Toward a mechanism-based in vitro safety test for pertussis toxin

Abstract

Pertussis vaccines are routinely administered to infants to protect them from whooping cough. Still, an adequate safety test for pertussis toxin (PT), one of the main antigens in these vaccines, is not available. The histamine sensitization test is currently the only assay accepted by regulatory authorities to test for the absence of active PT in vaccines. This is however, a lethal animal test with poor reproducibility. In addition, it is not clear whether the assumed underlying mechanism, i.e., ADP-ribosylation of G proteins, is the only effect that should be considered in safety evaluation of PT. The in vitro safety test for PT that we developed is based on the clinical effects of PT in humans. For this, human cell lines were chosen based on the cell types involved in the clinical effects of PT. These cell lines were exposed to PT and analyzed by microarray. In this review, we discuss the clinical effects of PT and the mechanisms that underlie them. The approach taken may provide as an example for other situations in which an in vitro assay based on clinical effects in humans is required.

Keywords: :

• pertussis toxin
• vaccine
• safety
• in vitro
• microarray

Introduction

Pertussis, or whooping cough, is a highly contagious, acute respiratory disease that is caused by the gram-negative bacterium Bordetella pertussis. Prior to the availability of vaccines, pertussis was a major cause of childhood morbidity and especially childhood mortality. Pertussis vaccines have proven very efficient in reducing the incidence of pertussis.¹ Pertussis vaccines are either killed intact bacteria (whole-cell vaccines; WCV) or specific, inactivated components from the bacterial cell wall (acellular vaccines; ACV), the latter being preferred because they appear less reactogenic.¹ One of the principal antigens present in all pertussis vaccines is pertussis toxin (PT).² Detoxification of PT is essential for the safe use of pertussis vaccines because native PT has several undesirable biological activities. On the other hand, detoxification should not be too rigorous to avoid losing the antigenic properties of PT. Thus, detoxification is a balanced process and regulatory authorities therefore require efficacy and safety testing of pertussis vaccine batches before market release.^3-5

Two tests for residual PT are named in the European Pharmacopeia, i.e. the histamine sensitization test (HIST) and the Chinese Hamster Ovary (CHO) cell clustering assay.⁴ Because aluminum adjuvants appear to be toxic for CHO cells, the HIST is the only test that can be used for final vaccine formulations. The HIST is based on the principle that mice vaccinated with biologically active PT are sensitized for histamine, resulting in a decrease of the lethal dose of histamine.⁶ The HIST is a lethal mouse test, although modifications have been proposed to avoid lethality.^7,8 Additionally, the HIST suffers from large variations in test performance depending on mouse strain, number, age, injection and challenge route.^9,10 Together, this makes replacement of the HIST highly desirable.

Clinical Effects of Pertussis Toxin Should Be the Basis for the Development of Novel Safety Tests

Several in vitro alternatives have been developed as possible replacements for the HIST or the CHO clustering assay.^11-14 They are based on the assumed mechanism underlying increased histamine sensitization and CHO cell clustering, i.e., PT-induced ADP-ribosylation of G proteins. However, these in vitro alternatives assume that this mechanism is the only effect of PT that should be taken into consideration. It is not clear, however, whether other mechanisms or cellular signaling routes also play a significant role in the clinical effects of PT and should therefore be taken into account for safety testing. We therefore first determined which clinical effects of PT are relevant and thus need to be prevented.¹⁵ Thus, instead of developing a replacement for the HIST and the CHO clustering assay we developed a safety test that is based on the clinically relevant effects of PT.

Identifying clinical effects of PT in humans is challenging as direct infusion of PT in humans is not performed for obvious reasons, although there is one exception.¹⁶ Besides PT, B. pertussis infection introduces endotoxin and other B. pertussis toxins and adhesins, which all have clinical effects, making distinction with PT-specific effects difficult. Nonetheless, linking data from clinical observations of B. pertussis infections (ref. 1 and references herein) to data obtained from in vivo and in vitro experiments indicates a probable role for PT in hypoglycemia, pneumonia, lymphocytosis, intractable pulmonary hypertension, hypotonic responses, seizures and encephalopathy.

A classic effect of PT in mice is increased insulin secretion by pancreatic β cells resulting in hypoglycemia.¹⁷ Toyota et al. administered PT to 6 healthy volunteers and observed increased insulin levels, confirming this effect in humans.¹⁶ In addition, neonates with severe pertussis presented with marked hypoglycemia, suggestive of increased insulin levels.^18,19

The role of PT in developing pneumonia is more complex and appears temporally differentiated. In the initial infection, PT increases the circulating white blood cell mass by increasing the release of cells from extravascular sites²⁰ and by preventing subsequent extravasation of these cells.^21,22 Both humans and animals infected with B. pertussis and mice injected with PT present with leukocytosis.^1,23,24 In mice, after initial inhibition of the influx of immune cells into pulmonary tissue, high numbers of neutrophils are recruited to the lung.²⁵ Lung tissue from children with a fatal pertussis infection showed abundant aggregates of leukocytes, indicating massive infiltration of these cells in the later stages of infection²⁶ suggesting a similar process in humans.

The increased circulating white blood cell mass induces increased vascular resistance which may contribute to the induction of pulmonary hypertension. Infection-induced hypoxia appears to be a second cause of pulmonary hypertension.²⁶ In both these aspects of pulmonary hypertension, possibly the most catastrophic clinical complication of pertussis in children, PT may play a role.

PT perturbs vascular smooth muscle contraction resulting in vasodilatation, which is associated with the hypotonic response observed in pertussis. Pulmonary hypertension is the opposite effect of vasodilatation. Indeed, in a study of 15 infants aged <4 mo who died of B. pertussis infection, 11 patients showed severe systemic hypotension while pulmonary hypertension was also specifically documented for 5 patients²⁶ suggestive of co-existence of these effects.

The involvement of PT in inducing the seizures and encephalopathy that have been observed during pertussis infections^27,28 is less clear. Although seizures and encephalopathy are most likely due to cerebral hypoxia related to extreme coughing fits, PT may play a role in facilitating brain inflammatory responses by increasing vascular permeability.^29-31

Taken together, the main clinical effects in humans where PT is involved are (1) increased insulin secretion with resulting hypoglycemia, (2) leukocytosis, (3) lung edema and (4) inflammatory responses, together resulting in pulmonary hypertension and pneumonia. Moreover, PT can induce (5) systemic hypotension, and is possibly involved in inducing (6) neurological problems.

ADP-Ribosylation of G Proteins is a Common Mechanism Underlying the Identified Clinical Effects of Pertussis Toxin

PT is released into the extracellular medium by Bordetella pertussis. It is functionally divided into an A and a B subunit; a classic AB-toxin.² The A, or enzymatic subunit, contains the ADP-ribosyltransferase activity, whereas the B, or receptor subunit, is required for cellular entry.^2,32,33 The A subunit catalyzes the transfer of ADP-ribose from NAD to the α subunit of heterotrimeric, guanine nucleotide binding (G) proteins in eukaryotic cells.^34-37 ADP-ribosylation of these G proteins uncouples G protein signaling from ligand binding to G protein-coupled receptors, and is implicated in all of the above-mentioned clinical effects of PT.

PT-mediated ADP-ribosylation of G proteins in pancreatic β cells disrupts Gi α-mediated inhibition of adenylyl cyclase activity resulting in unrestrained cAMP formation with consequent increased insulin release^23,38-40 leading to hypoglycemia. Likewise, PT mediates ADP-ribosylation of G proteins involved in vasoconstriction in vascular smooth muscle cells. This decreases the sensitivity for contraction-inducing agonists.^41-43 Consequently, maintaining blood pressure is impaired, possibly leading to the observed hypotonic responses, especially in the presence of vasodilating agents such as histamine. This PT-induced inability of the vascular system to respond to histamine-induced vasodilation, that leads to decreased blood pressure and mortality in mice is very likely the underlying mechanism of the HIST.⁴¹ PT-mediated ADP-ribosylation also plays a significant role in increasing barrier permeability involved in lung edema. The increased permeability can also facilitate passage of immune cells implicated in pneumonia and, in case of the blood-brain barrier, neurological problems. Cells involved in PT-induced increased permeability include pulmonary epithelial cells,^44,45 pulmonary endothelial cells,⁴⁶ and brain microvascular endothelial cells.^29,31 In these cells PT-mediated ADP-ribosylation of G proteins appears not to lead to increased cAMP levels, as observed in many other cells including pancreatic β cells, but involves phosphatidylinositol 3-kinase (PI-3 kinase) and protein kinase C (PKC) signaling.^31,45,46 Neutrophils and (alveolar) macrophages are also targets of PT-mediated ADP-ribosylation.^47-49 Taken together, PT-mediated ADP-ribosylation plays a significant role in all clinical effects of PT, and should thus be incorporated in safety evaluation of pertussis vaccines. However, ADP-ribosylation may not be the only factor involved in the clinical effects of PT (Table 1).

Table

Clinical effect	Cellular mechanism	Cell types involved	Involvement of ADP-ribosylation	Involvement of receptor-mediated pathways
Hypoglycemia	Increased pancreatic insulin secretion	Pancreatic β cells	Yes^23^,^38^-^40	Not likely
Pulmonary hypertension	Leukocytosis and infection-induced hypoxia	Circulating neutrophils and leukocytes	Yes^21^,^49	Possibly
Pneumonia	Increased pulmonary infiltration of immune cells	Pulmonary endothelial and epithelial cells, alveolar macrophages, circulating immune cells	Yes^29^,^44^-^46^,^48	Possibly^62^,^65
Hypotonic responses	Perturbed contractile apparatus of vascular smooth muscle cells	Vascular smooth muscle cells	Yes^41^-^43	Not likely
Seizures/ encephalopathy	Increasing vascular permeability and attraction of leukocytes	Brain microvascular endothelial cells and circulating leukocytes	Yes^29^,^31	Possibly^30]

ADP-Ribosylation may not be the Only Mechanism Underlying the Clinical Effects of Pertussis Toxin

Mechanisms other than ADP-ribosylation may play a role in the other clinical effects in which PT is implicated. These effects can be ascribed to the B subunit alone or to the complete toxin and appear to be receptor-mediated.

Compared with the well-studied mechanism of PT-mediated ADP-ribosylation of G proteins and its clinical consequences, much less is known about receptor-mediated effects of PT. Shortly after the isolation of PT in 1978,⁵⁰ the B subunit was found to not only facilitate cellular entry of the A subunit, but also to exhibit biological activity by itself.³³ The B subunit was shown to have mitogenic activity in lymphocytes and an insulin-like action to enhance glucose oxidation.³³ Both the holotoxin and the B subunit were found to induce a rapid increase in cytosolic free [Ca²⁺] associated with an increase in cellular diacylglycerol and inositol triphosphate.⁵¹ It was suggested that PT interacts with a specific receptor in the T lymphocyte membrane.⁵¹ The nature of specific receptors for the B subunit that facilitate entry of the A subunit and/or exert biological activities have been subject of debate since then. Different receptors in various cell types have been suggested as candidates for B subunit-mediated effects.^51-64 However, except for glycoprotein Ib⁶¹ specific binding of PT to any of these receptors could not be shown. A role for Toll-like receptor 4 (TLR4) in mediating PT effects independent of ADP-ribosylation has also been demonstrated. In an important study by Nishida et al. siRNA-induced inhibition of TLR4 expression⁶⁴ abolished PT-induced Rac activation and enhancement of angiotensin type 1 receptor function but failed to affect ADP-ribosylation, showing that indeed a functional effector pathway of PT next to ADP-ribosylation does exist. Besides in vitro effects, TLR4 was also shown to mediate effects of PT in vivo.³⁰

Regarding the cell types that are implicated in the clinical effects of PT, an ADP-ribosylation-independent pathway induced by PT was demonstrated in endothelial cells.⁶⁵ Both the purified B subunit and a PT holotoxin mutant that lacks ADP ribosyltransferase activity were shown to activate endothelial cell p42/p44 MAP kinases.⁶⁵ Although this pathway was later shown not to increase endothelial permeability by itself⁴⁶ it may contribute to increased vascular permeability. Dendritic cells, implicated in generating the Th1/Th17 response observed during B. pertussis infection,²⁵ have been shown to maturate by PT in an ADP-ribosylation-independent manner and very likely through TLR4 signaling.^62,63,66,67

In conclusion, 2 signaling pathways exist by which PT can elicit biological responses, the first one being dependent on ADP-ribosylation and the second one occurring independent of ADP-ribosylation. This second pathway occurs through binding of the B subunit with cell surface receptors. In this regard, it is important to note that ADP-ribosylation-dependent effects appear to occur at very low PT concentrations, while the ADP-ribosylation-independent effects require significantly higher concentrations. This difference is complemented by the fact that the onset of action is slow for ADP-ribosylation-dependent effects, while it is fast for ADP-ribosylation-independent effects.⁶⁰

Approach Toward a Mechanism-Based in Vitro Safety Test for Pertussis Toxin

Not all effector mechanisms of PT are completely understood, nor are all of its clinical effects. It is therefore important that in order to develop new animal-free safety tests the molecular mechanisms underlying these effects should be unraveled. In our view, employing omics-based techniques to assess the involvement of molecules at different levels (i.e. genes, metabolites, proteins) is the way forward to develop animal-free testing systems. As a starting point toward this mechanism-based testing, we studied PT-induced gene expression changes using microarray analysis. In order for our analysis to encompass a maximum number of the pathways utilized by PT to exert its clinical effects, we included multiple human cell lines in our screen. These cell lines were based on the cell types known to be involved in the clinical effects of PT.

Thus, using microarray analysis we screened 6 human cell types (bronchial epithelial cell line BEAS-2B, fetal lung fibroblast cell line MRC-5, primary cardiac microvascular endothelial cells, primary pulmonary artery smooth muscle cells, hybrid cell line EA.Hy926 of umbilical vein endothelial cells and epithelial cell line A549, and immature monocyte-derived dendritic cells) for differential gene expression induced by PT.⁶⁸ Immature monocyte-derived dendritic cells (iMoDCs) were the only cells in which PT induced significant differential expression of genes. The results were confirmed using different donors and were further extended by showing specificity for PT in comparison to Escherichia coli lipopolysaccharide (LPS) and Bordetella pertussis lipo-oligosaccharide (LOS). Statistical analysis indicated 6 genes, namely IFNG, IL2, XCL1, CD69, CSF2, and CXCL10, as significantly upregulated by PT and this was demonstrated at the protein level for those genes that encode secreted proteins. IL-2 and IFN-γ gave the strongest response. The minimal PT concentrations that induced production of IL-2 and IFN-γ in iMoDCs were 12.5 and 25 IU/mL, respectively. High concentrations of LPS slightly induced IFN-γ but not IL-2, while LOS and detoxified PT did not induce production of either cytokine. In conclusion, using microarray analysis we evaluated 6 human cell lines/types for their responsiveness to PT and found 6 PT-responsive genes in iMoDCs of which IL-2 is the most promising candidate to be used as a biomarker for the detection of residual PT.

Thus, instead of focusing on developing an alternative method to test for a single known mechanism by which PT induces clinical effects, we took a step back and focused on outlining all effects on gene expression by using microarray analysis in relevant human cell lines. This approach proved to be successful, although the extent of gene expression changes was limited. As a next step, other omics-based techniques such as metabolomics, proteomics or kinomics could be employed in the same way. This mechanism-based approach may have the potential to contribute to animal replacement, not only in pertussis vaccine safety testing, but in vaccine safety testing in general.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

We would like to thank Dr D Xing (NIBSC) and Dr FR Mooi (RIVM) for helpful suggestions.

10.4161/hv.28001