Magnetic nanoparticles enable clinically relevant in vitro diagnosis of β-lactam allergy Authors: Jose A. Céspedes, Maria I. Montañez, Isabel M. Jiménez, Violeta Gil-Ocaña, Esther Matamoros, María Salas, Adriana Ariza, Gádor Bogas, Cristobalina Mayorga, Yolanda Vida, Ezequiel Pérez-Inestrosa, Maria J. Torres Published: 2026 ABSTRACT β-lactam antibiotics are the leading cause of drug-induced hypersensitivity reactions. Yet available in vitro diagnostics still show limited sensitivity, specificity and reproducibility. In this work, we report the use of magnetic silica-coated nanoparticles functionalized with dendrimeric antigen conjugates (DeAn@Np) for an improved in vitro detection of β-lactam–specific IgE. These nanoparticles possess a high surface area and combine the chemical versatility of silica with the magnetic properties of iron oxide, enabling efficient magnetic separation and reducing assay handling times compared with centrifugation-based workflows. The chosen nanoparticle size (30 nm) was evaluated in immunoassays using sera from 50 amoxicillin-allergic patients and 44 tolerants, achieving 98% sensitivity and outperforming both conventional RAST (38%) and commercial ImmunoCAP (17%) in this cohort, with sensitivity gains of up to 60% (p < 0.0001). Remarkably, DeAn@Np discriminates effectively between selective and cross-reactive allergic patients, minimizing false positives and negatives with a 100% of specificity compared to conventional RAST (70.4%) (p = 0.005). These results highlight the potential of DeAn@Np as a sensitive and modular platform for β-lactam allergy diagnosis, combining improved analytical performance with practical advantages for future translational immunodiagnostic applications and safer penicillin allergy delabelling. Beyond drug allergy, these modular dendrimer-based magnetic nanoparticles can be extended to other antibody targets, offering a versatile strategy for immunodiagnostic applications. Introduction β-Lactam antibiotics (BLs) are among the most widely prescribed treatment options for bacterial infections. However, they are also the leading cause of drug-induced hypersensitivity reactions (DHRs), representing a significant clinical and public health challenge worldwide [1,2]. Between 8% and 25% of individuals are globally labelled as penicillin-allergic. However, in only 1–10% of them a true allergy is demonstrated [3,4]. Mislabeling has important clinical consequences, as it leads to the prescription of alternative antibiotics that are less effective or more toxic. This contributes to antimicrobial resistance, delays in appropriate therapy, and increased healthcare-associated risks [5,6]. Consequently, there is a growing global emphasis on incorporating penicillin allergy de-labelling into antimicrobial stewardship programs as a strategy to optimize antibiotic use and improve patient outcomes [3].For this purpose, precise diagnostic tools are critical. BL allergy can be triggered by various immunological mechanisms, with IgE-mediated mechanisms being the most common. Clinical manifestations range from mild cutaneous symptoms such as urticaria and exanthema, to severe, life-threatening conditions like anaphylactic shock [7]. Diagnosis is often challenging, as it requires a thorough clinical history and the use of in vivo diagnostic tools such as skin tests (STs) or drug provocation tests (DPTs), which involve risks for the patient. In vitro assays, including basophil activation test and the measurement of serum-specific IgE (sIgE), have emerged as valuable complementary diagnostic tools, although they generally exhibit low sensitivity [8]. To improve diagnostic safety and accuracy while minimizing the need for DPTs, current recommendations support the combined use of in vitro techniques with STs. This approach enhances diagnostic reliability and reduces the likelihood of adverse reactions during the evaluation process [1,9,10]. Among currently available in vitro diagnostic tools, the quantification of drug-sIgE in serum is the main method for evaluating suspected BL allergy [11]. Lately, immunoassays have been used to detect BL-sIgE, with differences in their antibody detection methods and solid phase. The ImmunoCAP system (Thermo Fisher), a standardized fluorescence enzyme immunoassay (FEIA), is the most widely used technique in clinical practice due to its accessibility and automation. However, its diagnostic performance for BL allergy is suboptimal, with variable and low sensitivity [[12], [13], [14]] (between 15.8 and 52.4% [15]) and false positive results in the case of benzylpenicillin (BP) [16,17]. The radioallergosorbent test (RAST) is a technique that has shown improved sensitivity (43–75%) and specificity (68–83%) for detecting BL-sIgE. [18]. However, it is currently restricted to specialized research settings due to safety and regulatory issues related to radioisotopes. Immunoassays for BL allergy diagnosis rely on analogous BL–carrier conjugates immobilized on a solid phase to capture serum BL-sIgE in vitro. These conjugates mimic the covalent drug–protein complexes formed in vivo during IgE-mediated allergic reactions, which occur when BL molecules bind to proteins. This process generates structures of sufficient molecular size to interact with the immune system and elicit the production of BL-sIgE, which specifically recognizes epitopes of the modified drug [8]. In this regard, solid phases of both immunoassays—commercial ImmunoCAP and conventional RAST— consist of a cyanogen bromide-activated cellulose polymer covalently bound to drug–poly-L-lysine (PLL) conjugates. While the ImmunoCAP configuration is established, standardized, and currently limited to five BLs, the RAST solid phase offers greater versatility, including a wider range of BLs and allowing for customization of carriers, solid phases, and activation chemistries [19]. The sensitivity and specificity of these immunoassays are influenced by the drug derivative structure after covalent binding to the carrier, the nature of the carrier molecule itself, the solid phase material and its chemical functionalization [8,11,19]. Moreover, drug density within the carrier molecule and exposure to the sIgE binding constitute critical factors for sensitivity. In the case of PLL (a carrier molecule widely used in immunoassays), although it has the capacity to carry a significant number of drug molecules, this amount can be variable and not all of them are exposed, decreasing the possibility of capturing drug-sIgE. Therefore, the development of improved solid phases remains an important objective for increasing analytical sensitivity in BL immunoassays. RAST has been optimized over recent years, particularly through the incorporation of nanomaterials aimed at enhancing sensitivity [20]. In this context, the growing success of nanotechnology in biomedical applications stems from the ability to design nanomaterials with tailored physicochemical properties. Specifically, nanoparticles offer distinct advantages for in vitro diagnostics as solid supports, including a high surface-to-volume ratio, tunable surface chemistry, and dimensions comparable to biological molecules, contributing to improved analytical performance [21,22]. When applied to diagnostic assays, and compared to conventional systems, these properties translate into enhanced sensitivity and specificity, offering opportunities to improve BL-sIgE detection performance [23,24]. Polyamidoamine (PAMAM) dendrimers have been proposed as effective substitutes for conventional carriers due to their structural homogeneity, well-defined multivalency, and chemical versatility, which allow for the reproducible attachment of high densities of drug haptens on their surface. These dendrimeric antigen conjugates (DeAns) have demonstrated strong recognition with serum sIgE from patients allergic to BLs such as benzylpenicillin (BP) or amoxicillin (AX) [25,26]. DeAns have been covalently immobilized onto conventional solid phase of RAST, cellulose discs [27], and onto novel materials acting as solid platforms for DeAns in immunoassays, as zeolites [28], and silica particles [29], to construct RAST systems, showing a promising performance. Additionally, silica (SiO2) composite nanoparticles (Nps) are among the most widely explored nanomaterials for diagnostic purposes [30]. Their synthesis is inexpensive and the particle size can be efficiently controlled. Besides, they are biocompatible and allow versatile chemical surface modification. Therefore, they provide a robust solid support for the immobilization of different biomolecules, making them ideal platforms in immunoassay applications [31,32]. SiO2 Nps exhibit a high surface area-to-volume ratio compared to cellulose discs, which may result in an increased density of DeAns on their surface and, consequently, improved sIgE binding and detection. In our first study, as a proof of concept, SiO2 Nps coated with DeAns (DeAn@SiO2) were prepared and evaluated in RAST using sera from three allergic patients and two tolerant controls, showing highly specific and sensitive AX-sIgE detection [33]. Furthermore, in a second study, multivalent mono- and bi-epitope materials were developed. They consisted of SiO2 Nps decorated with DeAns, where AX and/or BP were used as drug determinants. RAST evaluation, using sera from 21 allergic patients and 10 tolerant controls, enabled the simultaneous detection of allergy to the two different BLs, with a significant improvement compared to the corresponding BL-PLL discs [29]. Given the promising results obtained with the DeAn@SiO2 Nps in RAST, further efforts were directed towards translating such assays into clinical practice. However, a major limitation in the experimental procedure itself was identified during the multiple washing steps required, which considerably increases the overall assay time and can affect reproducibility. In clinical practice, this translates into handling numerous tubes and, consequently, long assay times. Moreover, the Nps size will encounter an additional limitation, since smaller sizes (< 400 nm) would exponentially increase the time required for clinical testing (due time-consuming centrifugation steps). The use of smaller Nps would result in a higher surface area/gram, thereby increasing the density of immobilized DeAns, improving sensitivity. Herein, the previously described DeAn@SiO2 strategy was translated into a magnetic Np-based platform to improve immunoassay handling and BL-sIgE detection performance. PAMAM-G2 dendrimers were selected as carrier molecules, and AX and/or BP were chosen as the drug determinants [34]. The Np surface was carefully functionalized, assessing the Np composition and quality. DeAns were anchored to Nps with methodology aimed at minimizing nonspecific interactions and facilitating the effective exposure of DeAn conjugates to sIgE. Combined with the larger interaction surface provided by the Nps, this approach is expected to improve antigen accessibility and enhance the probability of capturing low-abundance BL-sIgE. Importantly, an optimal Np size was selected to achieve the best balance between sensitivity and specificity as well as an easier synthesis procedure. The diagnostic performance of these magnetic core-shell Nps was evaluated by quantifying AX-sIgE and BP-sIgE levels in sera from allergic patients and tolerant subjects. The results were compared with those obtained using available immunoassay techniques (commercial ImmunoCAP and conventional RAST both using poly-L-lysine as carrier molecule). Results and discussions A key aspect in the development of efficient diagnostic immunoassays is ensuring reproducible handling and practical assay workflow. In this regard, it is essential to develop a stable and reproducible material for an easy handling by hospital staff. In this study, we have optimized two key aspects for clinical transfer: the size of the Nps (achieving the maximum hapten density which can influence the sensitivity of the test) and the method for recovering them from the supernatant during test development (the centrifugation step must be avoided, as it creates a bottleneck and increases the time required to perform the test that can affect reproducibility). Both issues can be solved using magnetic Nps. These will be easy to recover from the supernatant simply by applying a magnetic field and will permit reducing the size of Nps, as they do not depend on centrifugation. To maintain the properties of the previously studied Nps, the magnetic Nps have been coated with a layer of silica, which will allow us to apply our established chemical modification protocols. Iron oxide Nps are ideal candidates in this regard because of their magnetic properties, stability, ease of surface functionalization and environmental safety [29]. Furthermore, these magnetic properties have been exploited in the development of sensors for clinical diagnostics, yielding very promising results [[35], [36], [37]]. In this context, the silica coating provides a stable and chemically inert interface suitable for diagnostic applications. Fe3O4@SiO2 preparation and characterization The synthesis of the iron oxide Nps was carried out using the thermal decomposition of iron (III) chloride at high temperatures in an organic medium and in the presence of stabilizing surfactants [38]. This way, particles of four different sizes (Fe3O4 with average diameters of approximately 15, 30, 50 and 200 nm, Table S1) were obtained by varying solvent, iron precursor/surfactant concentration ratio, and temperature. Nps of approximately 200 nm in average diameter (SiO2) were prepared to confirm that the iron oxide core does not affect the RAST results. For comparison, both silica Nps with and without a magnetic core were prepared. The Stöber methodology was used for the silica coating of all iron oxide Nps. To this end, Fe3O4 were suspended in an ethanol/water/ammonium hydroxide mixture and treated with a solution of tetraethyl orthosilicate (TEOS) in ethanol. The obtained Nps (Fe3O4@SiO2) were analyzed to verify the composition and uniformity of the layer formed (as well as the size) and resulted in a solid that was easy to handle. The Fourier transformed infrared (FTIR) spectrum of Fe3O4 (Fig. S1) shows their characteristic peaks at 570 cm−1, whereas additional bands appear in the case of Fe3O4@SiO2, which are characteristic of the Si-O-Si and Si-OH stretching vibrations (1088 cm−1) [39]. In the case of 200 nm SiO2 particles, only the vibrational bands characteristic of Si-O-Si and Si-OH bonds can be observed. Transmission electron microscopy (TEM) images show a homogeneous distribution of the Fe3O4 Nps, with the expected average diameter in all cases (Fig. S3). STEM-EDX (Scanning Transmission Electron Microscope - Energy-Dispersive X-ray) was used to analyze the composition of the core and outer shell of Fe3O4@SiO2. The obtained images show the formation of a silica layer with an average thickness of approximately 5 nm around the Fe3O4 core (Fig. 1 and S4, Table S2). Furthermore, Zeta potential (ξ) measurements indicate a variation in the charge of Nps, from positive values for Fe3O4 to negative ones for Fe3O4@SiO2, thus confirming the formation of the silica outer shell (Tables S3 and S4) [34].Fig. 1STEM-EDX images of 30 nm Fe3O4@SiO2. A) STEM-HAADF images, B) Fe signal obtained after EDX analysis of the Nps, C) Si signal obtained after EDX analysis of the Np, D) superimposition of the three previous images. STEM-EDX images of 30 nm Fe3O4@SiO2. A) STEM-HAADF images, B) Fe signal obtained after EDX analysis of the Nps, C) Si signal obtained after EDX analysis of the Np, D) superimposition of the three previous images. The field-dependent magnetization (M vs H) of Fe3O4 and Fe3O4@SiO2 was measured at 298 K in a field of 19 kOe (Fig. S5), showing a superparamagnetic behavior in both cases. The naked Fe3O4 Nps exhibit a maximum magnetization of 28 emu/g Nps at 19 kOe, while the magnetization of Fe3O4@SiO2 were calculated to be approximately 10 emu/g Nps. Although coating with SiO2 reduced the magnetization, as a result of an increase in nonmagnetic components, the efficacy of the magnetic separation required under the assay conditions was satisfactory. DeAn@Np preparation and characterization Surface chemical modification is essential when Nps are intended for use in biological samples. In this study, Nps were functionalized using a previously established methodology that involved inexpensive, straightforward and well-known procedures. Furthermore, all reaction steps can be verified, ensuring the reproducibility of the final Nps composition [34,40].However, some of the conditions used during the chemical treatment differed slightly from those previously described, owing to the specific handling requirements of the magnetic material, which requires continuous agitation to maintain dispersion and prevent Nps aggregation. The coupling of the DeAn conjugates to the Nps surface was achieved thanks to the covalent binding between the dendrimer and previously modified Nps, following the sequence showed in Fig. 2.Fig. 2A) Chemical modification of the Nps surface to obtain DeAn@Np, B) Zeta potential (ξ) values obtained, C) Ninhydrin tests values obtained and estimation of the functionalization degree of the Np. A) Chemical modification of the Nps surface to obtain DeAn@Np, B) Zeta potential (ξ) values obtained, C) Ninhydrin tests values obtained and estimation of the functionalization degree of the Np. Primary amino groups were grafted onto the surface of the Nps (NH2@Np) by reaction of Fe3O4@SiO2 with a large amount of 3-(aminopropyl)-triethoxysilane (APTES). NH2@Np were then reacted with succinic anhydride, obtaining CO2H@Np, in which a large number of carboxylic groups was inserted in the Nps. PAMAM dendrimers (Generation-2) were covalently anchored to CO2H@Np via amide bonds formation, resulting in De@Np. Finally, the reaction of the free amino groups on the periphery of the dendrimers with the selected BLs resulted in the formation of DeAn@Np. Herein, two different composites were prepared using AX (DeAXO@Np) and BP (DeBPO@Np) as BLs [34]. The same procedure has been used to modify 200 nm SiO2 particles, preparing in this case only the AX derivative (DeAXONp). To control each step of the chemical modification, different standard techniques were used. Although direct characterization techniques such as X-ray photoelectron spectroscopy (XPS) or elemental analysis do not provide sufficiently reliable results because of the predominant contribution of Fe3O4 or SiO2 relative to the organic surface layer, indirect approaches including ζ-potential measurements and the ninhydrin test have been employed to estimate the degree of surface functionalization. Zeta potential measurements (ξ) are very useful for monitoring the entire process, as it was expected that the particle charge would change in response to the different chemical reactions. The observed ξ values (Fig. 2B–Tables S3 and S4) agree with the chemical modification of the particle surface. Negative values were observed for both Fe3O4@SiO2 and CO2H@Np [33,40]. In contrast, positive ξ values were observed for Nps with terminal amino groups in their surface (NH2@Np and De@Np) [28,34,41]. To prepare DeAn@Np, the amino terminal groups of the dendrimers in De@Np were reacted with the β-lactam ring of the corresponding antibiotic (AX or BP), forming the final conjugate with opened ring [42]. The selected incubation conditions are consistent with previous studies, where the conjugation reaction reaches a plateau under prolonged incubation, with previous reports supporting near-maximal functionalization under these conditions [27]. [40] Considering that all amino terminal groups of the dendrimer reacted, the charge of the Nps depends on the functional groups of the BL moiety. Therefore, the charge of the amoxicilloyl epitope (AXO, Fig. 2A) in DeAXO@Np was determined by both the carboxyl and the amino groups. In the case of the benzylpenicilloyl epitope (BPO, Fig. 2A) in DeBPO@Np, the charge will be determined only by the carboxyl group present in the structure. This agrees with data shown in Fig. 2B, where for all Nps sizes, the Zeta potential values (ξ) follow the same trend. The FTIR spectrum shows two remarkably peaks around 1640 and 1550 cm−1 in the case of De@Np and DeAn@Np, which are characteristic of the C=O and N-H vibrations in amide groups, thus confirming the presence of the dendrimers in the surface of the Nps (Fig. S2). An important variable for monitoring the chemical modification of Nps is the quantification of free amino groups, which is also very useful to estimate the degree of functionalization. Consequently, ninhydrin tests have been carried out to all Nps (Fig. 2C and Tables S5 and S6). As it can be seen, the amount of terminal primary amino groups inserted per gram of Nps increases as the size of the Nps decreases. After reaction with succinic anhydride a significant reduction in the number of terminal amino groups was observed for all Nps. We can estimate that for 200 nm Nps, around half of the amino groups on the particle surfaces reacted with succinic anhydride (insertion of around 60 μmol of carboxylic units/gram). For 50 nm Nps, the number of reacted amino groups increases (approximately 122 μmol of carboxylic units/gram). However, this data remains unchanged when the size of the Nps decreases to 30 nm, and varies only slightly when it decreases to 15 nm (around 134 μmol of carboxylic units/gram). A similar trend is observed after the reaction with the dendrimer (Tables S5 and S6). With these data in mind, we can estimate the insertion of approximately 9 μmol PAMAM-G2/gram for 200 nm Nps, around 13 μmol PAMAM-G2/gram for 50 nm Nps and a similar value for 30 or 15 nm Nps (Fig. 2C) [34,40,42]. These values indicate that, although reducing the size of the Nps increases their surface area per gram (and therefore the degree of surface modification), in the sizes from 30 to 15 nm reducing the size does not translate into a greater insertion of dendrimes per gram of Nps. As expected, the data obtained for both series of 200 nm Nps (with or without a magnetic core) are alike, indicating a similar level of functionalization. This demonstrates that even when the manipulation of the Nps is different, the methodology is reproducible, and that the presence of the magnetic core does not influence the chemical modification process. As mentioned before, the handling of magnetic material has led to some changes in the protocol. For chemical modification, vigorous stirring is essential to ensure proper Nps dispersion within the medium. In addition, continuous agitation is required to prevent magnetic aggregation and to maintain stable dispersions of DeAn@Nps in aqueous media. In all cases, the modified Nps were rapidly recovered from the reaction suspension simply by bringing closer a magnet to the reaction flask, causing the Nps to immediately accumulate against the flask wall. This magnetic responsiveness represents a major advantage of the overall system, as it enables a fast and straightforward recovery process without the need for any centrifugation steps, significantly simplifying and accelerating the washing procedure. The functionalized Nps can be stored as a dry solid at −18 to −20°C and readily redispersed prior to use without observable aggregation, preserving their functional properties. Freshly prepared dispersions exhibit good colloidal stability under the assay conditions. No visible aggregation was observed after several months of storage, indicating stable behavior under the conditions relevant for their use. 30 nm DeAXO@Np provides optimal performance for AX-sIgE detection To investigate the potential of DeAXO@Np as a solid phase for immunoassays, it is crucial to assess its capability to bind and measure AX-sIgE from serum samples. DeAXO@Np with different diameters (15, 30, 50 and 200 nm) were selected to systematically evaluate the influence of Np size on key assay parameters, including colloidal stability and suspension behavior (which directly affect interaction with IgE), magnetic separation efficiency, and antigen density on the Np surface. Smaller Nps provide higher surface area-to-volume ratios, which can increase DeAn loading. However, they may also exhibit reduced magnetic responsiveness and a higher tendency to aggregate. In contrast, larger Nps offer improved magnetic separation but lower surface area and potentially reduced hapten accessibility. Therefore, this size range was selected to evaluate the balance between these effects and identify the optimal Np size for IgE capture. RAST was performed using DeAXO@Np of different sizes (15, 30, 50 and 200 nm) at different amounts (1.5, 2.0 and 2.5 mg) in five sample types: (i) PBS as blank sample; (ii) pool of sera from confirmed allergic patients to AX, with high, or low values of AX-sIgE in RAST, as positive controls; (iii) pool of sera from tolerant subjects to AX, as negative controls, with standard values of total IgE or high values of total IgE to rule out possible unspecific results. First, when we compared the results obtained for 200 nm Nps with or without a magnetic core (Fig. S6), no differences were observed, indicating that the magnetic core does not affect recognition and IgE binding in RAST. For all sizes and concentrations of DeAXO@Np (Table 1), the % of RAST were higher for positive controls (allergic patients) than for negative controls (tolerant subjects). This difference increases when the higher amounts of DeAXO@Np (2.0 or 2.5 mg) were used in the assay compared to the lower amount (1.5 mg). Although a larger amount of DeAXO@Np in the assay was included in the study (3 mg), results were not favorable because DeAXO@Np aggregated at the concentration of the suspension and did not disperse properly. Regarding the DeAXO@Np size, an improvement of sIgE capture is observed with the smaller Nps (of 15, 30 and 50 nm) compared with the biggest one (200 nm Np) for both positive controls. Larger differences between Nps size were found when measuring samples from positive controls with high sIgE levels, for which the RAST results increased from 16.8 to 15.9% with the 200 nm DeAXO@Np (at 2.0-2.5 mg per assay, respectively) to 19.9-19.2%, 20.3-20.7% and 20.9-20.9% with the 50, 30 and 15 nm DeAXO@Np. It should be noted that % of RAST obtained with the negative controls show values around 0.1-0.2% with 30, 50 and 200 nm DeAXO@Np, which correctly discriminate a non-allergic patient. Despite bigger Nps show good results, their daily handling is more difficult in terms of preparation of homogeneous dispersion and therefore reproducible in vitro assays. On the other hand, using the smaller one, 15 nm DeAXO@Np used at 2.0-2.5 mg, values in negative controls showed a slight increase to 0.6-0.7%. Therefore, the best balance on sensitivity and specificity was achieved with DeAXO@Np of 30 nm when using a suspension of 2 mg in 20 μL of PBS in a test tube (using 100 mg/mL suspension), showing the highest % of RAST for both positive control pools and values very close to zero for PBS and negative control pools (Table 1).Table 1Comparison of RAST results, using different size and concentration of DeAXO@Np, performed with five sera samples containing different levels of sIgE. The selected optimal condition is highlighted.Table 1Sample conditionDeAXO@NpNp Φ (nm)200503015Amount (mg)1.52.02.51.52.02.51.52.02.51.52.02.5Levels of sIgE or tIgEPBS0.500.10.100.10.1000.50.10.25Positive controlsaHIGH sIgE8.616.815.912.419.919.213.720.320.714.720.920.9LOW sIgE4.89.07.86.59.810.97.111.412.58.212.812.8Negative controlsbNo sIgE000000000000No sIgE and HIGH tIgE0.10.20.10.40.20.10.20.10.200.60.7aPools of sera from allergic patients to AX (positive controls, with high and low levels of sIgE).bPools of sera from tolerant subjects (negative controls, with standard and high levels of tIgE). Table 1Comparison of RAST results, using different size and concentration of DeAXO@Np, performed with five sera samples containing different levels of sIgE. The selected optimal condition is highlighted.Table 1Sample conditionDeAXO@NpNp Φ (nm)200503015Amount (mg)1.52.02.51.52.02.51.52.02.51.52.02.5Levels of sIgE or tIgEPBS0.500.10.100.10.1000.50.10.25Positive controlsaHIGH sIgE8.616.815.912.419.919.213.720.320.714.720.920.9LOW sIgE4.89.07.86.59.810.97.111.412.58.212.812.8Negative controlsbNo sIgE000000000000No sIgE and HIGH tIgE0.10.20.10.40.20.10.20.10.200.60.7aPools of sera from allergic patients to AX (positive controls, with high and low levels of sIgE).bPools of sera from tolerant subjects (negative controls, with standard and high levels of tIgE). Comparison of RAST results, using different size and concentration of DeAXO@Np, performed with five sera samples containing different levels of sIgE. The selected optimal condition is highlighted. Pools of sera from allergic patients to AX (positive controls, with high and low levels of sIgE). Pools of sera from tolerant subjects (negative controls, with standard and high levels of tIgE). The superior performance of 30 nm DeAXO@Np can be attributed to a balance between surface area, DeAn density, and colloidal stability. While smaller Nps increase surface area, no significant gain in DeAn functionalization was observed below 30 nm, and a slight increase in non-specific signals was detected. In contrast, larger Nps provide lower surface area and reduced DeAn accessibility. The 30 nm architecture likely favors a more efficient multivalent presentation of the dendrimeric epitopes and improves their accessibility for low-abundance BL-sIgE recognition. Therefore, 30 nm represents an optimal compromise between sensitivity and specificity. These assays demonstrate low inter-assay variation coefficients of 5% or 6% for DeAXO@Np and DeBPO@Np, respectively, suggesting a high precision of the assay (see ESI). The implementation of magnetic Nps (DeAn@Np) introduces a highly efficient diagnostic workflow by substantially reducing nanoparticle collection time, a factor that remains independent of operator skill. As depicted in the schematic comparison, conventional non-magnetic silica Nps require 10 min of centrifugation per cycle to ensure they are properly pelleted for supernatant removal. This estimation accounts only for the physical collection time, excluding the additional, variable time required to transfer tubes into centrifuge racks, which depends on sample number and equipment capacity. With the RAST protocol demanding three washing cycles after both the serum incubation and the 125I-anti-IgE incubation stages, the total processing time for non-magnetic Nps accumulates to 60 min. Conversely, the magnetic system facilitates separation via a magnet in just 1 min per cycle, slashing the total collection time to 6 min and achieving at least a 10-fold reduction in overall washing time. Beyond the quantified speed, this approach offers superior clinical practicality, as it eliminates the need for centrifugation equipment by allowing the operator to simply engage or disengage a magnet from the existing support rack, further streamlining the process for high-throughput clinical batches. (Fig. 3).Fig. 3Schematic representation of the washing process using magnetic and non-magnetic Nps. The whole assay involves 2 washing steps: 1st after adding and incubating with the patient serum, the Np is washed three times to remove unbound antibodies. A second, separate washing step with three repetitions is performed after incubating with the secondary, 125I-anti-IgE to wash away any unbound labelled antibody. In table, it is included a time calculation of the processing a clinical batch of samples focussing on those steps independent to operators (washings using centrifugation vs magnets). Schematic representation of the washing process using magnetic and non-magnetic Nps. The whole assay involves 2 washing steps: 1st after adding and incubating with the patient serum, the Np is washed three times to remove unbound antibodies. A second, separate washing step with three repetitions is performed after incubating with the secondary, 125I-anti-IgE to wash away any unbound labelled antibody. In table, it is included a time calculation of the processing a clinical batch of samples focussing on those steps independent to operators (washings using centrifugation vs magnets). DeAXO@Np and DeBPO@Np enhances the sensitivity and specificity in comparison to conventional solid-phase immunoassays Once the optimum Np size (30 nm), concentration (100 mg of Np/mL), and protocol conditions were established, the capacity of DeAn@Np (DeAXO@Np and DeBPO@Np) to quantify serum BL-sIgE antibodies was evaluated in individual serum in a case-control study. Moreover, the sensitivity and specificity of DeAn@Np for detecting BL-specific IgE were compared to those of other solid phases, BL-PLL or ImmunoCAP. Fifty patients with a confirmed allergic reaction to AX and 44 BL-tolerant subjects were included (Table 2 and Table S7, respectively). The mean age of the 50 AX-allergic patients was 45.48 years and 70% were female; 21 (42%) showed urticaria (Grade I), 27 (54%) anaphylaxis (Grade II) and 2 (4%) anaphylactic shock (Grade III). Regarding the clinical response to BLs in STs and/or DPT, 20 patients showed a selective response to AX, not showing any allergic reactions to other BLs like BP or Penicillin V (PV) and 8 patients confirmed as cross-reactors since they also react to PG and/or PV determinants. Twenty-two patients could not be classified in their clinical response due to the severity of the reactions and contraindication of DPT performance. The median and interquartile range of time interval of sample collection after the index reaction was 285 IR: 150-810 days.Table 2Demographic clinical data and in vitro results of allergic patients.Table 2PatientAgeSexAX Selective/penicillin cross-reactor (CR)Grade of severityReactionINT (days)Skin testDPTDeAn@NpPLL-discImmunoCAPAXBPAXBPDeAXDeBPOAXBPAXBPP157FAXIIANA270+-ND-33.70.826.117.23.724.29P257FAXIIANA150+-ND-13.10.83.020.590.260.07P330MAXIIANA180+-ND-28.602.663.030.190.23P458FAXIURT/ANG8280+-ND-110.55.962.930.10.01P566FAXIIANA150+-ND-23.4018.700.180.08P630MAXIURT/ANG300+-ND-21.20.92.340.550.190.12P733FAXIIANA300--+-12.20.90.240.530.030.01P842FAXIURT/ANG3960--+-2801.770.830.020.02P933FAXIIANA5010--+-26.40.93.092.040.010.01P1055MAXIURT1440--+-23.200.6400.030.01P1167FAXIURT/ANG81+-ND-6.080.065.712.3300P1269FAXIURT/ANG394+-ND-5.261.322.7000P1321FAXIURT/ANG216+-ND-1.6602.941.1500P1427FAXIURT/ANG234+-ND-3.66.443.630.250.730P1541FAXIIANA99+-ND-2.20.92.20.6200P1668MAXIIANA62+-ND-4.180.16.290.7500P1756FAXIURT/ANG681+-ND-5.703.234.52NDNDP1849FAXIIANA144+-ND-402.69000P1967FAXIIANA660+-ND-39.42.936.612NDNDP2058FAXIIANA455+-ND-3.4206.661.1500P2134MCRIIANA150--ND+27.78.30.171.390.070.04P2264FCRIURT/ANG90++NDND27.315.12.563.970.760.34P2346MCRIURT/ANG300++NDND35.41951.137.20.740.75P2416FCRIIANA330-+NDND145.9000.050.005P2555FCRIIANA90++NDND176.25.881.180.830.27P2658MCRIURT/ANG5790++NDND17.28.42.652.520.120.46P2740FCRIIANA690--ND+30.60.50.820.712.691.26P2855FCRIIANA90++NDND20.80.4000.050.03P2959MUndIIANA90+-NDND22.98.47.010.440.250.36P3031FUndIURT/ANG2130--+ND22.73.42.081.370.130.02P3147FUndIIANA13,200--+ND16.47.41.80.450.020.01P3238FUndIIANA60+-NDND282.35.220.940.670.09P3344FUndIIANA210+-NDND18.80.071.210.260.570.24P3443FUndIANG330+-NDND20.903.270.590.030.002P3545FUndIIANA300+-NDND10.901.54−0.080.090.03P3619FUndIIANA180+-NDND15.502.330.780.040.002P3753MUndIURT/ANG2130+-NDND11.102.360.210.060.05P3820FUndIIISHOCK90+-NDND23.91.16.681.450.290.01P3961MUndIURT/ANG210+-NDND27.30.311.21.250.110.02P4031MUndIIANA120+-NDND27.70.84.161.070.040.01P4120FUndIURT5220--+ND26.32.50.3400.040.05P4248FUndIIANA1170--+ND16.85.31.0300.10.004P4315MUndIIANA5160--+ND26.81.40.930.140.040.02P4441FUndIIANA180--+ND21.12.80.500.020.01P4559FUndIURT/ANG12,090--+ND24.45.51.510.070.060.05P4658FUndIIISHOCK180--+ND28.26.33.960.470.090.01P4753MUndIURT/ANGUNK--+ND17.21.21.51.250.030.01P4853MUndIURT/ANG210NDND+ND13.90.7000.180.09P4947MUndIURT/ANG365--+ND220.60.8500.030.02P5037FUndIIANAUNK--+ND32.33.21.230.530.020.02F: female; M: male; AX: amoxicillin; BP: benzylpenicillin; Penicillin CR: penicillin cross-reactor; Und: undetermined; URT: urticaria; ANG: angioedema; ANA: anaphylaxis; SHOCK: anaphylactic shock; PLL: poly-L-lysine; DPT: drug provocation test; ND: not done; INT: interval of time between reported reaction and extraction for assay. Positive results are represented in bold. Table 2Demographic clinical data and in vitro results of allergic patients.Table 2PatientAgeSexAX Selective/penicillin cross-reactor (CR)Grade of severityReactionINT (days)Skin testDPTDeAn@NpPLL-discImmunoCAPAXBPAXBPDeAXDeBPOAXBPAXBPP157FAXIIANA270+-ND-33.70.826.117.23.724.29P257FAXIIANA150+-ND-13.10.83.020.590.260.07P330MAXIIANA180+-ND-28.602.663.030.190.23P458FAXIURT/ANG8280+-ND-110.55.962.930.10.01P566FAXIIANA150+-ND-23.4018.700.180.08P630MAXIURT/ANG300+-ND-21.20.92.340.550.190.12P733FAXIIANA300--+-12.20.90.240.530.030.01P842FAXIURT/ANG3960--+-2801.770.830.020.02P933FAXIIANA5010--+-26.40.93.092.040.010.01P1055MAXIURT1440--+-23.200.6400.030.01P1167FAXIURT/ANG81+-ND-6.080.065.712.3300P1269FAXIURT/ANG394+-ND-5.261.322.7000P1321FAXIURT/ANG216+-ND-1.6602.941.1500P1427FAXIURT/ANG234+-ND-3.66.443.630.250.730P1541FAXIIANA99+-ND-2.20.92.20.6200P1668MAXIIANA62+-ND-4.180.16.290.7500P1756FAXIURT/ANG681+-ND-5.703.234.52NDNDP1849FAXIIANA144+-ND-402.69000P1967FAXIIANA660+-ND-39.42.936.612NDNDP2058FAXIIANA455+-ND-3.4206.661.1500P2134MCRIIANA150--ND+27.78.30.171.390.070.04P2264FCRIURT/ANG90++NDND27.315.12.563.970.760.34P2346MCRIURT/ANG300++NDND35.41951.137.20.740.75P2416FCRIIANA330-+NDND145.9000.050.005P2555FCRIIANA90++NDND176.25.881.180.830.27P2658MCRIURT/ANG5790++NDND17.28.42.652.520.120.46P2740FCRIIANA690--ND+30.60.50.820.712.691.26P2855FCRIIANA90++NDND20.80.4000.050.03P2959MUndIIANA90+-NDND22.98.47.010.440.250.36P3031FUndIURT/ANG2130--+ND22.73.42.081.370.130.02P3147FUndIIANA13,200--+ND16.47.41.80.450.020.01P3238FUndIIANA60+-NDND282.35.220.940.670.09P3344FUndIIANA210+-NDND18.80.071.210.260.570.24P3443FUndIANG330+-NDND20.903.270.590.030.002P3545FUndIIANA300+-NDND10.901.54−0.080.090.03P3619FUndIIANA180+-NDND15.502.330.780.040.002P3753MUndIURT/ANG2130+-NDND11.102.360.210.060.05P3820FUndIIISHOCK90+-NDND23.91.16.681.450.290.01P3961MUndIURT/ANG210+-NDND27.30.311.21.250.110.02P4031MUndIIANA120+-NDND27.70.84.161.070.040.01P4120FUndIURT5220--+ND26.32.50.3400.040.05P4248FUndIIANA1170--+ND16.85.31.0300.10.004P4315MUndIIANA5160--+ND26.81.40.930.140.040.02P4441FUndIIANA180--+ND21.12.80.500.020.01P4559FUndIURT/ANG12,090--+ND24.45.51.510.070.060.05P4658FUndIIISHOCK180--+ND28.26.33.960.470.090.01P4753MUndIURT/ANGUNK--+ND17.21.21.51.250.030.01P4853MUndIURT/ANG210NDND+ND13.90.7000.180.09P4947MUndIURT/ANG365--+ND220.60.8500.030.02P5037FUndIIANAUNK--+ND32.33.21.230.530.020.02F: female; M: male; AX: amoxicillin; BP: benzylpenicillin; Penicillin CR: penicillin cross-reactor; Und: undetermined; URT: urticaria; ANG: angioedema; ANA: anaphylaxis; SHOCK: anaphylactic shock; PLL: poly-L-lysine; DPT: drug provocation test; ND: not done; INT: interval of time between reported reaction and extraction for assay. Positive results are represented in bold. Demographic clinical data and in vitro results of allergic patients. F: female; M: male; AX: amoxicillin; BP: benzylpenicillin; Penicillin CR: penicillin cross-reactor; Und: undetermined; URT: urticaria; ANG: angioedema; ANA: anaphylaxis; SHOCK: anaphylactic shock; PLL: poly-L-lysine; DPT: drug provocation test; ND: not done; INT: interval of time between reported reaction and extraction for assay. Positive results are represented in bold. These results were compared with those obtained with the cellulose BL-PLL-discs (AXO-PLL-disc and BPO-PLL-disc) used in conventional RAST. Results obtained for quantification of AX-sIgE (Fig. 4) showed that both DeAXO@Np and AXO-PLL-disc were able to discriminate between patients and controls with a statistically significant difference (p < 0.0001 and p = 0.0097, respectively). A comparison of the two solid phases revealed significant higher levels of AX-sIgE detection when using DeAXO@Np in allergic patients whereas lower in controls (p < 0.0001).Fig. 4Percentage of Radio Allergo Sorbent Test (RAST) for DeAXO@Np and AXO-PLL-disc in allergic to AX patients (n = 50) and tolerant subjects (n = 44). Statistical significance was determined using a non-parametric Mann-Whitney U test for unpaired data and Wilcoxon test for paired data. Percentage of Radio Allergo Sorbent Test (RAST) for DeAXO@Np and AXO-PLL-disc in allergic to AX patients (n = 50) and tolerant subjects (n = 44). Statistical significance was determined using a non-parametric Mann-Whitney U test for unpaired data and Wilcoxon test for paired data. To analyze positivity of the test using different solid phases DeAXO@Np and AXO-PLL-disc, we first established the cut-off points for considering positive results taking into account the best sensitivity/specificity balance using ROC curves analysis. For DeAXO@Np, the established cut-off was ≥1.93, offering an area under the curve (AUC) of 1.00, with 98% of sensitivity, 100% of specificity, 100% of positive predictive value (PPV), 97.8% of negative predictive value (NPV) and statistically significant differences (p < 0.0001) (Fig. 5A and B and Table 3). For AXO-PLL-disc, the established cut-off was ≥3.08, offering an AUC of 0.65, 38% of sensitivity, 81.8% of specificity, 70.4% of PPV, 53.7% of NPV and statistically significant differences (p < 0.01). ImmunoCAP cutoff is established at 0.35 kUA/L as recommended by manufacturer and studies [15]. The sensitivity of ImmunoCAP was 17%, with a specificity of 100%, a PPV of 100%, and a NPV of 51.3% (Fig. 5B and Table 3).Fig. 5A) ROC curve of AXO-PLL-disc and DeAXO@Np in AX-allergic patients (n = 50) and tolerant controls (n = 44). B) Comparison of the sensitivity and specificity of immunoassays for the determination of AX-sIgE. C) ROC curve of BPO-PLL-disc and DeBPO@Np in penicillin cross reactive patients (n = 8, BP-AX-allergic) and tolerant subjects (n = 44). D) Comparison of the sensitivity and specificity of immunoassays for the determination of BP-sIgE. Statistical significance was determined using Cochran's Q test with subsequent pairwise comparisons using McNemar's test with Holm-Bonferroni correction. ∗p > 0.05; ∗∗∗∗p > 0.0001.Table 3Calculated diagnostic values of the different immunoassays approaches to AX and BP.Table 3AUCCut-offSensitivitySpecificityPPVNPVp-valueAX ImmunoassayAXO-PLL-disc0.653.0838%81.8%70.4%53.7%∗DeAXO@Np1.001.9398%100%100%97.8%∗∗∗∗ImmunoCAP-0.3517%100%100%51.3%-BP ImmunoassayBPO-PLL-disc0.612.450%88.6%37.5%83%n.s.DeBPO@Np0.863.875%100%100%95.65%∗∗∗ImmunoCAP-0.3537.5%100%100%92.6%-∗p < 0.05; ∗∗∗p < 0.001 ∗∗∗∗p < 0.0001; n.s.: not significant; AX: amoxicillin; BP: benzylpenicillin; AUC: area under curve; PPV: positive predictive value; NPV: negative predictive value. A) ROC curve of AXO-PLL-disc and DeAXO@Np in AX-allergic patients (n = 50) and tolerant controls (n = 44). B) Comparison of the sensitivity and specificity of immunoassays for the determination of AX-sIgE. C) ROC curve of BPO-PLL-disc and DeBPO@Np in penicillin cross reactive patients (n = 8, BP-AX-allergic) and tolerant subjects (n = 44). D) Comparison of the sensitivity and specificity of immunoassays for the determination of BP-sIgE. Statistical significance was determined using Cochran's Q test with subsequent pairwise comparisons using McNemar's test with Holm-Bonferroni correction. ∗p > 0.05; ∗∗∗∗p > 0.0001. Table 3Calculated diagnostic values of the different immunoassays approaches to AX and BP.Table 3AUCCut-offSensitivitySpecificityPPVNPVp-valueAX ImmunoassayAXO-PLL-disc0.653.0838%81.8%70.4%53.7%∗DeAXO@Np1.001.9398%100%100%97.8%∗∗∗∗ImmunoCAP-0.3517%100%100%51.3%-BP ImmunoassayBPO-PLL-disc0.612.450%88.6%37.5%83%n.s.DeBPO@Np0.863.875%100%100%95.65%∗∗∗ImmunoCAP-0.3537.5%100%100%92.6%-∗p < 0.05; ∗∗∗p < 0.001 ∗∗∗∗p < 0.0001; n.s.: not significant; AX: amoxicillin; BP: benzylpenicillin; AUC: area under curve; PPV: positive predictive value; NPV: negative predictive value. Calculated diagnostic values of the different immunoassays approaches to AX and BP. ∗p < 0.05; ∗∗∗p < 0.001 ∗∗∗∗p < 0.0001; n.s.: not significant; AX: amoxicillin; BP: benzylpenicillin; AUC: area under curve; PPV: positive predictive value; NPV: negative predictive value. Comparison of the results based on Cochran's Q test among the three immunoassay approaches revealed statistically significant differences for specificity and sensitivity between the three techniques (p < 0.0001). AXO-PLL-disc showed a significantly lower specificity (81.8%) compared to ImmunoCAP and DeAXO@Np with a 100% (p = 0.04) for both (Fig. 5B). Regarding sensitivity, DeAXO@Np offered the highest sensitivity (98%) compared to AXO-PLL-disc and ImmunoCAP (p < 0.0001 for both). In addition, the sensitivity of AXO-PLL-disc was significantly higher compared to ImmunoCAP (p = 0.015) (Fig. 5B and Table 3). These data comparing DeAXO-SiO2 Np and AXO-PLL-disc were in line with those described previously [34]. In this mentioned manuscript, comparisons with ImmunoCAP were not performed since it was a patient's selection criterion. For a further analysis relative to detection of BP-sIgE, allergic patients were clinically classified as selective to AX (n = 20), if they presented positive in vivo tests to AX and negative to BP; and cross-reactor to penicillin (n = 8) if they were positive to BP in any in vivo tests. The rest of patients allergic to AX (n = 22) were not able to be tested for allergy to BP in DPT so were not classified, described in Table 1 as undetermined. ROC curves including cross-reactor patients were performed for DeBPO@Np and BPO-PLL-disc to establish a cutoff. Cut-off was established for DeBPO@Np at ≥ 3.8, offering an AUC of 0.86, 75% of sensitivity, 100% of specificity, 100% of PPV, and 95.65% of NPV, with statistically significant differences (p = 0.0002) (Fig. 5C and. D and Table 3). Cut-off was selected for BPO-PLL-disc at ≥ 2.4 with an AUC of 0.61, 50% of sensitivity, 88.6% of specificity, 37.5% of PPV and 83% of NPV, with no significant differences. The ImmunoCAP cut-off established by manufacturer at 0.35 kUA/L for BP-sIgE showed a sensitivity of 37.5%, a specificity of 100%, a PPV of 100% and a NPV of 92.6% (Fig. 5D and Table 3). Therefore, DeBPO@Np showed the highest value of sensitivity (75%) and specificity (100%), followed by ImmunoCAP with 37.5% of sensitivity and 100% of specificity, and, lastly, BPO-PLL-disc with 37.5% of sensitivity and 88.6% of specificity. No statistically significant differences were reported between immunoassays (Fig. 5D). DeAXO@Np and DeBPO@Np enhance the diagnostic accuracy and severity in comparison to conventional solid-phase immunoassays We afterwards analyzed how the results obtained with all solid phases correspond to the clinical response of the patients classified as selective and cross-reactors. Related to level of RAST with DeAXO@Np, results of selective reactors and cross-reactors differed from those of controls (p < 0.0001 for both). In terms of positivity, according to the chosen cut-off, we observed positive RAST values in 95% of selective reactors, 100% of cross-reactors and 0% of tolerant controls, with significant differences between selective reactors as well as cross-reactors versus controls (p < 0.0001 for both). In contrast, AXO-PLL-disc only revealed significant differences between selective reactors and controls in both %RAST and % of positivity (p = 0.0017 and p = 0.01, respectively) (Fig. 6). When using ImmunoCAP, we have observed significant differences in AX-sIgE levels between selective reactors as well as cross-reactors against controls (p < 0.0001 for both). However, when analyzing the results in terms of positivity, we detected a significantly higher percentage in cross-reactors compared to both selective patients and controls (p = 0.038 and p = 0.0003, respectively).Fig. 6Results (sIgE levels (upper line), and positivity (lower line)) of the three immunoassay approaches of A) DeAXO@Np, AXO-PLL-disc and AX-ImmunoCAP in controls (n = 44), selective AX-allergic patients (n = 20) and cross-reactors to penicillin (n = 8). B) DeBPO@Np, BPO-PLL-disc and BP-ImmunoCAP in controls (n = 44), selective AX-allergic patients (n = 20) and cross-reactors to penicillin (n = 8). Statistical significance for non-parametric quantitative data was determined using the Kruskal-Wallis test with subsequent pairwise comparisons via Dunn's post-hoc test adjustment. Categorical results were analyzed using Fisher's exact test. Results (sIgE levels (upper line), and positivity (lower line)) of the three immunoassay approaches of A) DeAXO@Np, AXO-PLL-disc and AX-ImmunoCAP in controls (n = 44), selective AX-allergic patients (n = 20) and cross-reactors to penicillin (n = 8). B) DeBPO@Np, BPO-PLL-disc and BP-ImmunoCAP in controls (n = 44), selective AX-allergic patients (n = 20) and cross-reactors to penicillin (n = 8). Statistical significance for non-parametric quantitative data was determined using the Kruskal-Wallis test with subsequent pairwise comparisons via Dunn's post-hoc test adjustment. Categorical results were analyzed using Fisher's exact test. When BPO was evaluated, %RAST with DeBPO@Np immunoassay showed higher levels in cross-reactors compared to controls and selective AX-patients (p = 0.0016 and p = 0.0092, respectively), with no differences between controls and selective patients (as expected). Interestingly, BPO-PLL-disc was unable to discriminate between cross-reactors and the other two groups (controls and/or selective). Regarding ImmunoCAP results we observed significant higher levels of BP-sIgE in both cross-reactors and selective reactors compared to controls (p < 0.0001 for both). These results indicate a lower specificity when using this solid phase as higher levels were detected in selective reactors to AX compared to controls. In terms of positivity, cross-reactors showed a sensitivity of 75% for DeBPO@Np, whereas in controls and selective reactors, it was of 0% and 5%, respectively. There was a significant increase in cross-reactors compared to selective patients to AX and controls (p = 0.0005 and p < 0.0001, respectively). Conversely, results with BPO-PLL-disc showed a 37.5% of positivity in cross-reactors, 11.4% in controls and 25% in selective reactors, with no significant differences between groups, indicating a lower specificity (Fig. 6B). Regarding ImmunoCAP, 37.5% of positivity was obtained in cross-reactors, 0% in controls and 10% in selective reactors, with a significantly higher percentage in cross-reactors compared to controls (p = 0.0025), but not to selective responders, as it would have been expected (Fig. 6B). All these data indicate that compared to DeAXO@Np, the sensitivity was lower in DeBPO@Np, 100% vs 75%, which could be due to the fact that the primary drug involved in the reaction was AX and the BP determinants were clinically detected due to cross reactivity. Remarkably, whereas DeBPO@Np results agree with the patients’ clinical response to BP (cross-reactors), with a specificity of 95%, BPO-PLL-disc produce significant cases of false positive results, since 25% of AX-selective patients (tolerant to BP) showed positive results with this solid phase. Moreover, immunoCAP also exhibited false positive results (10%) in AX-tolerant patients, which is in line with previous studies [15,16]. The results of DeBPO@Np demonstrate improved agreement with the clinical classification compared to BPO-PLL-disc and ImmunoCAP (with which, a percentage of cases would be misdiagnosed due to both false negative and false positive cases). These results are very important for detecting safe BL alternatives (BP in this case), in patients with confirmed reactions to AX, which could be of utmost importance for patient management. The enhanced sensitivity observed for DeAn@Np may result from the combination of several complementary features of the platform. The homogeneous distribution of the functionalized Nps throughout the sample volume increases the probability of interaction between exposed BL determinants and low-abundance BL-sIgE antibodies. In contrast, conventional immunoassays employ larger solid supports on which the determinants are immobilized, a configuration that may introduce local steric constraints and potentially limit access to a fraction of the available binding sites. In addition, the high surface-area-to-volume ratio of the Nps provides a high density of accessible binding sites, while the dendrimeric architecture enables a multivalent and spatially organized presentation of the determinants. Together, these characteristics may improve antibody capture efficiency compared with conventional solid-phase supports. We are aware that this work presents several limitations that should be addressed in further studies. First, although the use of radiolabeled anti-IgE enabled highly sensitive detection, the use of radioactivity limits the applicability of this format in routine laboratory settings. Importantly, this limitation is related to the detection system employed in the present study rather than to the DeAn@Np solid phase itself. In this context, the DeAn@Np platform is compatible with alternative non-radioactive readout strategies, including amplified enzymatic, fluorescence, or chemiluminescence-based immunoassays. Future work will therefore focus on adapting the current RadioImmunoAssay (RIA) based system into more accessible amplified immunoassay formats compatible with standard laboratory practice while maintaining high analytical sensitivity for BL-sIgE detection. Second, this study was designed as a preliminary case-control analysis to evaluate the capacity of DeAn@Nps as solid phases for BL-sIgE detection compared to conventional solid phases such as cellulose discs or ImmunoCAP. Therefore, additional studies addressing large-scale reproducibility (intra- and inter-assay variability), long-term storage stability, and assay standardization will be necessary for technical validation. This must be followed by evaluation in prospective cohorts for clinical validation following the regulatory considerations for implementation in hospital laboratories. Conclusions Herein, core–shell Np consisting of a magnetic core coated with silica and functionalized with DeAn conjugates (DeAn@Np) were successfully developed and evaluated as solid phases for BL allergy diagnosis. The magnetic configuration maintained the favorable physicochemical properties and chemical reproducibility that had been previously reported with silica Np while introducing significant practical advantages for future clinical implementation. Magnetic separation eliminated the need for centrifugation steps, substantially reducing washing time and improving handling reproducibility. Among the Np sizes evaluated (from 200 to 15 nm), 30 nm provided the most favorable balance between sensitivity and specificity, enabling efficient antigen presentation and stable dispersion. When evaluated in serum from patients with confirmed BL allergies, the magnetic Np demonstrated better outcomes in terms of positivity minimizing false-negative results compared to both conventional RAST and the commercial ImmunoCAP. Furthermore, the platform discriminated between selective and cross-reactive patients, reducing false-positive results associated with conventional solid phases. Overall, the modular detection strategy of DeAn@Np represents a promising strategy for improving BL-sIgE detection, while offering practical advantages in assay handling and workflow efficiency. Although additional technical and clinical validation studies will be required, the platform shows potential for adaptation to non-radioactive immunoassay formats and could be extended to drug allergies and immune-mediated diseases. Materials and methods Materials All chemicals were purchased from Merck, Avantor or Thermo Scientific Chemicals, and used as received unless stated. PAMAM dendrimers were purchased from Merck. Milli-Q™ water was obtained by using the ultrapure, Millipore® Direct-Q® 3. Phosphate buffer saline (PBS) and 4M Lithium acetate buffer pH 5.3 was prepared as described [26]. [43]. Np preparation Iron oxide Np synthesis (Fe3O4·Np) In all cases, Nps were synthetized in a 250 mL autoclave under an air atmosphere. Following the heating step, the system was allowed to cool naturally to room temperature. The resulting Nps were isolated from the reaction medium using a neodymium magnet. Np of 200 nm diameter: Ethylene glycol (EG, 80 mL) and iron chloride hexahydrate (FeCl3·6H2O, 2.7 g, 10 mmol) were placed in an autoclave and stirred for 15 min. After the addition of sodium acetate (NaOAc, 7.2 g, 87.7 mmol) and polyethylene glycol (PEG, 2.0 g, 0.5 mmol) the autoclave was closed and heated at 200°C for 16 h. The obtained Nps were washed with water (x 4 times) and 800 mg of 200 nm-Fe3O4 were recovered. Np of 50 nm diameter: EG (80 mL) and FeCl3·6H2O (1.1 g, 3.9 mmol) were placed in an autoclave and were stirred for 15 min. After the addition of NaOAc (5.2 g, 64.0 mmol) and Milli-Q™ water (5.4 g, 300.0 mmol) the autoclave was closed and heated at 200°C for 16 h. The obtained Nps were washed with water (x 4 times) and 600 mg of 50 nm-Fe3O4 were recovered. Np of 30 nm diameter: EG (80 mL) and FeCl3·6H2O (0.8 g, 2.9 mmol) were placed in an autoclave and were stirred for 15 min. After the addition of NaOAc (5.2 g, 64.0 mmol) and Milli-Q™ water (5.4 g, 300.0 mmol) the autoclave was closed and heated at 200°C for 16 h. The obtained Nps were washed with water (x 4 times), and 500 mg of 30 nm-Fe3O4 were recovered. Np of 15 nm diameter: EG (80 mL) and FeCl3·6H2O (0.5 g, 2.0 mmol) were placed in an autoclave and were stirred for 15 min. After the addition of NaOAc (5.2 g, 64.0 mmol) and Milli-Q™ water (5.4 g, 300.0 mmol) the autoclave was closed and heated at 200°C for 16 h. The obtained Nps were washed with water (x 4 times) and 300 mg of 15 nm-Fe3O4 were recovered. Silica coating Fe3O4 (0.5 g) were suspended in an ethanol/water/ammonium hydroxide mixture (45/15/1.25 mL respectively) and sonicated for 5 min. Then, a solution of TEOS (0.5 mL, 2.5 mmol) in ethanol (15 mL) were added dropwise, and the reaction was stirred for 16 h at room temperature. The obtained Nps were washed with water (x 4 times) and 400 mg of Fe3O4@SiO2 were recovered. Silica Np synthesis (SiO2·Np) SiO2 Np were prepared using previously stablished protocols [34], where ammonium hydroxide (5 mL) and TEOS (2.4 mL) where reacted in ethanol (110 mL) for 2 h. Np surface chemical functionalization Amino modification Fe3O4@SiO2 were suspended in toluene (350 mg Np/20 mL) and sonicated for 15 min. Then, APTES (1.75 mL, 7.4 mmol) was added and mixture was sonicated for 1 h. Then an additional amount of APTES was added (1.00 mL, 4.2 mmol) and the dispersion was refluxed under stirring for 16 h. The obtained particles were washed with toluene (x 3 times) and Milli-Q™ water (x 3 times) to obtain the desired amino functionalized particles (NH2@Np). Carboxy modification NH2@Np were suspended in DMF (300 mg Nps/10 mL) and sonicated for 15 min. Then, a solution of succinic anhydride (660 mg, 6.6 mmol) in DMF (10 mL) was added to the previous mixture and sonicated for an additional 15 min. Triethylamine (TEA, 96 μL, 0.84 mmol) was then added and the solution stirred at room temperature for 16 h. The obtained particles were washed with DMF (x 3 times) and Milli-Q™ water (x 3 times) to obtain the desired carboxy functionalized particles (CO2H@Np). Dendrimer modification 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDCI, 86.2 mg, 0.44 mmol) and N-Hydroxysuccinimide (NHS, 100 mg, 0.87 mmol) were added to a suspension of CO2H@Np in anhydrous DMF (260 mg Nps/4 mL), and the resulting mixture was sonicated for 15 min. Then, PAMAM-G2 dendrimer (42.8 mg, 1.3·10−2 mmol) in anhydrous DMF (4 mL) was added to the previous mixture and sonicated for additional 15 min after which the solution was stirred at room temperature for 16 h. The obtained particles were washed with DMF (x 3 times) and Milli-Q™ water (x 3 times) to obtain the desired dendrimer functionalized particles (De@Np). The selected conditions were based on previously optimized and validated protocols, in which no further increase in dendrimer conjugation was observed upon increasing PAMAM equivalents, indicating plateau conditions and near-maximal functionalization [34,40,44]. β-Lactam (BL) modification De@Np were suspended in a freshly prepared solution of BL (10 mg/mL) in 0.05 M Na2CO3/NaHCO3 aqueous buffer at pH 10.2 (4 mL/150 mg of Np). The suspension was then sonicated for 15 min and stirred at 4°C for 7 days. During this period, 10 mg of BL were added at approximately 24 h intervals. The obtained particles were washed with PBS (x 3 times) and Milli-Q™ water (x 3 times) and finally lyophilized to obtain the desired functionalized particles (DeAXO@Np or DeBPO@Np for AX or BP respectively). The selected incubation time and conditions were based on previously optimized and validated protocols, in which no further increase in conjugation was observed upon prolonged incubation, indicating plateau conditions and near-maximal functionalization [27,40]. Np storage The functionalized nanoparticles (DeAn@Nps) were stored as a dry solid at −18 to −20°C. For the assay, nanoparticle dispersions were freshly prepared before use. Under these conditions, no aggregation was observed upon redispersion, indicating preservation of surface functionalization during storage. Freshly prepared dispersions exhibited good colloidal stability at the working concentrations used, ensuring reliable performance. No visible aggregation was observed for up to 6 months of storage under these conditions. Np characterization Fourier transformed infrared (FTIR) A Nicolet Nexus spectrometer (Thermo Fisher Scientific) equipped with a Smart Golden Gate attenuated total reflectance (ATR) accessory was used. Measurements were carried out with 5 mg of the corresponding particles. Transmission electron microscopy (TEM) TEM images were recorded in a PHILIPS CM-200 instrument (Ibima-Plataforma BIONAND) and were used to analyze particle size and morphology. Np size has been determined through the diameter measurement of ten randomly selected particles from TEM images. Transmission electron microscopy experiments with energy dispersive X-ray spectroscopy (TEM-EDX) were recorded in a FEI Talos F200X transmission microscope (TEM/STEM) with an acceleration voltage of 200 kV and an EDX Super-X microanalysis system (Central Research Services (SCAI), UMA). In all cases, using a 1 mg Nps/1 mL of Milli-Q™ water suspension. Vibrating Sample Magnetometer (VSM) The magnetization curves of the dried Nps were obtained using a Vibrating Sample Magnetometer (VSM, Lake Shore Cryotronics (Inc. 575 McCorkle Blvd. Westerville, Ohio 43,082-8888 USA); Ibima-Plataforma BIONAND). Measurements were performed at 298 K under an applied magnetic field of 19 kOe using 0.1 mg of Nps (prepared by drying 100 μL of a 1 mg/mL Nps dispersion). All measurements were conducted in duplicate. Saturation magnetization (Ms): Fe3O4: 28 emu/g Nps (value calculated as the average of 27.6 and 27.2); Fe3O4@SiO2: 10 emu/g Nps (value calculated as the average of 9.3 and 10.4) at 19 kOe. The coercivity (Hc) and remanent magnetization (Mr) values were found to be nearly zero in all cases, as expected for a superparamagnetic material. Z potential measurements Z Potential measurements were carried out using a Malvern Zetasizer Nano ZS90 instrument. Measurements were done in duplicate using 1 mg Nps/3 mL of aqueous dispersion samples. For 30 nm Np, measurements of 5 different batches are shown in Table S4. Quantification of free primary amino groups The quantification of free primary amino groups present on the Np surface was carried out following a previously described procedure [28,34]. The ninhydrin test solution was prepared by dissolving ninhydrin (300 mg) and hydrindantin (45 mg) in water-free DMSO (11.25 mL). Then, 4 M Lithium acetate buffer pH 5.3 (3.75 mL) was added, and the solution was deoxygenated. The Np (1 mL of a 3 mg Nps/1 mL of Milli-Q™ water suspension) were placed in a sealed tube, and 1 mL of ninhydrin solution was then added. The solution was heated to 100°C for 15 min. Next, 5 mL of 50% water/EtOH was added. The reaction mixture was filtered, and the absorbance (λ = 570 nm) was measured with a VARIAN CARY 100BIO UV-Visible spectrophotometer. A calibration was performed with aqueous n-butylamine solutions in a concentration range of 100-650 μM. The measurements were carried out in triplicate. For 30 nm Np, measurements of 5 different batches are shown in Table S6. Patients selection This work was designed as a retrospective case-control study. The study group was selected from subjects attending the Allergic Unit at the Hospital Regional Universitario de Málaga (HRUM). Fifty patients with a confirmed diagnosis of immediate hypersensitivity reaction (IHR) to AX after an allergological work-up were included in the study according to European Network of Drug Allergy (ENDA) and European Academy of Allergy and Clinical Immunology (EAACI) [45,46] guidelines [47]. Patients were included after a confirmed diagnosis of immediate allergic reaction to AX when they showed a positive ST to BLs, and in the case of negative STs, a positive DPT to these drugs. Severity of the reaction was evaluated according to Ref. [48] [49]as grade 1 (mild or URT/ANG), grade 2 (moderate or anaphylaxis (ANA), and grade 3 (severe or SHOCK). AX-allergic patients (n = 50) were further classified as selective to AX (n = 20), diagnosed by ST or DPT positive to AX and DPT negative to penicillin V (PV), cross-reactors (n = 8) to penicillins (AX and BP), diagnosed by positive ST and/or DPT to BLs. Forty-four age- and sex-matched subjects with confirmed tolerance to BLs were included as negative controls. Skin test Skin prick test (SPT) and, if negative, intradermal test (IDT) were carried out as previously described [46,48], using daily prepared solutions for AX (Diater S.A, Madrid, Spain) and BP determinants, from DAP Penicillin® test kit (Diater S.A, Madrid, Spain). The maximum concentrations reached were, respectively: (i) AX at 20 mg/mL; (ii) major determinant benzylpenicilloyl-octa-L-lysine (BP-OL) at 0.04 mg/mL, equivalent to an 8.64·10−5 M concentration of the benzylpenicilloyl (BPO) moiety; and (iii) minor determinant (MD) at 0.5 mg/mL, equivalent to a concentration of 1.5·10−3 M of sodium benzylpenilloate. Readings were done after 20 min and considered positive if: (i) in SPT a wheal larger than 3 mm surrounded by erythema appeared, together with a negative response to the control saline; (ii) in IDT, if the diameter of the initially marked wheal area increased more than 3 mm and it was surrounded by erythema. Positive data were expressed as the mean diameter by measuring the largest and the smallest diameters at angles right to each other. Drug provocation test DPT was carried out in a single-blind procedure according to the EAACI guidelines [47,50,51] in consenting patients with negative-ST to AX, PPL/BP-OL and MDM/MD, DPTs with AX and/or PV. Increasing doses of PV (ERN, Barcelona, España) or AX (Normon Laboratories, Madrid, Spain) were given by oral route at incremental dose every 30 min until reaching the total cumulative full dose of 500 mg [52]. In vitro sIgE determination sIgE determinations were performed in vitro with the described customized or commercial materials using sera from allergic and tolerant subjects. ImmunoCAP: To determine serum sIgE to penicilloyl G (c1) and amoxicilloyl (c6) FEIA ImmunoCAP® (Thermo-Fisher) was used following the manufacturer's instructions in patients and controls serum. Results of drug-sIgE were considered positive according to the cutoff ≥0.35 kUA/L. RAST protocol for BL-PLL-dics RAST was carried out adapting conventional methodology for both, using the drugs concerned (BP or AX) conjugated to poly-L-lysine (PLL) over the cellulose disc solid support [53], or the prepared Nps, to determine drug-sIgE in vitro. RAST protocol for PLL-discs: Briefly, 30 μL of subject undiluted sera was added directly onto each disc. For each subject, three type of discs were assayed in parallel: AXO-PLL-disc, BPO-PLL-disc and blank disc (PLL-functionalized disc without drug) in duplicate. Incubation was performed for 3 h. All incubation steps were performed at room temperature under gentle agitation (800 rpm) in a humid chamber. After incubation, discs were washed using PBS containing 0.1% Tween 20, with 3 washing steps of 3 mL each. After washing, 25 μL of 125I-anti-IgE solution (a pool of antihuman IgE antibodies kindly provided by Thermo Fisher Scientific and radiolabeled with iodine-125 in our laboratory according to previously described procedures) [54] was added and incubated overnight. The 125I-anti-IgE solution was used at a final concentration of 15 μg/100 mL, delivering approximately 20,000–30000 counts per minute (cpm) per assay tube. The following day, discs were washed under the same conditions, and radioactivity was measured in a Multi Crystal Gamma Counter LB 2111 (Berthold Technologies, Wildbad, Germany). Each assay was performed in duplicate and average values were used for calculating the % RAST result, as a percentage of the maximum of the count per minute (cpm), according to the following equation.%RAST=cpm(drugsolidphase)−cpm(blanksolidphase)cpmmaximum Results were considered positive over corresponding cutoffs calculated according to Receiver Operating Characteristic (ROC) curves. RAST protocol for DeAn@Nps The nanoparticle-based RAST was adapted from the disc-based assay. Different suspensions of particles of 30 nm in PBS were prepared just before the assay and sonicated: DeAn@Nps at 100 mg/mL. Undiluted serum (30 μL) was added to nanoparticle suspension (20 μL) in duplicate. Mix of nanoparticles and serum was incubated overnight. All incubation steps were performed at room temperature under gentle agitation (800 rpm) in a humid chamber. On the following day, nanoparticle suspension was washed, and 25 μL of 125I-anti-IgE solution [54] was added for an overnight incubation under gentle agitation (800 rpm). The same radiolabeled anti-human IgE preparation described above was used in all experiments. Finally, nanoparticles were washed. Washing steps for magnetic nanoparticles consisted of adding 3 mL of washing buffer (PBS + Tween 20 0.1%), followed by magnetic separation using a neodymium magnet placed at the base of the tube and removal of the supernatant. This washing step was repeated three times after serum incubation and again after radiolabeled anti-IgE incubation. Radioactivity were measured as described above, and %RAST values were calculated using the same formula as for the disc-based assay. Results were considered positive when values were above the corresponding ROC-derived cut-off values. The possible endotoxin contamination effect in the immunoassay was ruled out as no significant non-specific IgE binding was observed in control samples. Statistical analysis Normality was assessed by Kolmogorov-Smirnov test for the percentage of RAST with the different solid phases and as they showed non-normal distribution, the comparisons were performed by the Mann-Whitney U test for unpaired samples (Patients vs controls) and by the Wilcoxon test for pair samples (within patient or control groups) with different solid phases. For comparisons involving more than two groups, such as the classification of selective reactors, cross-reactors, and controls, the Kruskal-Wallis test was employed, followed by Dunn's post-hoc test for subsequent pairwise adjustments. Categorical and qualitative variables were analyzed using Fisher's exact test. To determine the diagnostic performance of the different solid phases, Receiver Operating Characteristic (ROC) curves were generated to establish optimal cut-off points and calculate sensitivity, specificity, area under the curve (AUC), positive predictive value (PPV), and negative predictive value (NPV). Additionally, differences in sensitivity and specificity among the three immunoassay techniques (ImmunoCAP, conventional RAST, and DeAn@Np) were evaluated using Cochran's Q test, with further pairwise comparisons conducted via McNemar's test with Holm-Bonferroni correction. Statistical significance was defined as p < 0.05, and all analyses and graphical representations were performed using GraphPad Prism 7. Ethical statement The study was conducted according to the principles of the Declaration of Helsinki and approved by the Ethics Committee of Malaga (AC19/00082, 26/03/2020). All subjects included were orally informed about the study and signed the corresponding informed consent. CRediT authorship contribution statement Jose A. Céspedes: Data curation, Formal analysis, Writing – original draft. Maria I. Montañez: Data curation, Formal analysis, Investigation, Methodology, Writing – original draft. Isabel M. Jiménez: Data curation, Formal analysis. Violeta Gil-Ocaña: Data curation, Formal analysis. Esther Matamoros: Data curation, Formal analysis, Investigation. María Salas: Data curation, Formal analysis, Validation. Adriana Ariza: Data curation, Formal analysis, Validation. Gádor Bogas: Data curation, Formal analysis, Validation. Cristobalina Mayorga: Conceptualization, Data curation, Formal analysis, Investigation, Writing – original draft, Writing – review & editing. Yolanda Vida: Conceptualization, Formal analysis, Investigation, Supervision, Writing – original draft, Writing – review & editing. Ezequiel Pérez-Inestrosa: Funding acquisition, Project administration, Writing – original draft. Maria J. Torres: Funding acquisition, Project administration, Writing – original draft. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.