Penicillin causes non-allergic anaphylaxis by activating the contact system

Immediate hypersensitivity reaction (IHR) can be divided into allergic- and non-allergic-mediated, while “anaphylaxis” is reserved for severe IHR. Clinically, true penicillin allergy is rare and most reported penicillin allergy is “spurious”. Penicillin-initiated anaphylaxis is possible to occur in skin test- and specific IgE-negative patients. The contact system is a plasma protease cascade initiated by activation of factor XII (FXII). Many agents with negative ion surface can activate FXII to drive contact system. Our data showed that penicillin significantly induced hypothermia in propranolol- or pertussis toxin-pretreated mice. It also caused a rapid and reversible drop in rat blood pressure, which did not overlap with IgE-mediated hypotension. These effects could be countered by a bradykinin-B2 receptor antagonist icatibant, and consistently, penicillin indeed increased rat plasma bradykinin. Moreover, penicillin not only directly activated contact system FXII-dependently, but also promoted bradykinin release in plasma incubated-human umbilical vein endothelial cells. In fact, besides penicillin, other beta-lactams also activated the contact system in vitro. Since the autoactivation of FXII can be affected by multiple-factors, plasma from different healthy individuals showed vastly different amidolytic activity in response to penicillin, suggesting the necessity of determining the potency of penicillin to induce individual plasma FXII activation. These results clarify that penicillin-initiated non-allergic anaphylaxis is attributed to contact system activation, which might bring more effective diagnosis options for predicting penicillin-induced fatal risk and avoiding costly and inappropriate treatment clinically.

Scientific RepoRtS | (2020) 10:14160 | https://doi.org/10.1038/s41598-020-71083-x www.nature.com/scientificreports/ The contact system is a plasma protease cascade initiated by activation of factor XII (FXII). Activated FXII (FXIIa) initiates two branches of downstream events that mediate the interface between coagulation and inflammation. The coagulation branch-activated FXI initiates fibrin formation and the inflammatory branch (also called kallikrein-kinin system)-activated prekallikrein cleaves the nonenzymatic cofactor high molecular weight kininogen (HK) to liberate oligopeptide bradykinin (BK) 18 . BK is involved in the regulation of inflammatory processes, vascular permeability, and blood pressure. Recent data have linked FXIIa-driven formation of BK and the downstream activation of the G-protein-coupled receptor B2 (B2R) signaling to anaphylaxis 19,20 . The name "contact system" comes from the mode of FXII being activated, as "contact" with negatively charged surfaces triggers FXII activation via conformational rearrangement. Activators that initiate FXIIa formation in vivo include polyphosphate, heparin, misfolded proteins, collagen, nucleic acids (DNA and RNA), oversulfated chondroitin sulfate and artificial surfaces 21,22 .
In the present study, we demonstrate for the first time that penicillin-driven FXII contact activation triggers kallikrein-kinin system, thus releasing BK to cause anaphylaxis. Our findings suggest that targeting contact system-produced BK or its downstream signaling is a promising strategy for prevention and treatment of anaphylaxis triggered by penicillin.

Materials and methods
Materials and reagents. Penicillin G sodium salt (1,550 U/mg) and other beta-lactams were from Solarbio Life Sciences (Beijing, China) and Yuanye Biotechnology Ltd. (Shanghai, China), respectively. Icatibant was from MedChemExpress (Monmouth Junction, NJ, USA). S-2302™ was from Chromogenix (Milano, Italy). BK ELISA kit was from DS Pharma Biomedical Co., Ltd. (Osaka, Japan). Kaolin, dextran sulfate sodium salt (DS; Mr ~ 500,000), propranolol, triprolidine, SB290157 and CV3988 were from Sigma-Aldrich (Darmstadt, German). PMX53 and pertussis toxin (PTX) were from GL Biochem Ltd. (Shanghai, China) and Enzo Life Sciences (Farmingdale, NY, USA), respectively. Citrate-anticoagulant standard human plasma and FXII-depleted plasma were from Boatman Biotech Co., Ltd. (Shanghai, China). Antibodies for human FXII and transferrin were from GeneTex Inc. (San Antonio, TX, USA). Shrimp tropomyosin (ST) and its IgE monoclonal antibody (mAb) were prepared as we previously described 23 . Human plasma was obtained from eight healthy volunteers. cells and animals. Primary human umbilical vein endothelial cells (HUVEC) were from Lonza (Basel, Switzerland). The male Balb/c mice (18-20 g) and SD rats (180-200 g) were purchased from Vital River Experimental Animal Services (Beijing, China) and housed in an SPF laboratory.
The anaphylactic shock assay in PTX-pretreated mice was performed according to previously described procedures 27 . Briefly, the mice were injected (i.p.) with PTX (200 ng/mouse) on days 1 and 3. On day 7, the mice were injected (i.v.) with normal saline (NS) or penicillin (30 KU/mouse). Thirty minutes later, the rectal temperature was measured. evans blue extravasation assays. Evans blue extravasation assays in mouse hindpaws and rat dorsal skin were measured as previously described 28,29 . Measurement of blood pressure and BK generation in rats. The rats were anesthetized and placed in a supine position. Catheters were placed into the carotid artery. Systolic arterial blood pressure was monitored in the left carotid artery with an arterial catheter connected to a polygraph (RM6240C, Chengdu, P.R. China). After equilibrating, the rats were injected (i.v.) with penicillin (1,250 KU/kg) or NS, and blood pressure was continuously monitored. For the antagonist experiment, rats were treated with icatibant (200 μg/kg, i.v.) 5 min prior to penicillin challenge.
For the determination of plasma BK, once penicillin (1,250 KU/kg) or NS was injected (i.v.) into the anesthetized rats, arterial blood was immediately collected and added to ice-cooled absolute ethanol. BK level was determined by ELISA.
The overlapping effect of penicillin-and antigen/IgE-caused hypotension was measured using a rat undergoing passive systemic anaphylaxis. To clarify the respective role of penicillin and antigen, we deliberately chose anti-ST IgE, a non-penicillin IgE mAb, to induce IgE/antigen hypotension. Rats were passively sensitized intravenously (i.v.) with 1 mg/kg of anti-ST IgE. Twenty-four hours later, the rats were challenged (i.v.) with ST (10 mg/ kg) with or without penicillin (1,250 KU/kg), and the arterial blood pressure values were measured.
Amidolytic activity assay. Amidolytic activity was assayed according to previously described procedures 30 .
Briefly, 100 μL of plasma was pretreated with 100 μL of penicillin at various concentrations (diluted by Tris buffer: 50 mM Tris-HCl, 0.117 M NaCl, pH 7.8) at 37 °C. Ten minutes later, 100 μL of the chromogenic substrate S-2302 (1.5 mg/mL) was added and further incubated at 37 °C for 30 min. The reaction mix was centrifuged at 3,000 × g for 5 min. Supernatant absorbance was monitored at 405 nm. Kaolin and DS were used as positive control activators of the contact system. Buffer alone was included as a negative control. Data presentation. The data reported as the mean ± SD from a representative experiment. All of the experiments reported in this work were repeated at least three times with the same pattern of results. The statistical analysis was performed using a one-way ANOVA. A student's t-test was used when only two groups were compared. P < 0.05 was considered significant.

Results
penicillin lowers rectal temperature in propranolol-or ptX-pretreated mice. Given that penicillin can provoke severe and life-threatening anaphylaxis clinically, we first evaluated whether penicillin could cause anaphylactic shock (detected as hypothermia) 7,24 . Mice were made more sensitive to penicillin-induced shock by pretreating propranolol or PTX, which does not induce shock by itself but can augment the susceptibility of shock via enhancing vascular permeability [25][26][27]32 . Penicillin contributed to obvious hypothermia in both propranolol-and PTX-pretreated mice ( Fig. 1), indicating that penicillin can also cause a non-allergic anaphylactic shock.

Penicillin increases microvascular permeability in a BK-dependent manner.
The local effect of penicillin on microvascular permeability was determined using the Evans blue extravasation assay. As shown in Fig. 2A, penicillin significantly (diameter of blue spot ≥ 5 mm) increased rat dorsal skin vasopermeability when the dose was ≥ 100 KU/mL (100 μL per spot). Moreover, it potently caused mouse plantar blue dye (right hindpaw) compared with NS (left hindpaw) (P < 0.01). To address the underlying mechanism, we next screened 6 antagonists (CV3988, SB290157, PMX53, triprolidine, fasudil and icatibant). The results showed that penicillin-increased microvascular permeability was independent of PAF, C3a, C5a, histamine and ROCK 33 (data not shown), but could be countered by the B2R antagonist icatibant (Fig. 2B).  www.nature.com/scientificreports/ penicillin contributes to anaphylaxis by contact system activation (cSA) in rodents. Based on the above results, we next evaluated whether icatibant could block penicillin-caused non-allergic anaphylaxis. First, we evaluated the effect of icatibant on penicillin-induced hypothermia in propranolol-pretreated mice. As expected, icatibant indeed significantly countered penicillin-caused mouse hypothermia (P < 0.01; Fig. 3A). Given that acute hypotension is an underlying symptom of anaphylaxis 34 , the effect of penicillin on rat arterial blood pressure was then determined. Unlike intra-arterial administration of heparin 19 , intravenous injection of penicillin could trigger a rapid and transient drop in systemic arterial blood pressure. Pretreatment with icatibant almost completely blocked penicillin-caused hypotension (Fig. 3B,C). Along with the hypotension, plasma BK was markedly increased after penicillin challenge (Fig. 3D). These findings demonstrate that BK plays a crucial role in penicillin-initiated non-allergic anaphylaxis. In addition, to determine a possible overlapping effect of penicillin-and antigen/IgE-caused hypotension, SD rats were sensitized passively with a non-penicillin IgE (anti-ST IgE). The overlapping effect assay was carried out by intravenously injecting penicillin and antigen (ST) into the ST-IgE-sensitized rats. As a result (Table 1), penicillin and ST could respectively cause hypotension in normal and antigen/IgE-sensitized rats. Unexpectedly, penicillin did not cooperate with ST to trigger more drastic hypotension in antigen/IgE-sensitized rats.
penicillin activates the contact system in an fXii-dependent manner. Direct activation of the contact system by penicillin was further investigated in vitro. Amidolytic activity was assessed in the standard plasma and FXII-deficient plasma by the addition of the S-2302 chromogenic substrate that can be hydrolyzed by plasma kallikrein. Similarly to the FXII activator kaolin, penicillin directly activated the contact system in standard human plasma, and this effect almost disappeared in FXII-deficient plasma (Fig. 4A). Consistently, owing to being converted into FXIIa consisting of a 52-kD heavy chain and a 28-kD serine protease domain, prototypical FXII (80-kD) in penicillin treated-plasma was also significantly decreased (Fig. 4B). BK, a short-lived bioactive peptide 35 , is the terminal product of CSA cascade. Given that the entire BK-forming cascade can be assembled and activated along the surface of endothelial cells 36,37 , HUVEC were incubated with standard human plasma and then the plasma was removed. The obtained PI-HUVEC, which had combined all elements for BK generation, could be used for the measurement of BK in vitro. As shown in Fig. 4C, unlike non-PI-HUVEC, PI-HUVEC could also release a small quantity of BK without penicillin. In contrast to non-PI-HUVEC, penicillin markedly promoted BK release in the PI-HUVEC in a concentration-dependent manner.   18,40 or the negative charge density of FXII activators 41 ), etc. Thus, we reasoned that plasma from different individuals had distinct amidolytic activity. To confirm this point, we assayed respective amidolytic activity in plasma from volunteers. As shown in Fig. 5, eight plasma samples (nos. 1-8) showed a consistent concentration-response to DS, including a relatively weak response of no. 5 (Fig. 5A). In contrast to DS, the majority responsiveness to penicillin was lower, except for no. 6. And nos. 1, 2,  www.nature.com/scientificreports/ 3, 7 and 8 did not change significantly after penicillin treatment (≤ 250 KU/mL). Particularly, nos. 4 and 6, who had positive skin-test histories, reacted differently in response to penicillin: no. 4 had a high basal level but low sensitivity, while no. 6 had a low basal level but high sensitivity (Fig. 5B). These findings obviously indicate that the activation level of the contact system varies with diverse activators (e.g., penicillin and DS).

Effects of other beta-lactams on CSA.
To date, beta-lactams are the most successful chemical class of antibiotic widely used in the clinic to treat infections in humans 42 . They are mainly divided into 6 subclasses including penicillins (e.g., penicillin G, penicillin V, cloxacillin, amoxicillin and piperacillin), cephalosporins (e.g., cephalexin), oxacephems (e.g., latamoxef), single-ring beta-lactams (e.g., aztreonam), cephamycins (e.g., cefoxitin), and carbapenems (e.g., imipenem). We found that the effects of different beta-lactams on plasma amidolytic activity were various. Penicillin G was found to be the most effective, and cloxacillin, latamoxef and piperacillin had similar impacts. Other beta-lactams were found to be less effective (Fig. 6).

Discussion
Immediate hypersensitivity reaction (IHR) can be divided into allergic-and non-allergic-mediated 43 . According to the World Allergy Organization proposition, "anaphylaxis" is reserved for severe IHR whose symptoms include hypotension, vascular leakage, or even cardiac arrhythmia and bronchial constriction in severe cases 20,44 . Previous studies have demonstrated that the FXII-driven contact system critically contributes to the pathogenesis of anaphylaxis in murine models 20 . Clinically, anaphylaxis is associated with transient decreased plasma FXII, suggesting that the factor is consumed. The severity of anaphylaxis is correlated with the intensity of CSA and BK formation 20 . The abnormal production of BK leads to anaphylaxis and angioedema via its ability to increase inflammation and vessel permeability 21,45 . There are common triggers for anaphylactic reactions such as food, medications or insect venom, of which 44-57% of the fatal anaphylaxis was triggered by medications, far more than other inducements 3,46-48 . The most common lethal reasons for fatal anaphylaxis are asphyxia and shock, which can be exactly caused by BK [49][50][51][52] . An established concept is that penicillin can serve as a complete antigen to induce sIgE through irreversibly forming drug-protein conjugates in vivo 53,54 . Upon a secondary challenge, the penicillin-protein conjugate crosslinks its sIgE which bound to the high affinity receptor, FcɛRI, and leads to the degranulation of mast cells and basophils (Fig. 7), thus inducing type I hypersensitivity (e.g., anaphylaxis). In fact, true PenA is rare clinically, and most reported PenA is "spurious" 55,56 . Moreover, penicillin-induced IHR is possible to occur in both skin test-and sIgE-negative patients 57,58 , suggesting that a non-allergic mechanism exists. In fact, although penicillin could not directly induce human mast cell LAD2 degranulation in vitro (data not shown), it caused non-allergic anaphylaxis (Figs. 1, 3B). Our data showed that penicillin increased microvascular permeability, which could be countered by a B2R antagonist icatibant, rather than other antagonists (e.g., triprolidine and fasudil 33 ) (Fig. 2B). Moreover, icatibant could also block penicillin induced-anaphylactic shock (Fig. 3A) and  www.nature.com/scientificreports/ hypotension (Fig. 3B,C). It is worth mentioning that heparin, an endogenous FXII activator, causes hypotension only through intra-arterial application 19 , which may be attributed to its rapid catabolism in blood circulation. In the present study, we found that penicillin triggered hypotension via both intra-arterial and intravenous injection. Given the fact that intra-arterial application of penicillin, an exogenous stimulus, is clinically meaningless, only the results of intravenous challenge were shown. In line with the in vivo result (Fig. 3D), in vitro study indicated that penicillin initiated CSA through the activation of FXII (Fig. 4A,B) and finally led to BK formation (Fig. 4C). Together, penicillin-initiated non-allergic anaphylaxis was attributed to CSA in which FXIIa activates prekallikrein leading to BK release via cleavage of its precursor HK. The fact that penicillin can trigger both IgE-and non-allergic-mediated anaphylaxis raises the reasonable speculation about whether penicillin-induced severe or fatal anaphylaxis results from an overlapping effect. Unexpectedly, penicillin did not cooperate with antigen to trigger more drastic hypotension in antigen/IgE-sensitized rats (Table 1). In allergen/IgE-mediated type I hypersensitivity, mast cell or basophil-derived heparin can also initiate FXII contact activation and subsequently generate BK, which binds to B2R, triggering anaphylaxis 19,20 . Obviously, both allergen/IgE-and non-allergic-mediated anaphylaxis have the same pathway, namely the FXIIa-BK-B2R axis. Perhaps it is because the same axis that the overlapping effect did not occur. In addition, only one dose of penicillin was used in the rat blood pressure test due to the fact that penicillin caused hypotension in a non-dose-dependent manner (data not shown), which is similar to heparin 19 . As we know, FXIIa cleaves prekallikrein to form active kallikrein, which in turn reciprocally activates more FXII, forming a powerful activation amplification loop (Fig. 7). It is because of the "positive feedback loop" that, once initiated, the penicillin-induced cascade reaction reaches completion independent of the dosage of penicillin. In this context, other factors (e.g., plasma FXII or C1INH level, etc.) may be more crucial for reaction intensity.
FXII autoactivation can be influenced by multi-factors, such as FXII 38 and C1INH levels 39 , ionic milieu 18,40 , or the negative charge density of FXII activators 41 , etc. Indeed, our data also showed that plasma from different human individuals had vastly different amidolytic activity in response to penicillin (Fig. 5B and Supplementary file 2), of which high responders might be the potential anaphylaxis sufferer (e.g., no. 6 in Fig. 5B and no. 5 in Supplementary file 2). In fact, some clinical evidences also support our findings. For example, penicillin-induced IHR was much more likely with parenteral administration than oral administration 58 . Angioedema, BK-mediated swelling 45 , is also common in penicillin-induced IHR 59,60 . Perhaps, not coincidentally, Caucasians, whose FXII levels are significantly higher than Japanese 61 , suffer from more penicillin-induced IHR compared with Asians including Japanese 60 .
Given that true PenA is rare clinically and the FXIIa-BK-B2R axis is the same pathway for penicillin-triggered allergic-and non-allergic-IHR (Fig. 7) 19 , compared with the commonly used skin test or sIgE assay, it is perhaps www.nature.com/scientificreports/ more effective to determine the potency of penicillin (or beta-lactams) to induce individual plasma FXII activation or BK formation for predicting the risk of penicillin-induced IHR. We notice that the World Health Organization has established a "first international standard for FXII, Plasma, human" 62 , which also ought to be beneficial for predicting the risk of penicillin-induced IHR, even including that of other beta-lactams (e.g., penicillin G, cloxacillin, latamoxef, piperacillin and aztreonam, etc.; Fig. 6). Additionally, in the same volunteer, the degree of FXII activation in response to penicillin at the same concentrations is disparate over time (no. 5 in Fig. 5 vs. Supplementary file 2), which might be attributed to the change of plasma milieus. Thus, instant prediction, rather than long-term predication, may be available for predicting the IHR risk of penicillin. In summary, our study demonstrates, for the first time, that as an activator of the FXII-driven contact system, penicillin can lead to hypersensitivity reactions in rodent models, which can be hampered by icatibant, a B2R pharmacological inhibitor. These findings not only warrant further exploration of the penicillin-driven contact system as a source of therapeutic targets for treatment of anaphylaxis triggered by penicillin, but also might bring more effective diagnosis options for the prediction of penicillin-induced fatal risk.

Data availability
The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.