iGIST - a kinetic bioassay for pertussis toxin based on its effect on inhibitory GPCR signaling

Detection of pertussis toxin (PTX) activity is instrumental for the development and manufacturing of pertussis vaccines. These quality and safety measures require annually thousands of mice. Here, we describe iGIST (Interference in Gαi-mediated Signal Transduction) - an animal-free kinetic bioassay for detection of PTX by measuring its effect on inhibitory G protein-coupled receptor (GPCR) signaling. PTX ADP-ribosylates inhibitory α-subunits of the heterotrimeric G proteins, thereby perturbing the inhibitory GPCR signaling. iGIST is based on HEK293 cells co-expressing a somatostatin receptor 2 (SSTR2), which is an inhibitory GPCR controllable by a high affinity agonist octreotide, and a luminescent 3’5’-cyclic adenosine monophosphate (cAMP) probe. iGIST has a low sensitivity threshold in picogram/ml range of PTX, surpassing by 100-fold in a parallel analysis the currently used in vitro end-point technique to detect PTX, the cluster formation assay (CFA) in Chinese hamster ovary cells. iGIST also detects PTX in complex samples, i.e. a commercial PTX- toxoid containing pertussis vaccine that was spiked with an active PTX. iGIST has an objective digital readout and is observer-independent, offering prospects for automation. iGIST emerges as a promising animal-free alternative to detect PTX activity in the development and manufacturing of pertussis vaccines. iGIST is also expected to facilitate basic PTX research, including identification and characterization of novel compounds interfering with PTX.


4
The most widely debated animal-free alternative to HIST builds on the early findings by Hewlett et al., who observed phenotypic alterations, described as cell rounding and cell cluster formation, in Chinese Hamster Ovary (CHO) cells exposed to PTX 19 (Supplementary Videos 1-2). The resulting test, designated as a cluster formation assay (CFA), is based on visual grading of the cell clustering in CHO cell monolayers upon PTX treatment and can detect ng/ml levels of PTX 17 . However, the CFA is an observer-dependent end-point test, suffering from subjectivity bias and considerable interassay variability 17 . Also, the molecular basis of the PTX-evoked clustering in CHO cells, similar to the mechanism of histamine hypersensitivity in HIST, remain poorly understood. Despite these limitations, the European Pharmacopoeia Commission has decided that CFA can be used instead of HIST for safety assurance of the currently marketed pertussis ACVs 17 , based largely on work of Isbrucker et al. 20 , effective as of January 2020. Recently, Biological Reference Preparation batch 1 (BRP1) of PTX was introduced in order to control the inter-assay variability of CFA 21 .
Improved alternatives to CFA, based on the mechanistic understanding of PTX cellular effects, have been actively sought for 17 . Available biochemical assays for PTX measure either the PtxS1-catalyzed ADP-ribosylation of a C-terminal peptide of Gαi with HPLC 22 or binding of the pentameric PtxS2-S5 oligomer to carbohydrate structures with ELISA 23 . Both the assays have objective readouts, but capture only distinct PTX activities under artificial cell-free in vitro conditions. DNA microarrays have been utilized to identify PTX-induced gene expression signatures either in rat tissues 24,25 or in in vitro cultured human cells 26 . Practical applications have not yet emerged from these studies.
Hoonakker et al. exposed rat vascular smooth muscle cells (A10 cells) to PTX and determined the amount of cAMP in cell lysates with an end point ELISA 27 . PTX did not increase the amount of cAMP when incubated alone with the cells, but it potentiated isoproterenol-induced elevation of cAMP 27 . Isoproterenol binds to β-adrenergic receptors 28 , which leads to activation of Gαs and thereby to subsequent stimulation of the cAMP-producing ACs. In an extension of their work, Hoonakker et al. detected PTX effects in A10 and CHO cells with a cAMP response element (CRE)driven luciferase reporter 29 . In agreement with their earlier cAMP ELISA study 27 , PTX did not increase the CRE-reporter activity by itself, but it did enhance cAMP responses to isoproterenol or forskolin (FSK) 29 . FSK activates ACs by intercalating the C1 and C2 subunits of ACs into the catalytically active cAMP-producing form 30 . Although the detailed molecular basis of the CREreporter assay was not reported, the PTX-mediated blockage of basally active Gαi signaling was probably sufficient to allow enhanced cAMP accumulation upon pharmacological AC stimulation.
The CRE-reporter assay has a low ng/ml-range sensitivity for PTX 29 , comparable to CFA 19 , yet the question of its practical use in vaccine industry awaits further studies.

5
In this work, we set out to establish a sensitive microtiter plate format bioassay for PTX, based on kinetic measurement of intracellular cAMP levels in living cells in combination with a defined and tightly controllable inhibitory GPCR pathway.
Cell lines -Human embryonic kidney cell line (HEK293) was obtained from American Type Culture Collection (ATCC, #CRL-1573). HEK293 with stable overexpression of Gs22/cAMP probe, as well as the derived sensor cells with stable overexpression of SSTR2 (aka HEK-Gs/SSTR2_HA), were developed and characterized by us earlier [31][32][33]  were added to the wells as 20 μl of 10x solutions to yield the desired 1x working concentration. Nontreated controls received 20 μl of the specified HEPES (25 mM, pH 7.4) buffer per well. Further, the plates were placed back in the incubator and kept at the above-specified conditions for the scheduled time to allow for PTX to act. Once the desired exposure time had elapsed, the plates were retrieved from the incubator, the medium was removed and the wells were refilled with 45 μl of the freshly For initial inspection and qualitative analyses of iGIST data, the captured luminescent reads were plotted as intracellular cAMP kinetic curves (luminescence vs time) and subjected to visual assessment. Subsequent quantitative analyses involved several steps of data transformation and were carried out as follows. Firstly, cAMP kinetic curves were processed to obtain baseline signal-

RESULTS AND DISCUSSION
iGIST bioassay robustly detects PTX-induced abrogation of Gαi signaling -iGIST is based on stably-transfected HEK293 sensor cells (HEK-Gs/SSTR2_HA), co-expressing somatostatin receptor 2 (SSTR2), which is a well-characterized GPCR negatively regulating the cAMP-producing ACs through PTX-targeted Gαi, and a luminescent cAMP probe GloSensor-22F 35,36 . GloSensor-22F, originally introduced by Wood et al. 35,36 , represents a cAMP-binding domain of protein kinase A fused to a circularly permuted Photinus pyralis luciferase, jointly functioning as a sensitive and reversible cAMP probe in living cells. The sensor cells were earlier established in-house in HEK293 background, which has low endogenous expression of SSTR2 32,33 , and used to measure SSTR2mediated signaling upon exposure to various ligands. In iGIST, activities of SSRT2 and ACs are controlled with a high-affinity synthetic peptide agonist octreotide (Oct) 37  We incubated the sensor cells with a commercial PTX#1 preparation and subsequently challenged them with FSK or FSK + Oct 10 nM. Importantly, in view of the earlier noted high sensitivity of the sensor cells to certain compounds such as organic solvents and alcohols 32 , iGIST bioassay followed a strict parallel design with every dose of PTX#1 evaluated against the matched dose of the PTX#1 solvent (SolC#1; 50% glycerol in H2O with 50 mM Tris, 10 mM glycine and 0.5 M NaCl). iGIST luminescence readout was firstly plotted as raw signals vs time (Figure 2A-C), which allows for quick visual assessment of the effects. Then, to obtain quantitative observer-independent estimate of the effects, we rendered the raw luminescence signals into numerical area under the curve (AUC) values, normalized to AUC of FSK response in the control sensor cells (not exposed to PTX#1 or SolC#1 before FSK stimulation). FSK response in the control cells served as an internal calibrator in the assay and was taken for 100% for every given run. The derived values were denoted as AUC%values and utilized for deduction of PTX effects on Gαi signaling through pair-wise PTX#1 vs SolC#1 comparisons (Figures 2D-E and S2). Finally, to characterize Gαi signaling across a range of PTX#1 and SolC#1 exposures, we calculated the Gαi signal relay index (Gαi-SRI), expressed as a ratio of AUC%-values for FSK vs combination of FSK + Oct (AUC% FSK / AUC% FSK + Oct 10 nM; Figure 2F-H) at every given PTX#1 and SolC#1 dose. At full abrogation of Gαi signaling by PTX, the sensor cells are expected to lose responsiveness to Oct, with Gαi-SRI approaching 1.0. Schematics of iGIST output values and of their calculations is shown in Figure S1.
iGIST robustly registered PTX-induced abrogation of Gαi signaling that was proportional to the PTX#1 dose and time in contact with the cells. iGIST demonstrated the highest sensitivity at the longest PTX incubation studied (24 h), revealing a nearly complete abrogation of Gαi signaling at already 10 ng/ml of PTX (Figures 2F and S2C-D). With shorter incubations, the PTX dose required for abrogation of Gαi signaling increased with the assay reliably capturing PTX#1 activity at 100 ng/ml with 8h incubation (Figure 2D-E/G), and at around 1000 ng/ml with the 4h incubation ( Figures 2H and S2A-B). The global pattern of FSK response in PTX#1-treated sensor cells closely followed the one of SolC#1. Although comparisons of FSK responses at 100 ng/ml PTX#1 vs SolC#1 after 8h and 24h reached statistical significance (Figures 2D and S2C), the actual differences were very small, and thus likely had no practical relevance. The cAMP levels in the sensor cells without FSK stimulation were not significantly affected by PTX#1 across the dose range studied (Figures 2A-C, luminescent signals before black arrowhead), which is in line with the earlier reports 27 29 .
As all the above evidence was obtained with a single PTX preparation (PTX#1), we validated the iGIST bioassay with another PTX formulation, from a different vendor and having a different solvent composition (PTX#2; in 10 mM Na2HPO4 and 50 mM NaCl in H2O). The response pattern of iGIST to PTX#2 was virtually identical to the one of PTX#1. Though the effect started to emerge at already 1 ng/ml of PTX#2, abrogation of Gαi signaling became profound at 10 ng/ml of the toxin (Figure 3B-C) -the same threshold dose as with PTX#1 after 24h incubation. Apart from a borderline increase at 10 pg/ml, PTX#2 did not alter the pattern of FSK response, recapitulating the effects of SolC#2 ( Figure 3A). Basal cAMP levels before FSK addition also stayed unaffected with PTX#2.
Collectively, iGIST reliably detected PTX activity with two unrelated PTX preparations, revealing PTX-induced abrogation of Gαi signaling at ng/ml levels of the toxin.
iGIST reveals an unexpected potentiation of Gαi signaling at low PTX dose -When comparing PTX dose responses after 24 h, we unexpectedly detected a potentiation of Gαi signaling with 100 pg/ml of PTX, manifested as an increment of Gαi-SRI (Figures 2F and 3C). This effect seems paradoxical and difficult to understand from a standpoint of the canonical PTX activity, i.e. abrogation of Gαi signaling. Yet, the potentiation of Gαi signaling at 100 pg/ml PTX dose was highly reproducible, pronounced and consistently detected with both the toxin preparations (PTX#1 and PTX#2). Our initial model of PTX effect, based on Gαi-SRI and fitted through a non-linear regression (four-parameter logistic curve for an inhibitory response with a variable slope) could not accommodate this outlier. The resulting sigmoid curves (the red ones, Figures 2F and 3C) predicted a simple unidirectional inhibitory response from low-ng/ml levels of PTX onwards. The data urged us to consider an alternative model of PTX effect, which could be described by a bell-shaped curve with a truncated left arm when fitted through a 5 th order polynomial regression ( Figure 3D). The resulting alternative model of PTX effects on Gαi signaling fits the experimental data much better.
According to the alternative model, PTX exerts no effects on Gαi signaling at the lowest exposure tested (10 pg/ml), potentiates at 100 pg/ml dose and starts to abrogate at higher doses. Canonical abrogation of Gαi signaling with PTX is manifested first by a drop in Gαi-SRI back to the baseline level at around 1 ng/ml of the toxin (Figure 3D). This roughly corresponds to Gαi-SRI of 2 -the value reflective of Gαi signaling across the studied dose range of SolCs. Then, the effect continues to increase dose-dependently, reaching saturation with a nearly-complete abrogation of Gαi signaling at 10 ng/ml of PTX (Gαi-SRI of 1) (Figure 3D).
The potentiation of Gαi signaling by low-dose PTX in 24 h incubation, as revealed by the iGIST, is highly intriguing. Admittedly, the exact molecular basis remains a matter of subsequent studies. As for now, we hypothesize that the phenomenon relates to the dynamics of how the different G protein α-subunits are complexed and functionally regulated with G protein βγ-subunits 38 . However, in view of the time scale of the potentiation effect, more complex compensatory mechanisms could be involved such as a transcriptional and/or translational response. Irrespectively of the nature of the underlying molecular mechanisms, detection of the potentiation effect has a profound application potential. First, it increases the sensitivity of iGIST two orders of magnitude, from ca 10 ng/ml down to 100 pg/ml of PTX. Secondly, it adds to the specificity of iGIST as reconstruction of the truncated bell-shape PTX response curve through serial dilution of an analyte would ensure the specific nature of the observed signal ( Figure 3D). Collectively, iGIST reveals a hitherto undescribed potentiation effect of PTX on Gαi signaling, which improves specificity and increases the sensitivity of the iGIST to pg/ml range of PTX.
iGIST bioassay is more sensitive than CHO cluster formation assay to detect PTX -Sensitivity of the iGIST bioassay was compared with that of the cluster formation assay (CFA), originally introduced by Hewlett et al. 19 . To allow for direct comparison with iGIST bioassay, we obtained two strains of wild-type CHO cells from two different sources (designated as CHO#1 and CHO#2), and utilized the cells for time and dose-range studies with the earlier used toxin preparation (PTX#1).
Through parallel use of the two CHO strains, representing the progeny of the same maternal CHO culture, we strived to mitigate the risks, associated with possible genetic and phenotypic drift in immortalized cell lines upon extended culturing 39,40 .
Both strains of CHO were subjected to live cell imaging with IncucyteHD imager under regular incubator conditions up to 72h from the moment of PTX addition. The derived phase-contrast images were visually graded by six independent observers, using a 3-tier scale (0 -no effect  Figure S3). Here, a minor decline in estimated confluence was already noticeable at 1 ng/ml of PTX#1 at 48 h of treatment, with the effect becoming clear and pronounced from 10 ng/ml of PTX#1 onwards. The data underlines the subjective nature of visual grading in the conventional CFA, suggesting that observerindependent software-driven image analysis might make a better option for CFA. Most importantly, however, the data demonstrates that the iGIST bioassay is more sensitive to detect PTX#1 than CFA (ca 100-fold, with a threshold of 100 pg/ml of PTX) (Figures 2F and S2C-D).

13
iGIST bioassay detects PTX spiked into Boostrix pertussis acellular vaccine -The commonly acknowledged limitation of CFA 19 and other proposed animal-free bioassays to detect PTX 27, 29 is their poor compatibility with the final vaccine product due to cytotoxicity of the aluminum-based adjuvants 17 . This problem is pending, despite that several approaches to mitigate adjuvant toxicity, e.g. by means of vaccine dilution or barrier methods such as semi-permeable transwell inserts for culture plates, have been proposed 17 . To analyze the applicability of iGIST for PTX detection in complex samples, i.e. commercial pertussis vaccines, we prepared ACV dilution series supplemented with a fixed PTX concentration. As industry-grade PTX-toxoid was not available, we spiked a known dose of the active PTX#1 to achieve a final concentration of 100 ng/ml into serial dilutions of PTXtoxoid-containing vaccine (Boostrix; includes 16 μg/ml of PTX-toxoid, admixed with tetanus and diphtheria toxoids). Despite the complexity of Boostrix, including significant levels of aluminum (≤ 0.08% w/v), detergent (Tween80) and two other toxoids (diphtheria and tetanus) with possible residual activity, iGIST successfully detected the spiked PTX#1. This was evidenced by the complete abrogation of Gαi signaling with Gαi-SRI of 1 in all the analyzed Boostrix dilutions (³1:10) ( Figure   5). The unspiked Boostrix was not neutral in terms of its effects on Gαi signaling in iGIST ( Figure   5C, blue dots), but in the absence of the matched SolC the nature of the observed responses remains unknown. Importantly, we did not observe overt cytotoxicity (i.e., cell detachment, cell death) even at the most concentrated Boostrix solution tested (1:10 dilution) within the time window of the assay (24h) (Figure S4). Taken together, our results with the PTX#1-spiked Boostrix underline functional robustness of iGIST and highlight iGIST as a promising tool for PTX detection in complex samples.
iGIST -objective digital readout and prospects for automation -An important advantage of iGIST is its observer-independency. This feature, combined with the microtiter plate format and the digital nature of iGIST readout, opens avenues for automatization, e.g. by means of robotic platforms for plate handling and luminescence acquisition. Data processing at a higher throughput might be approached through utilization of tailored scripts, rendering iGIST luminescence signals into numerical values, such as Gαi-SRI (AUC% FSK / AUC% FSK + Oct 10 nM). For an alternative numerical index of PTX activity that streamlines data interpretation and thus might be more compatible with automated data processing, we propose a comparative Gαi-SRI, calculated as a ratio of (AUC% FSK / AUC% FSK + Oct 10 nM) values for PTX-exposed and matched SolC-exposed samples (Figure S5).
Reflective of the relative change in Gαi signaling in the sensor cells, be it potentiation or abrogation, and accounting for the effects of SolC, comparative Gαi-SRI should readily highlight PTX exposures.
Comparative Gαi-SRI can also be used to measure iGIST inter-assay variability, i.e. we obtained a 3-point composite estimate of 15,23% (equaling average coefficient of variation for comparative Gαi-SRIs for PTX 10 pg/ml, 100 pg/ml and 10 ng/ml at 24h of exposure, taken for no-effect level, maximal stimulation and inhibition, respectively; for 4x independent runs). Subsequent studies with appropriate controls, i.e. individual vaccine components and SolCs from different steps of PTX vaccine manufacturing process, are required to delineate the industry-scale applicability of iGIST.

CONCLUSIONS
We established iGIST (Interference in Gαi-mediated Signal Transduction), a kinetic microtiter plate format bioassay to detect PTX at pg/ml levels by measuring its effect on inhibitory GPCR signaling.
iGIST is observer-independent, has an objective digital readout and exceeds in sensitivity by 100fold the currently used in vitro end-point technique to detect PTX activity, the cluster formation assay in Chinese hamster ovary cells 17 . iGIST also detects PTX in complex samples, i.e. a commercial PTX-toxoid containing pertussis vaccine Boostrix that was spiked with an active PTX. We conclude that iGIST is a useful new tool for PTX basic research 5 , PTX-targeted drug development 43

Figure 2. iGIST detects PTX#1-induced abrogation of Gαi signaling. A-E) FSK and Oct responses
in the sensor cells after PTX#1 (w/v dose) or matched SolC#1 (corresponding stock dilution) exposure for 8h at +37 °C. Luminescence signals from a single representative experiment with the selected doses (A-C) and integrated results (as AUC%-values) of several independent runs (D-E). For curves in A-C, depicting raw luminescence reads, error bars denote +/-SD (only upper half shown) and yand x-axes denote luminescence signal (AU) and time (s), respectively. The moment of FSK and Oct addition is indicated with the black arrow. For bar diagrams in D-E, y-axis depicts AUC%-values derived from the luminescence signals in several independent runs (response to FSK in control sensor cells that were not subjected to PTX#1 or SolC#1 is taken for 100%). Error bars represent average values +/-SEM. F-H) Oct/SSTR2-mediated effects on cAMP levels in sensor cells, measured as Gαi signal relay index (AUC% FSK / AUC% FSK + Oct 10 nM at a given dose of PTX#1 or SolC#1) after exposure to PTX#1 or matched SolC for 24, 8 and 4 h at +37 °C, respectively. Panels F-G depict integrated results of several independent runs (average values +/-SEM; see also Figure S2C-D). Panel H is based on data from a single representative experiment in 3x technical replicates (mean +/-SD; refer also to Figure S2A-B). Dose-response curves were fitted with a non-linear regression. A state of complete abrogation of Gαi signaling (AUC% FSK / AUC% FSK + Oct 10 nM = 1) is indicated with the black dotted line. All the assays were run in standard conditions, in 3x technical replicates. The number of individual assay repeats (n#) for bar diagrams ≥3, if not indicated otherwise. Significant differences for comparisons of responses at corresponding doses of PTX#1 vs SolC#1 are indicated with asterisks (further information in Experimental Section). Inferential statistics was only performed when n≥3 for individual assay repeats. The yaxis depicts AUC%-values derived from the luminescence signals (response to FSK in control sensor cells that were not subjected to PTX#2 or SolC#2 is taken for 100%). All the assays were run at standard conditions. Significant differences for comparisons of responses at corresponding doses of PTX#2 vs SolC#2 are indicated with asterisks (further info in Experimental Section). C) Oct/SSTR2-mediated effects on cAMP levels in the sensor cells, measured through Gαi signal relay index (i.e., ratio of AUC% FSK and AUC% FSK + Oct 10 nM) after exposure to PTX#2 (w/v dose) or matched SolC (corresponding stock dilution) for 24 at +37 °C. Curves are based on the same data as shown in panels A-B (average values +/-SEM), and fitted through a non-linear regression (four-parameter logistic curve with a variable slope). The state of complete abrogation of Gαi signaling (AUC% FSK / AUC% FSK + Oct 10 nM ratio of 1.0) is depicted with the black dotted line. D) Gαi signal relay index (AUC% FSK / AUC% FSK + Oct 10 nM) at different toxin doses, from PTX#1 and PTX#2 experiments with iGIST (same data points as on Figure 2F and Figure 3C); curve fitting through a 5 th order polynomial regression.