the YB core, has been shown to exhibit greater strength when incorporated into synthetic promoters, though often with higher background activity [3]. To determine if tandem repeats of CRE sequences proximal to the minCMV core promoter could also enable adenosine/CREB inducible gene expression, we created promoters with 6, 9, or 12 CRE sequence elements upstream of the minCMV core, using similar design constraints as above. We hypothesized that synthetic promoters containing more than 12 CRE sequence elements would have reduced function, as distal elements would be too far from the TATA box (>170bp) for effective function [14,19]. We found that the 9x-CRE_minCMV promoter had the highest gene expression level and fold-induction when induced with 20μM forskolin, and as expected, the 12x-CRE_minCMV synthetic promoter with more distal CRE sequence elements afforded lower expression (Figure 1D). Compared to 12x-CRE_YB, the 9x-CRE_minCMV promoter enables roughly double the level of protein expression, though with higher background expression. Still, fold-induction was comparable (5-fold compared to 6.5-fold) for both promoters. As observed with the YB core promoter without CRE elements, the minCMV core promoter drove minimal gene expression by itself and did not exhibit forskolin-induced gene activation responsiveness. Therefore, we demonstrated that tandem repeats of CRE sequence elements mediate inducible gene expression when proximal to other core promoters, and we created stronger CREB-activated promoters using the minCMV core (Figure 1C and 1D).

Synthetic CREB promoters are induced by adenosine

We next confirmed that our CRE sequence element-containing synthetic promoters could be induced by extracellular adenosine, testing for reporter protein expression after eight hours of induction. As expected, the minCMV and YB core promoters did not activate gene expression in response to adenosine stimulation and had low background expression level (Figure 1E–Figure 1G). Both minCMV and YB-based CRE-containing synthetic promoters induced significant eGFP expression in response to incubation with 0.750mM adenosine, with the 12x-CRE_YB promoter exhibiting 32-fold induction and the 9x-CRE_minCMV exhibiting 24-fold induction (Figure 1E & Figure 1F). We further determined that optimum adenosine level to add to media to induce expression from the 12x-CRE_YB promoter to be ~1mM, which resulted in a ~13-fold activation, higher than ~6-fold activation induced by forskolin (Figure 1G). The optimal adenosine concentration was unchanged by the amount of response elements in a YB-based promoter (e.g. 6x-CRE vs. 12x-CRE). The difference in fold-activation due to adenosine stimulation from the 12x-CRE_YB as measured by eGFP median fluorescent intensity (MFI) using flow cytometry and as measured in bulk supernatant via GLuc activity indicated more robust expression changes could be detected at the single cell level than in bulk. The COX2 native adenosine-responsive promoter again had high background expression and limited adenosine-induced expression. In sum, our synthetic CRE containing promoters are fully functional and strongly activated by adenosine, and greatly outperform an endogenous CREB-responsive promoter.

FBS in cell culture media degrades adenosine, altering promoter responsiveness

It has been shown that HEK293T cells likely express the lower affinity adenosine receptor, A_2BR (EC₅₀ = 23.5μM) and not the higher affinity A_2AR, (EC₅₀ = 0.73μM) [29–31]. Still, considering adenosine’s affinity for A_2BR, we expected to observe maximal promoter responsiveness at micromolar concentrations, instead of ~1mM (Figure 1G). Therefore, we next considered the possibility that a component in our media could be degrading supplemented adenosine, increasing required adenosine levels for maximal promoter activation. In this vein, adenosine deaminase activity, mediated by the bovine adenosine deaminase I, can be detected in cattle serum, so we hypothesized that fetal bovine serum (FBS) might also be able to degrade adenosine [32,33].

To test our hypothesis, we supplemented either (1) complete cell culture media, i.e., with 10% FBS or (2) cell culture media lacking FBS, with 1mM adenosine and analyzed how adenosine concentration changed over time in each media type (Figure 2A). We found that cell culture media containing 10% FBS degraded more than half of the adenosine in just 12 hours and nearly all adenosine withing 24 hours, while cell culture media without FBS did not degrade adenosine over the course of the full 24 hours. We further showed that heat deactivating the FBS could eliminate its ability to mediate adenosine degradation, but surprisingly, neither a 30-minute nor a six-hour, high temperature (56°C) incubation was adequate to eliminate adenosine degradation. Instead, a 24-hour heat-deactivation/incubation was required to prevent FBS-mediated adenosine degradation (Figure 2A). After discerning that our prior analyses of promoter induction may have been impacted by concurrent FBS-mediated adenosine depletion, we retested the adenosine-responsiveness of the 12x-CRE_YB and 9x-CRE_minCMV synthetic promoters in media that lacked adenosine degradation ability. Both promoters were strongly induced by much lower adenosine levels, with near full induction of 12x-CRE_YB at <50μM and of 9x-CRE_minCMV at <100μM when measuring both GLuc and eGFP (Figure 2B–D). Therefore, HEK293T cells expressing our synthetic promoters are responsive to physiologically relevant, micromolar adenosine concentrations, but only when background adenosine degradation is eliminated from cell culture media.

Figure 2: FBS-mediated adenosine degradation affects cellular responsiveness to adenosine.

(A) Media background adenosine degradation. Standard HEK293T cell media (DMEM) was supplemented with 1mM adenosine and (i) no FBS (purple), (ii) 10% FBS (white diamonds), (iii) 10% FBS with 30-minute heat deactivation (HD) (light blue circles), (iv) 10% FBS with 6hr HD (medium blue triangles), or (v) 10% FBS with 24hr HD (dark blue squares). Media was sampled at the indicated time points, diluted in PBS, and read by plate reader for absorbance at 265nm. Each point represents n=3 experimental replicates and error bars represent standard error (SEM). * = P<0.05, ** = P<0.01, and **** = P<0.001 by ANOVA Tukey Test of both no FBS media and 24hr heat deactivated FBS media compared to all other samples. (B) Promoter dose responsiveness to adenosine in fully heat deactivated media. HEK293T cells were reverse transfected with the indicated constructs (9x-CRE_minCMV = green, 12x-CRE_YB = dark blue, minCMV = light green, and YB = light blue) in 96 well plates. 48 hours later, media was changed to media with the adenosine concentrations indicated on the x-axis. (C) 12xYB promoter responsiveness to fully heat deactivated (HD) or regular media. The 12x-CRE_YB promoter was transfected into HEK293T cells in 96 well plates. 48 hours later, adenosine was titered in either HD (blue squares) or regular media (light blue circles). 8 hours later, the cells were collected and analyzed by flow cytometry to eGFP median fluorescent intensity (MFI). Each point represents the mean of n=3 biological replicates and error bars represent SEM. (D) 9x-CRE_minCMV promoter responsiveness compared in HD or regular media. Experiment performed as in (C). For both experiments, ** = P<0.01 by ANOVA Tukey Test compared to non-heat deactivated control at the indicated point.

Combinatorial additions of cAMP-stimulating compounds amplify synthetic promoter induction

Because compounds like prostaglandin E₂ (PGE2) and norepinephrine (Norep) also stimulate the cAMP/PKA pathway, our engineered adenosine/CREB-responsive promoters could be activated by these compounds [11]. As a first pass, we tested the ability of the 12x-CRE_YB and 9x-CRE_minCMV promoters to be induced by PGE2 and Norep independently. Promisingly, we saw that both PGE2 and Norep could stimulate the synthetic promoters, with PGE2 having a 3.6-fold maximal induction at 50μM for 9x-CRE_minCMV and Norep having a 1.8-fold maximal induction at 25μM for the 12x-CRE_YB promoter (Figures 3A & 3B). In a direct comparison, fold induction of the 9x-CRE_minCMV promoter induced by a PGE2 and Norep was lower than that of adenosine (Figure 3C), but the ability of other cAMP-producing compound to induce expression of our synthetic promoters might imply that they can combine to enhance overall promoter strength, or that their combination might reduce the level of any individual component needed for maximal induction. In this vein, adenosine, PGE2, and Norep can all promote the synthesis of each other, meaning that they are rarely found in complete isolation [11].

Figure 3: Promoter responsiveness to other cAMP stimulating compounds.

(A, B) Promoter responsiveness to prostaglandin E2 (PGE2) or norepinephrine (Norep). HEK293T cells were reverse transfected with either the 9x-CRE_minCMV (dark blue circles) or 12xYB (light blue circles) promoters. 48 hours later, media was changed to fresh media with the indicated PGE2 or Norep concentration (x-axis) and analyzed by GLuc 8 hours later. Fold induction is relative to the untreated control. Each point is the average of n=3 experimental replicates and error bars are standard error (SEM). (C) Fold induction comparisons between adenosine (ADO), PGE2 and Norep. HEK293T cells were reverse transfected with the 9x-CRE_minCMV promoter. Media was changed 48 hours later to the indicated media, and then read for GLuc activity. Fold induction is quantified by untreated controls. Each bar is the average of n=3 experimental replicates and error bars = SEM. ** = P<0.01 by ANOVA Tukey Test. (D) Promoter responsiveness to different combinations of each cAMP-stimulating compound. HEK293T cells were transfected with the 9x-CRE_minCMV promoter and then cultured for 48 hours. Then, the media was changed to the supplemented media indicated on the x-axis and, after 8 hours, read for GLuc activity (by RLU, y-axis).

Therefore, we tested how well combinatorial additions of each cAMP stimulating compound could stimulate gene expression from our synthetic promoters. We tested the triple combination of the maximal, half maximal, or quarter maximal concentrations of adenosine, Norep, or PGE2 and compared the resulting promoter induction to that mediated by adenosine or forskolin only (Figure 3D), where for instance, the quarter maximal concentration is the inducer level that stimulated ~25% of the maximal gene expression attained by the inducer at any concentration. Maximal additions of each compound in combination exhibited the highest promoter induction, and perhaps most notably, quarter-maximal additions of each compound (e.g., 25μM compared to 250μM for adenosine) outperformed both maximal adenosine induction and forskolin induction (Figure 3D). Taken together, these results suggest that combinations of adenosine, PGE2, and norepinephrine can amplify CREB-responsive promoter activation and act in a synergistic manner towards enabling high gene expression levels at low inducer concentration.

Synthetic promoters display favorable on and off kinetics and enable higher gene expression than a strong, endogenous promoter

We wanted to further compare our synthetic promoters to a strong, endogenous promoter, and to determine their kinetics of induction. Therefore, we stably transduced HEK293T cells with constructs encoding the synthetic 9x-CRE_minCMV promoter, the minCMV core promoter, or the constitutively active endogenous murine PGK promoter, in which each promoter drove expression of the eGFP reporter. As a marker for transduction, a SFFV promoter driving expression of tagBFP was also included. After sorting BFP positive cells, we compared our synthetic 9x-CRE_minCMV promoter, in terms of maximum protein expression, to the PGK promoter. Promisingly, forskolin-induced activation of the 9x-CRE_minCMV yielded nearly 2.5-fold higher eGFP expression compared to PGK promoter, confirming that our synthetic promoters mediate high levels of gene expression that can surpass commonly used constitutive native promoter (Figure 4A). The PGK promoter generated slightly higher eGFP expression when incubated with forskolin but was slightly repressed by adenosine-only or adenosine/PGE2/Norep stimulation, while the minCMV core promoter was not impacted by an inducer (Figure 4A and Supplementary Figure 2).

Figure 4: 9x-CRE_minCMV Promoter outperforms endogenous promoters and has favorable on/off kinetics within engineered cell lines.

(A) The 9x-CRE_minCMV promoter outperforms a constitutively active endogenous promoter when in an induced state. 9x-CRE_minCMV and PGK viral constructs were transduced into HEK293T and sorted based on tagBFP. 5 × 10⁴ sorted cells were seeded in 96 well plates and treated with (dark blue) or without (light blue) 20μM forskolin (Fsk). After 24-hour treatment, the cells were analyzed by flow cytometry for eGFP median fluorescent intensity (MFI). Bars are the average of n=3 experimental replicates and error bars = standard error (SEM). ** = P<0.01 compared to all other groups by ANOVA Tukey Test (B) HEK293T cell lines engineered with the 9x-CRE_minCMV promoter were treated with either (i) 20μM Fsk, dark blue circles and line (ii) 250μM adenosine (ADO), medium blue squares and line every 12 hours or (iii) 25μM ADO, 4.17μM PGE2, and 6.125μM Norep every 12 hours (Combo), light blue triangles. After the indicated treatment times, the cells were analyzed by flow cytometry for eGFP MFI, normalized to untreated cells to achieve fold induction. The black arrow represents when media was changed to either adenosine + recombinant ADA1 or adenosine only media to observe promoter shut-off. Each point represents the average of 3 experimental replicates and error bars = SEM. ** denotes P<0.01 by ANOVA Tukey Test between ADA1 + ADO and ADO only treated cells. (C) HEK293T cell lines engineered to express human interleukin 2 (IL-2) from the indicated promoter on the x-axis were treated with the indicated compound. Cell lines were treated with 20μM Fsk, 250μM ADO every 12 hours, or left untreated (UTD). Supernatant was harvested for Fsk treated samples after 24 hours, while ADO and untreated samples were harvested after 48 hours. IL-2 concentration (ng/mL) was determined by ELISA, and ND = not detected within the lowest standard curve concentration. Bars represent the average of N=3 experimental replicates and error bars = SEM. ** denotes P<0.01 by ANOVA Tukey Test between PGK + ADO and 9x-CRE_minCMV + ADO groups. CREB-activated promoters are functional in other cells. (D-F) Each cell line was transduced with either the 9x-CRE_minCMV (dark blue) or 12x-CRE_YB promoter (light blue) and sorted for tagBFP. Each cell type was seeded at 5 × 10⁴ cells/well in a 96 well plate. For Jurkat cells, they were directly seeded in the titrated forskolin media, but for 4T1 and MAD109, the cell lines were allowed to attach before the media was changed to forskolin-containing media. After 24 hours, the media was sampled and analyzed for GLuc activity by RLU. Each point represents the average of n=3 experimental replicates and error bars = SEM.

Next, to test the speed of promoter induction, i.e., the time to achieve maximal eGFP expression for each promoter, we measured eGFP expression after 3, 6, 9, 12, 24, and 48 hours post inducer addition. Forskolin (20μM), adenosine (250μM), and the triple combination of quarter maximal PGE2 (4.17μM), Norep (6.125μM), and adenosine (25μM) were tested as inducers. In addition, we were interested in determining if inducer removal (or degradation) could result in a quenching of the promoter activity. To this end, after the 48-hour induction test, we added enough recombinant adenosine deaminase 1 to the adenosine-only induced cells to degrade the 250μM adenosine within minutes, and then tested for eGFP expression 12 hours later.

Regardless of stimulating compound, the 9x-CRE_minCMV promoter exhibited favorable on/off kinetics. Interestingly, the PGE2 + Norep + adenosine inducer combination reached maximal induction fastest, after only 12 hours, but had the lowest level of induced gene expression (Figure 4B). Forskolin resulted in the highest level of induced gene expression seen within 24 hours, though eGFP expression was substantially decreased by 48 hours, potentially due to overly persistently elevated intracellular cAMP, which can be counteracted by endogenous phosphodiesterases to maintain cAMP and cGMP homeostasis [34,35]. Using adenosine as the only inducer continuously increased gene expression of the course of 48 hours, reaching 47-fold induction, and addition of adenosine deaminase 1 resulted in a decrease in eGFP levels within 12 hours, suggesting that removal of adenosine could turn off our synthetic promoters (Figure 4B). Next, we wanted to test if we could use our promoters to drive expression of the immunomodulatory cytokine, IL-2, in response to adenosine. IL-2 administration, which can cause toxicity when delivered at high dosages systemically, has improved therapeutic value when delivered intratumorally [36]. We replaced the GLuc-P2A-eGFP gene cassette with the human IL-2 CDS and remade HEK293T cell lines. Upon 24-hour treatment with forskolin and 48-hour treatments of adenosine in our engineered cells, our (induced) 9x-CRE_minCMV promoter enabled more IL-2 production than the PGK promoter (Figure 4C).

In sum, our synthetic 9x-CRE_minCMV promoter exhibits rapid and strong induction of gene expression in response to a variety of adenosine associated stimulation, and its maximal expression level surpasses that of an often-used strong native promoter, pPGK. The 9x-CRE_minCMV also turns off when adenosine is eliminated, losing approximately 20% of eGFP signal in just 12 hours, despite the day long half-life of eGFP [37]. Finally, we demonstrated that we can repurpose an immunosuppressive signal like adenosine to produce an immunostimulatory protein that is conventionally suppressed by adenosine accumulation [38,39].

Confirming synthetic promoter function in additional cell lines

Finally, we wanted to confirm that our promoters retained function in cells lines apart from HEK293T. Therefore, we demonstrated that both the 12x-CRE_YB and 9x-CRE_minCMV promoters were strongly induced in both a second human cell line (Jurkat immortalized T cells), as well as two murine cells lines, the 4T1 mammary carcinoma and the Madison 109 lung cancer cell line (Figure 4D–Figure 4F) [40–42]. As before, the 12x-CRE_YB promoter generally afforded lower maximal gene expression levels but higher fold-induction, due to its lower off-state, than the 9x-CRE-minCMV promoter.

Enhancing the strength of CREB-activated synthetic promoters through rational introduction binding sites for CREB partners

We were also interested in why our 12x-CRE-YB promoter is weaker at full induction that the 9x-CRE_minCMV promoter, despite harboring more CRE elements in a useable proximity to its TATA box. Interestingly, a target of PKA other than CREB is the p65 subunit of NF-𝜅B (RELA), which has been shown to recruit a CREB interaction partner, CREB binding protein (CBP) [43]. This interaction can functionally increase gene expression induced by CREB, likely through its lysine acetyltransferase domain, which has been hypothesized to increase the strength of the full CMV promoter (Figure 5A) [44,45]. Indeed, we observed moderate (~1.5x) adenosine-mediated induction of the CMV promoter (Supplementary Figure 3) suggesting incorporating RELA response elements could further increase the strength of our CREB-activated promoters.

Figure 5: Improving the strength of the 12x-CRE_YB promoter through incorporation of RELA sites.

(A) Image depiction of the expected mechanism of RELA promoter strength augmentation. NF-𝜅B (RELA, purple) is recruited to proximal DNA binding sequences (dark blue), allowing recruitment of CREB binding protein (CREB-BP, light gray), which, through its lysine acetyltransferase (KAT) domain, can acetylate CREB dimers at palindromic CRE sites (light blue, dark gray) to improve their transcriptional activity. (B) Schematic of promoter designs. Three RELA sites (dark blue) were interspersed between 6xCREs (dark gray), yielding the 3x-RELA-12x-CRE_YB promoter. RELA sites were removed from 3’ to 5’ by one to form 2x and 1x sequences, respectively. (C) Comparison of RELA-supplemented 12x-CRE_YB promoters. HEK293T cells were transfected with the indicated promoter construct and titrated at different adenosine concentrations for eight hours. After which, the media was sampled for GLuc activity. Each point represents the average of n=3 experimental replicates and error bars are SEM.

Therefore, we interspersed the same consensus RELA sites (GGGACTTTCC) found within the CMV on the 5’ end, 3’end, and in the middle of the two 6xCREs in the 12xYB promoter (Figure 5B) and tested the resulting promoter’s adenosine responsiveness. We found that the 3x-RELA-12x-CRE_YB promoter was between 1.24x-1.35x stronger than the 12xYB alone at each tested adenosine concentration without increased gene expression in the promoter’s off state - 0μM adenosine added (Figure 5C). While the 3x-RELA-12x-CRE_YB was still not as strong as the 9x-CRE_minCMV, it did still show superior fold-induction due to its lower OFF state. The success in engineering RELA-supplemented promoters into the 12x-CRE_YB promoter motivated us to see if these elements could improve the 9x-CRE_minCMV promoter as well, but their incorporation surprisingly dampened the promoter’s sensitivity to adenosine (Supplementary Figure 4).

Discussion

We have engineered genetically compact and effective CREB-activated promoters that are capable of up to 47-fold gene expression induction, resulting in higher reporter expression than either constitutive promoters or other endogenous adenosine-responsive promoters.

We revealed key design considerations for future engineering efforts. For example, synthetic promoters harboring more than twelve CRE sequence elements for the YB core promoter had reduced expression, possibly because the additional CRE sequences are too far away from the core promoter’s TATA box. The additional three CRE elements within 15x-CRE_YB are at least 171 nucleotides from the YB core promoter’s TATA box, while the typical distance between functional CRE elements and an endogenous promoter’s TATA box is ~100 nucleotides [14,19]. As pCREB or other transcriptional co-activators may have a resource budget, pCREB binding the distal (and less effective) CRE sequence elements could reduce availability of pCREB proteins to bind the proximal and more effective CRE elements, dampening promoter activity.

Notably, these engineered promoters were highly induced by adenosine, a tumor-accumulating metabolite, as well as a combination of adenosine and other immunosuppressive molecules. We further showed that removal of adenosine could reduce gene expression within 12 hours, and that our promoters were functional in several mammalian cell types. In future work, we hope to demonstrate that adenosine-induced gene expression achieved by our promoters in other cell types, such as primary T cells, to be significantly more responsive, particularly if they express the A_2AR rather than the A_2BR, considering their differing affinities for adenosine (EC₅₀ = 0.73μM vs. EC₅₀ = 23.5μM) [29].

Importantly, we have also shown that background adenosine degradation from latent adenosine deaminase activity prevalent in FBS (tested across several lots) must be considered for in vitro studies involving adenosine. FBS-mediate degradation of adenosine may have been a confounding factor in prior efforts studying immunosuppression mediated by adenosine as well as adenosine dose-responsiveness [46–49]. Typical heat-deactivation protocols performed by FBS suppliers (30 minutes at 56°C) do not eliminate this adenosine degradation activity, potentially due to the higher than 56°C melting temperature of mammalian adenosine deaminase enzymes [50].

Adenosine and CREB-signaling are highly implicated in immune suppression of T cells in tumor environments and have been shown to limit T cell function via A_2AR [51,52]. We have shown that our synthetic promoters are strongly induced at physiologically relevant levels of adenosine, i.e., at levels of adenosine found in solid tumors, and that they have a low off-state expression level, i.e., without exposure to adenosine. We have also shown that it is possible to utilize an adenosine-activated synthetic promoter to drive cell-based, potentially tumor-localized expression of agents (e.g., IL-2) that have antitumor activity but are toxic when present systemically [53,54]. In this vein, our 9x-CRE_minCMV promoter is genetically compact, making it attractive to use in cell therapy applications. The full suites of YB- and minCMV-based promoters also fully sample a broad range of expression, which could also enable tunable, CREB-activated protein production. Still, we did explore preliminary strategies for improving our synthetic promoters, demonstrating that incorporation of RELA response elements could moderately improve the strength and fold-induction of the 12x-CRE_YB promoter. Strategies that incorporate other heterologous response elements (e.g., dioxin responsive element or hypoxic responsive element) that allow for response to other mechanisms of immune suppression tumors may also be promising routes to expand the function and utility of these synthetic promoters [55–57]. In sum, we engineered effective adenosine-activated promoters with high induction strengths that are activated in response to tumoral environmental conditions, as well as put forward considerations for testing adenosine-mediated cellular responses in vitro with FBS.

List of abbreviations

ADO

adenosine

cAMP

3’5’-cyclic adenosine monophosphate

CBP

CREB binding protein

CRE

cAMP response element motif

CREB

cAMP response element binding protein

COX2

cyclooxygenase-2

FBS

fetal bovine serum

MFI

median fluorescent intensity

Norep

norepinephrine

PGE2

prostaglandin E2

PKA

protein kinase A

RELA

p65 subunit of NK𝜅B

Data Availability Statement

Data will be made available by the authors upon reasonable request.