Penicillin and cephalosporin cross-reactivity: role of side chain and synthetic cefadroxil epitopes

Background Analysis of cross-reactivity is necessary for prescribing safe cephalosporins for penicillin allergic patients. Amoxicillin (AX) is the betalactam most often involved in immediate hypersensitivity reactions (IHRs), and cefadroxil (CX) the most likely cephalosporin to cross-react with AX, since they share the same R1 side chain, unlike cefuroxime (CO), with a structurally different R1. We aimed to analyse cross-reactivity with CX and CO in patients with confirmed IHRs to AX, including sIgE recognition to AX, CX, CO, and novel synthetic determinants of CX. Methods Fifty-four patients with confirmed IHRs to AX based on skin test (ST) and/or drug provocation test (DPT) were included. Serum sIgE to AX and benzylpenicillin was determined by Radioallergosorbent test (RAST). Two potential determinants of CX, involving intact or modified R1 structure, with open betalactam ring, were synthesised and sIgE evaluated by RAST inhibition assay. Results Tolerance to CX (Group A) was observed in 64.8% cases and cross-reactivity in 35.2% cases (Group B). Cross-reactivity with CO was only found in 1.8% cases from Group B. ST to CX showed a negative predictive value of 94.6%. RAST inhibition assays showed higher recognition to CX as well as to both synthetic determinants (66% of positive cases) in Group B. Conclusions Cross-reactivity with CX in AX allergic patients is 35%, being ST not enough for prediction. R1, although critical for recognition, is not the unique factor. The synthetic determinants of CX, 1-(HOPhG-Ser-Bu) and 2-(pyrazinone) are promising tools for determining in vitro cross-reactivity to CX in AX allergic patients.

Cross-reactivity has important clinical implications, especially for searching safe alternative for further treatments, and an accurate diagnosis based on skin testing (ST) is recommended, being the role of drug provocation tests (DPT) controversial [3,9,26]. In vitro evaluation of cross-reactivity to BLs, mainly based on immunoassays, is limited by the difficulty for studying the structure of cephalosporin-protein conjugates [27]. Although several reports have addressed this issue [28][29][30], the antigenic determinants of cephalosporins are currently not wellknown [31].
To our knowledge, structure-activity relationship (SAR) studies have been the unique successful approach for investigating cephalosporin epitopes [28][29][30]32]. In this context, we have elucidated precise epitope structures through synthesis and immunologic evaluation of well-defined structures proposed as antigenic determinants for cephalosporins with different R1, bearing different functionalities at the C-6 of the cephalosporin (methyl, hydroxymethyl, aldehyde, mercaptomethyl) and without involvement of the remaining dihydrothiazine ring [29,30]. Moreover, we have identified a novel synthetic pyrazinone structure as an antigenic determinant of cefaclor [28], formed after reaction of the amino group in the R1 with the likely aldehyde functionality at C-6 of the original cephalosporin [28,32]. CX is another aminocephalosporin that could follow the same fragmentation and reactivity pathways as cefaclor [32].
In this study we have evaluated the in vivo degree of cross-reactivity with CX and CO in patients with confirmed IHR to AX and the immunological recognition of AX and these cephalosporins by serum specific IgE (sIgE). The ultimate aim of this study was to evaluate if synthetic structures, proposed as potential antigenic determinants mimicking the fragment of CX, which would remain coupled to the protein, can help get insight into the structure responsible for CX allergies and, therefore, study cross-reactivities between AX and CX.

Patients
The studied group was obtained from the Regional University Hospital of Málaga Drug Allergy Database. This prospective cohort includes all patients with confirmed drug allergy from 1984 to 2019 after an allergological workup including clinical history, ST, and DPT.
Patients with IHR to AX (allergic to the whole penicillin group or AX selective reactors with good tolerance to penicillin V (PV)) were diagnosed following the European Academy of Allergy and Clinical Immunology (EAACI) recommendations [9,33]. Tolerance to CX and CO was evaluated and, based on CX tolerance, patients were classified into: Group A with tolerance (demonstrated by negative ST and DPT) and Group B with cross-reactivity (demonstrated by positive ST or DPT) (Fig. 2).

Skin test
Skin prick (SPT) and, if negative, intradermal tests (IDT) were performed as described [9,33], using benzylpenicilloyl-poly-L-lysine (PPL, DAP, Diater, Leganés, Spain) at 1.07·10 -2 M, minor determinant mixture (MDM: benzylpenicillin, benzylpenicilloate, and benzylpenilloate) at 1.5 M and AX (Diater laboratories, Madrid, Spain); CX (Lilly SA, Madrid) and CO (GlaxoSmithKline S.A, Madrid) all at 20 mg/mL. Since May 2011 DAP composition has changed and includes the major determinant benzylpenicilloyl-octa-L-lysine (BP-OL) at 0.04 mg/mL, equivalent to 8.64·10 -5 M concentration of the benzylpenicilloyl (BPO) moiety, and the minor determinant (MD) at 0.5 mg/mL, equivalent to 1.5·10 -3 M concentration of sodium benzylpenilloate. Cephalosporin reagents were prepared according to Romano [19,34]. Readings were done after 20 min and considered positive: (i) In SPT, if a wheal larger than 3 mm surrounded by erythema appeared, with a negative response to the control saline; (ii) In IDT, if the increase in diameter of the wheal area marked initially was greater than 3 mm surrounded by erythema. Positive data expressed as the mean diameter recorded by measuring the largest and the smallest diameters at right angles to each other [35].

Drug provocation test
In subjects with negative ST to PPL/BP-OL and MDM/ MD, oral DPT with PV was performed at incremental dose (50, 100, 100, 150 mg) each 40-min until reaching the total cumulative dose (TCD) of 400 mg, followed by a 2 day therapeutic course of PV of 400 mg/8-h at home [33]. If DPT with PV and ST to AX was negative, oral DPT with AX was performed (50,100,150,200 54 patients 150, 200 mg) until TCD of 500 mg, followed by a 2 day therapeutic course of CX 500 mg/8-h. Finally, CO was administered following this procedure. Patients were carefully monitored during DPT and for 2 h after the last dose, complete equipment for cardiopulmonary resuscitation was available [36].

In vitro sIgE determination by radioallergosorbent test (RAST)
It was done using BP and AX conjugated to Poly-L-Lysine (PLL) (Sigma, St. Louis, MO) resulting in BPO-PLL and AXO-PLL in the solid phase, as described [37,38], and radiolabeled anti-IgE antibody (kindly provided by Thermo Fisher Scientific and radiolabelled in our laboratory) [28]. Samples were considered positive if they were higher than 2.5% of label uptake, which was the mean + 2SD of a negative control group.

RAST inhibition assay
This was done as described [38], incubating sera from patients with RAST values higher than 7% with different BLs (AX, CX, and CO) and the synthetic determinants of CX (1 and 2) in two ten-fold decreasing concentrations (100 mM and 10 mM) for 18 h at room temperature. After this, the AXO-PLL disc was added, followed by the previous described RAST procedure. The results were expressed as percentage inhibition with respect to the non-inhibited serum. Comparison of the inhibition capacity of the different inhibitors was made at 50% inhibition.

Statistical analysis
Description of quantitative variable included the median, mean, standard deviation (SD), and interquartile range (IR). Differences in percentage between the groups were compared by Chi-square analysis, numeric demographic data by Student t test. Comparisons for variables without a normal distribution were performed by the Mann-Whitney test for non-related samples and by Friedman test for related samples. All statistical analyses were done using the software package GraphPad PRISM v7. A P < 0.05 was considered statistically significant.

Results
From 1393 patients with confirmed BL hypersensitivity evaluated from 1984 to 2019, 994 subjects were confirmed with IHRS to AX, from which we randomly selected 54 patients, whose cross-reactivity to CX and CO was evaluated and flow-charts analysed (Fig. 2 Table 1).
In all cases, ST with CX was done, with 37 (68.5%) cases negative and 17 (31.5%) positive (6 (35.3) by SPT and 11 (64.7%) by IDT) ( Table 2). From the 37 cases with negative ST to CX, DPT was done with this cephalosporin, being positive in 2 (5.4%) ( Table 3). P37 developed urticaria in trunk 50 min after 150 mg of CX and needed antihistamines and corticosteroids, and P53 developed anaphylaxis 30 min after 50 mg of CX and needed epinephrine.

Tolerance to cefadroxil happens in 65% of AX allergic patients
Based on CX study, 35 (64.8%) cases showed tolerance (Group A) and 19 (35.2%) cross-reactivity (Group B). Comparisons of the clinical characteristics of both groups showed no differences regarding type of the original reaction to penicillins, time interval between drug administration and symptom development, or time between last reaction and study. Comparisons of ST to AX showed that in Group A, 12 (34.3%) cases were positive by SPT, 19 (54.3%) by IDT and 4 (11.4%) negative whereas in Group B 15 (78.9%) cases were positive by SPT, 4 (21.1%) by IDT, and 0 (0%) negative, being differences statistically significant (p = 0.006).
The allergological study to CO showed negative STs in all cases (Table 2) and tolerance was also confirmed in all cases by DPT (Table 3) but only P34 (Group B) developed urticaria in trunk and arms 25 min after 50 mg of CO. Symptoms resolved 2 h after antihistamine administration. Therefore, cross-reactivity with CO was 1.8%, although a concomitant sensitisation rather than a crossreactivity could be hypothesized.

Significant differences of recognition are only found at the lower concentration of cefadroxil
The analysis of sIgE results indicated that the mean value of RAST to BPO-PLL and AXO-PLL was 1.12 ± 3.65 and 6.8 ± 9.4 respectively, with 4 out of 48 (8.3%) cases positive to BPO-PLL and 24 out of 48 (50%) to AXO-PLL (Table 2). Comparisons between groups A and B showed higher differences, although not discriminating, in terms of mean levels to AXO-PLL and the percentage of positive cases (76.5% vs 35.5%;) for AXO-PLL in Group B (p = 0.038 and p = 0.007, respectively).
To study CX specific recognition, we performed RAST inhibition assays on 6 cases from each group (Fig. 4a).  As inhibitors, we included AX, CX, and CO at two concentrations, 10 and 100 mM (Fig. 4b).
Results with AX showed, as expected, a high percentage of inhibition at both concentrations in all cases. Regarding CX, the percentage of inhibition was above 50% in most of patients at 100 mM, 5 out of 6 patients in each group, similarly to levels obtained with AX. However, these percentages decrease at 10 mM, being lower than those observed with AX especially Group A (Fig. 4a). In fact, comparison analysis of the percentage of inhibition between groups only shows significant differences for CX at 10 mM (p = 0.034) (Fig. 4c). Only one case (P38, Group B) showed a percentage above 50% with CO.

Synthetic determinants of cefadroxil showed better sIgE recognition in Group B
The design of the two synthetic determinants of CX was based on our degradation hypothesis of the aminocephalosporin-protein conjugate, using butylamine as a model nucleophile emulating protein lysine (Fig. 3a). After covalent protein conjugation through BL ring, the dihydrothiazine ring is unstable and could degrade producing structures in which carbon 6 presents different functionalities. Two relevant candidates, according to previous immunological recognition results [30], are structures bearing hydroxyl and aldehyde functionality in carbon 6. In the case of hydroxyl functionality, it would generate the molecule 1 as determinant; whereas the aldehyde functionality can react with the amino group of R1 side chain generating the pyrazinone 2 as a novel determinant. The synthesis of the molecule 2 was achieved following the Ugi/Deprotect/Cyclize strategy (Fig. 3b) [39]. First, starting reagents (an isocyanide, a protected amine, a protected aldehyde, and a N-protected aminoacid) were assembled by following the one-pot Ugi four-component reaction to produce the Ugi adduct. The latter acidmediated-cleavage of the protected groups may result in the amino-functionalised aldehyde intermediate that cyclises, through intramolecular imine formation, and aromatises affording target pyrazinone (2). This method allowed the straightforward synthesis of 2, for which other procedures resulted unsuccessful. Compounds 1 and 2 were purified and well-characterised, allowing the immunological recognition study of precise chemical structures. RAST inhibition assays were performed using CX and the two synthetic structures (1 and 2), as inhibitors (Fig. 5b), in two cases from Group A and 6 from Group B. There was no inhibition with these structures in Group A (Fig. 5a). Higher percentages of inhibition were observed in Group B, being greater than 50% in 4 out of 6 cases at 100 mM, in which similar levels of inhibition to those obtained with CX were observed (Fig. 5a). However, significant lower percentage of inhibition with these synthetic structures was observed performing the RAST inhibition at 10 mM (p = 0.0022 for both) (Fig. 5c).

Discussion
BLs are the most widely used antibiotics and the drugs most frequently involved in IHR [1] in adults and children [40][41][42]. All BL compounds can potentially induce a specific immunological response and, due to their wide prescription, BL allergy is nowadays a worldwide health issue with relevant implications [43][44][45]. One of the main issues is establishing the risk of developing an allergic reaction to cephalosporins prescribed in patients previously diagnosed of penicillin IHR, with different unsolved questions like if this risk can be predicted by ST and/or DPT, or the role of the chemical structure, specifically the side chain, in this recognition [10,[46][47][48][49]. The main difficulty is that, despite efforts [28][29][30], the antigenic determinants of cephalosporins are unknown [31]. In this study we have found that, for predicting cross-reactivity, ST is not enough and, R1 side chain, although critical for recognition, is not the unique factor. Moreover, the use of chemical tools for SAR study is a promising approach   10:57 for elucidating the chemical structures involved in these IHRs.
In this study we have found that the cross-reactivity of IHRs to AX with CX, a cephalosporin with the same R1, was 35% and with CO, cephalosporin with different R1, figures decrease to 1.8%. Results with CX are in agreement with those by Romano [19] reporting that 39.3% of patients with IHR to penicillins had positive tests for cephalosporins, with 37.7% positive to aminocephalosporins, including CX and/or cefamandole. These results are similar to our previous data, with 38% of cross-reactivity between AX and CX using DPT [21]. Relevantly, we did not detect differences in cross-reactivity to CX among those selective to AX compared to those reacting to the whole group, confirming that R1 is not the only factor influencing cross-reactivity. Regarding CO tolerance, all patients had negative ST and only one had a positive DPT, showing a high degree of CO tolerance, in line with previous data [16][17][18][19]25].
Comparisons of SPT results to AX between Group A and B (cross-reactivity to CX) showed a higher percentage of positivity (78.9 vs 34.3%) in the group tolerant to CX (Group A). These results agree with those by Romano [19] estimating an odds ratio of ST positivity to ampicillin for cross-reacting to at least one cephalosporin of 2.5 (CI, 1.4-4.5). Moreover, the analysis of the sIgE results showed significantly higher levels and positivity to AXO-PLL in Group B. This seems to indicate that patients that cross-react with cephalosporins have a high degree of reactivity, taking into account that the two cases that developed systemic symptoms after ST with penicillins belong to Group B and that patients reacted to small amount of CX (50 and 150 mg) and CO (50 mg) in DPT.
Regarding the role of ST for predicting cross-reactivity, a positive cephalosporin ST in patients allergic to penicillins may indicate not only cross-reactivity but also concomitant sensitivity. Of note, P34 with cross-reaction to CO also reacted to CX. Whether this patient has crossreacting or co-existing antibodies was something we cannot clarify in the present study as the RAST level was not enough for performing RAST inhibition with both drugs. However, cross-reactivity is more probable since this patient had not been previously treated with cefuroxime or any cephalosporin. This percentage is in agreement with previous data [18] that found 2.9% of cross-reactivity with CO in patients with prior histories involving only a penicillin. If a negative cephalosporin ST predicts good tolerance is controversial [50]. Different studies showed that patients with a well-established IgE-mediated allergy to penicillin and with ST negative to cephalosporins tolerate cephalosporins [15][16][17]. However, others demonstrated that less than 3% of cases can have a DPT positive with cephalosporin despite having negative ST [18,19]. In this study 2 out of 37 patients (5.4%) with ST negative to CX and 1 out of 54 patients (1.8%) with ST negative to CO had a positive DPT to CX and CO respectively, indicating a negative predictive value (NPV) of 94.6% for CX and 98.1% for CO. That means that although NPV are high, a negative ST does not mean tolerance even if R1 are different.
Our immunological study by RAST inhibition assays agrees with previous results on cross-reactivity between penicillins and cephalosporins, showing that AX presented a better recognition, followed by CX [21]. Data showed a discriminating capacity of the test between Group A and B using lower drug concentrations, 10 mM, observing a significantly lower recognition of CX in patients with good tolerance to CX (Group A).
Regardless of this discriminative capacity, these data indicate that, although important for IgE recognition, the R1 is not per se the only structure involved in the immunological response, as structural modifications or some fragments of the nuclear structure may be involved in the antigenic determinant. In penicillins, the penicilloyl  structure formed after protein conjugation is stable and, therefore, the thiazolidine ring could also play a role in the antigenic determinant [51][52][53]. On the contrary, the equivalent cephalosporyl structure is unstable, thus the R2 substituent is expulsed [54,55] and the dihydrothiazine ring suffers different fragmentations, producing a complex mixture in which structures are difficult to elucidate [29,31]. We have addressed this issue, by using chemical tools, for performing SAR studies in which precisely defined structures, consisting on the R1 side chain coupled to the open BL ring with the carbon 6 of the original drug represented by a methyl group, were recognised by sIgE from patients with IHR to the cephalosporin containing either the same R1 or one structurally similar [29]. Further SAR studies involved similar synthetic determinants but with different functionalisation in such carbon 6, finding that hydroxymethyl and aldehyde functionality, compared with methyl group, increased recognition [30]. Based on these results, synthetic determinants of CX, involving the whole intact R1 or a modified R1 side chain, have been immunologically evaluated, showing higher-recognition by sIgE from patients cross-reactive to CX (Group B). The structure 1 (HPhG-Ser-Bu), consisting on the R1 side chain of CX and open BL ring with hydroxymethyl functionality at carbon 6 [30], was not previously evaluated with sIgE to aminocephalosporins. These determinants containing the intact corresponding  aminocephalosporins R1 have been immunologically evaluated in a recent study with cefaclor-allergic patients (12% of positive cases) [28], and in the present study with AX-and/or CX-allergic patients (66% of positive cases at the maximum concentration), showing different extent of recognition depending on R1.
The pyrazinone 2 has been synthesised and immunologically evaluated in this study for the first time. Its structure derives from intramolecular reaction between the R1 amino group and the aldehyde at carbon 6. Inhibition results in six cases of Group B show that the pyrazinone 2, at 100 mM concentration, is recognised in 66% of cases, in agreement with IgE recognition observed for pyrazinones derived from cefaclor, with 63% of positive cases for the equivalent pyrazinone to that described here [28], and 60% of patients for an equivalent analog developed by Venemalm [32].
These synthetic determinants (1 and 2) were not recognised by the two selected patients with tolerance to CX (Group A). Importantly, greater differences in recognition  Fig. 4 Immunological recognition of different BLs evaluated by RAST Inhibition using AXO-PLL as solid phase. a Graphs show the percentage of inhibition produced by different BLs, including amoxicillin, cefadroxil, and cefuroxime at two concentrations 10 and 100 mM in 6 individual sera from Group A (left) and 6 individual sera from group B (right). b Chemical structures of BLs, amoxicillin, cefadroxil and cefuroxime, used in the fluid phase. c Mean of the percentage of inhibition induced by Amoxicillin (AX) and Cefadroxil (CX) at two different concentrations in sera from AX-allergic patients from Groups A and B. Statistical analysis for non-related samples was performed by Mann-Whitney U test and significance considered for p < 0.05 Bogas et al. Clin Transl Allergy (2020) 10:57 between CX and the synthetic structures were observed in Group A than in Group B, using the higher concentration. One could think that AX presents the amino group in R1 for the formation of additional determinants, as diketopiperazine, considered as a minor determinant of AX [56]. However, it did not show sIgE recognition in previous studies [57], which is consistent with its lack of reactivity with proteins [56].

Conclusions
We have confirmed that cross-reactivity between penicillin and cephalosporins occurs when the R1 side chain is identical as previously reported, and that negative ST is not enough for predicting tolerance, being DPT necessary. The primary determinant of immunochemical recognition of aminocephalosporins rested, with the structure of the R1, intact (molecule 1) or in its cyclised form as pyrazinone (molecule 2), although other parts of the molecule (excluding R2 substituents and most of the dihydrothiazine) are necessary for the formation of