Abstract
Background: Penicillin allergy delabeling initiatives are now
part of antibiotic stewardship programs and include the use of invasive
and risky in vivo tests. Instead, the quantification of specific
IgE is highly useful to confirm immediate allergy to penicillins.
However, discrepant results associated to the low sensitivity of thein vitro tests have limited their routine diagnostic use for
delabeling purposes. We aimed to tackle a novel diagnostic strategy for
specific IgE testing based on a homologous interpolation scheme, using
recombinantly produced standards.
Methods : Serum samples from a cohort of allergic patients and
controls were analysed by a chemiluminescence-based immunoassay, using a
bispecific binanobody as standard. The novel standard targets the major
antigenic determinant of penicillin G and the paratope of Omalizumab,
acting as human-like specific IgE.
Results : Testing a cohort of 65 human serum samples, the method
achieved a good agreement and strong positive relationship, reaching a
limit of detection below 0.1 IU/mL. The sensitivity of the in
vitro test significantly increased (66 %), doubling that of the
ImmunoCAP reference in vitro assay with an overall specificity of
100 %.
Conclusions: The new diagnostic strategy compares favourably
with the results obtained on the ImmunoCAP system, paving the way
towards the standardization of penicillin allergy testing. The
recombinant standards are potent calibrators, highly stable, easy and
inexpensive to produce, and overcome the limitation associated with
preparations derived from pooled human serum, expediting the production
of next generation standards with different specificities to
successfully tackle β-lactam allergy delabeling by in vitrotests.
Keywords: allergy, β-lactam antibiotics, binanobody,
diagnostics, Omalizumab.
Introduction
Currently over 31 million of the EU-27 population (6.9 %) carry a
β-lactam allergy label, which prevents them from receiving the first-line
antibiotics such as penicillin. This contributes to the prescription of
antibiotics that diverge from the guidelines, increasing antimicrobial
resistance, one of the biggest threats to global health. Delabeling
initiatives are now part of antibiotic stewardship programs and include
the use of invasive, time-consuming and risky in vivo tests. The
quantification of specific IgE to penicillins is highly valuable for
allergy monitoring, confirming or ruling out immediate allergy and finding
safe alternatives, however, the current in vitro tests lack
sensitivity (ca. 75% false negatives), which limits their routine
diagnostic use for delabeling purposes.
In vitro diagnosis of human allergies is based on the
immunoanalytical quantification of specific IgE (sIgE) in
serum1. Currently, several assays are commercially
available for the determination of sIgE, as ImmunoCAP (Thermo Fisher
Scientific), Immulite and Atellica (Siemens), Euroline (Euroimmun),
Noveos (Hycor), ALEX2 (Macro Array Diagnostics)2 and
ALFA (Dr. Fooke)3. However, scientific evidence has
shown that the results provided by these methods are not comparable with
each other4, so there is an urgent need to harmonize
protocols for the in vitro diagnosis of allergy.
The lack of standardized allergen-specific IgE reference preparations is
one of the weaknesses of in vitro diagnosis5.
The concentrations of sIgE are calculated from a total IgE calibration
curve by a heterologous interpolation. To do this, a set of calibrators
(WHO 11/234) is used to transform the assay signals into quantitative
allergen-specific IgE antibody units (IU/mL)6.
However, recent studies have revealed a better correlation among methods
when homologous calibration is used because of the significant
differences found between the interpolation schemes for quantifying
allergen-specific IgE.7,8 Therefore, the absence of
internationally standardized allergen-specific IgE reference
preparations is one of the reasons for the discrepant results between
methods and probably the most important unresolved issue to harmonizein vitro serological quantitative allergy diagnostics.
The standard for the homologous calibration is the sIgE present in
serum. Serum samples, however, show variability in terms of sIgE level,
the concentration to β-lactam antibiotics being very
low9. Furthermore, the sample volume is limited, its
accessibility difficult and expensive as well. So, standardized
reference material is currently not available. Besides, the use of
positive human sera is key to perform the so-called ”round robin” tests
to evaluate the accuracy and precision (repeatability and
reproducibility) of the in vitro allergy tests between
laboratories.
The first attempt approaching synthetic standards consisted on raising
specific antibody molecules by hyperimmunization to produce a fully
humanized chimera10. Later, these molecules were
produced recombinantly by binding the Fab to the human Fc
region11,12. An advantageous alternative to
conventional antibody-based standards is the use of nanobodies. These
are recombinant antibody fragments (15 kDa) derived from the variable
domain (VHH) of a particular type of antibodies found in camelids that
are devoid of light chain13. They have remarkable
physicochemical properties including thermal stability, solubility and
are produced with high yields by bacterial fermentation. In addition,
their monodomain nature facilitates the construction of phage display
libraries, which offers the possibility of isolating nanobodies with the
desired properties14. Thus, synthetic standards
consisting of a nanobody against the target antigen expressed in tandem
with a second nanobody against the paratope of the detection antibody
can function as highly standardized surrogate homologous calibrators for
the currently existing in vitro tests.
This study reports on the clinical evaluation of the ultra-sensitive
chemiluminescence immunoassay, using synthetic bispecific binanobodies
(rBBN) as the human IgE-like standard for the reliable quantification of
IgE to penicillin G in a cohort of 65 human serum samples. The results
were compared with those obtained with the ImmunoCAP system.
Materials and methods
Reagents, buffers, and consumables
Keyhole limpet hemocyanin (KLH), human serum albumin (HSA), histone H1,
penicillin G (PG), bovine serum albumin (BSA), D-biotin, isopropyl
β-D-1-thio-galactopyranoside (IPTG), polyethylene glycol 8000 (PEG), LB
Broth (Miller), 3,3′,5,5′- tetramethylbenzidine (TMB), tris base, Tween
20 and other chemicals were from Sigma (St. Louis, MO, USA).
Anti-hemagglutinin epitope (anti-HA) antibody conjugated to horseradish
peroxidase (HRP) was from Sigma (Cat no. 12013819001). Omalizumab was
from Novartis International AG (Basel, Switzerland). Anti-human IgG
antibody-HRP (Cat No. ab97225) and the HRP Conjugation Kit (ab102890),
the latter used to prepare Omalizumab-HRP, were from Abcam (Cambridge,
UK).
E. coli One Shot BL21 (DE3) cells were from Thermo Fisher
Scientific (Fremont, CA, USA). E. coli ER2738 electrocompetent
cells were purchased from Lucigen Corporation (Middleton, WI, USA) and
chromatography columns were from GE Healthcare (Piscataway, NJ, USA).
High binding polystyrene ELISA plates were from Costar Corporation
(Cambridge, MA, USA) and SuperSignal ELISA Femto Substrate, PCR Plate
96-well, and SYPR Orange Protein Gel Stain were acquired from Thermo
Fisher (Madrid, Spain). Coating buffer was 50 mM sodium
carbonate/bicarbonate, pH 9.6, and washing buffer 10 mM sodium phosphate
buffer, 150 mM NaCl, 0.05% Tween 20%, pH 7.4 (PBS-T).
Phage Display
An adult llama (Lama glama ) from the municipal zoo of Montevideo
was immunized subcutaneously with 4 doses of 500 μg of HSA-PG conjugate
in Freund’s Incomplete Adjuvant. Ten days after the last immunization,
200 mL of blood was drawn. From the extracted blood, lymphocytes were
purified by gradient centrifugation with Histopaque (Sigma) according to
the manufacturer’s protocol. From 108 cells, total RNA
was extracted using TRIZOL (Invitrogen, Carlsbad, CA USA), and
retrotranscribed to cDNA using the RevertAID Reverse Transcriptase
(Thermo Fisher Scientific, Carlsbad, CA, USA). Subsequently, the DNA
fragments encoding the VH/VHH of the IgGs were amplified by PCR as
described in the supporting information. The fragments obtained were
digested using SfiI, cloned into the pComb3X phagemid vector (from Dr.
Barbas, The Scripps Research Institute, La Jolla, USA) and
electroporated into E. coli ER2738 competent cells. Finally,
cells were cultured and superinfected with the helper phage M13KO7 to
generate the phage library. Next, two rounds of panning were performed.
The eluted phages were used to infect ER2738 cells and grown on LB/agar
plates. From the isolated bacterial colonies, 2 mL cultures of each
clone were grown in SB with 0.1 mg/mL ampicillin at 37 °C and shaking
until an optical density of 0.6 was reached. After that, nanobody
expression was induced by the addition of 1 mM IPTG and incubated with
shaking at 37 °C for 16 hours. After centrifugation, the supernatants,
containing the nanobodies, were collected and tested in ELISA plates
coated with KLH-PG conjugate. Bound nanobodies were detected with an
anti-HA-HRP conjugate following the protocol described in the supporting
information.
Cloning, expression, and purification of rBBN.
Clones that were specific to the PG antigen were selected and the VHH
genes amplified and subcloned into the pINQ-bis vector. This vector is
an in house modified version of the pET-28a(+) vector that allows the
expression of bispecific nanobodies in the periplasm of E. coliBL21(DE3). The modification resulted in an expression cassette coding
for ompA signal peptide, the selected anti-β-lactam VHH flanked by two
SfiI restriction sites, and a spacer of GQAGR(GGGGS)3TSEL, followed by
an anti-omalizumab VHH sequence and a 6xHis tag. The vector was
transformed into E. coli strain (DE3). Individual clones isolated
on LB-kanamycin plates were grown in 500 mL of LB with kanamycin (50
μg/mL) and induced with 10 µM IPTG for 16 hours at 28 °C under shaking.
Cultures were centrifuged and periplasmic proteins extracted by osmotic
shock according to previously established protocols15.
Then, rBBN were purified on 1 mL His-Trap columns on the ÄKTA
purification system (GE Healthcare, Uppsala, Sweden) according to the
manufacturer’s instructions. The imidazole-eluted fractions were
dialyzed against PBS and quantified spectrophotometrically (Abs 280 nm
0.1% = 1.9) and by Bradford’s protein assay16.
Biochemical characterization of rBBN
The purified rBBN was analyzed by 12 % sodium
dodecylsulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and
Coomassie brilliant blue staining method. The specificity and
functionality of the rBBN was evaluated by ELISA, blue native PAGE and
ThermoFluor assay as described in the supporting information.
Analysis of serum samples.
Serum samples from 65 adults were collected in red-top tubes (BD
Diagnostics, Madrid, Spain) and incubated at room temperature for 60
minutes to induce clotting. After centrifugation at 2000 rpm for 15
minutes, the serum was aliquoted into cryovials and stored at -80 ºC
until use. A cohort of 35 allergic patients who report a history of
penicillin allergy confirmed by in vivo tests and 30 non-allergic
individuals were included in the case-control study. The clinical
history included also the concentration of sIgE measured by the
ImmunoCAP system. This study was approved by the Hospital Universitari i
Politècnic La Fe (Valencia, Spain) ethical review committee (registry
no. COBIOPHAD). Signed informed consent were collected from the
participants.
The samples were analysed by the homologous chemiluminescence
immunoassay (hmCLIA). Briefly, a white flat-bottomed polystyrene ELISA
plate was coated with H1-PG conjugate solution (3 μg/mL) in coating
buffer. The next day, the plate was washed 4 times with PBS-T. After
that 25 μL of serum sample was added to each well, followed by
incubation for 30 min at room temperature. The calibration curve was
made using the produced synthetic standards diluted in a pooled control
sIgE-free human serum. All serum samples were analysed in triplicate.
Subsequently, after washing the plate, 25 μL/well of Omalizumab-HRP
solution (1/10,000 dilution) was incubated as before. After, peroxidase
activity was measured by adding 25 µL of enhanced chemiluminescent
substrate solution previously diluted 1/10 in PBS. The luminescent
signals were read at 450 nm, using the EnSpire Multimode Plate Reader
(PerkinElmer, Waltham, USA). The limit of detection (LOD) and limit of
quantification (LOQ) were calculated by measuring ten times the blank
signal and interpolating the mean signal plus 3 and 10 times the
standard deviation to the calibration curve, respectively.
Allergen-specific IgE levels expressed in PSU/mL (Penicillin
Standardised Unit) were determined by interpolating the luminescent
signals into the calibration curve.
Statistical analysis
Coefficients of variation and the parameters to evaluate the assay
sensitivity were calculated by standard descriptive statistical methods
using Microsoft Excel 365 (Microsoft Corporation). The standard data
points were plotted and a four-parameter logistic (4PL) curve was fitted
through the points, using SigmaPlot 12 (Systat Software Inc). ANOVA
statistical analysis, multiple regression analysis and the correlation
study were carried out using SPSS Statistics (IBM). Clinical sensitivity
and specificity were calculated using MedCalc (MedCalc Software) using
ImmunoCAP as reference method.
Results and discussion
Phage display.
The antibody titter of the llama serum, as it can be observed in Figure
S1, was high after the fourth boost, revealing the production of
specific antibodies to PG. Next, a phage-display VHH library with a size
of 1 × 108 transformants was constructed from the
lymphocytes of the immunized llama (see supporting information). Ten
individual clones were selected and their reactivity was tested against
a panel of antigen determinants to β-lactam antibiotics. As is shown in
Figure S2, seven out of ten clones were specific to penicillin G (PG1-5,
PG7-8), two clones were generic to the panel of determinants (PG9-10)
and one was a nonspecific clone (PG6) as it showed reactivity against
the carrier protein KLH. The PG8 clone, the best in performance, was
selected for further cloning an expression.
Cloning, expression, and
purification of rBBN.
After subcloning into the pINQ-bis vector, sequencing confirmed that the
clones possessed the two VHHs sequences of interest. Figure 1A depicts
the structural model of rBBN, using the Robetta web server and
visualized in UCSF Chimera. The model shows the two nanobodies connected
by a loop (yellow); the 17 kDa one (blue) targets the penicillin G
determinant and the 18 kDa one (green) the paratope of Omalizumab.
Purified rBBN was analyzed by 12 % SDS-PAGE gel electrophoresis, and a
clear and intense band was observed (lane 3, Figure 1B), corresponding
to the synthetic standard with an expected molecular weight around 35
kDa.
Biochemical characterization of rBBN
The functionality of rBBN was tested by ELISA, following the protocol
described in the supporting information. As it can be seen in Figure 2A,
the set of data fitted well to a four-parameter logistic model,
revealing the standard was functional, as it targeted both the antigenic
determinant to penicillin G and the paratope of Omalizumab. According to
the lowest detectable concentration, the limit of detection was 0.13
ng/mL (3.7 pM), under the ELISA conditions.
We also conducted a thermofluor assay to check if interaction of rBBN
with H1-PG stabilized the complex. As it can be observed in Figure 2B,
the Tm of rBBN alone is 59 °C while the Tm of rBBN in the presence of
H1-PG increased around seven-Celsius degrees. This indicates that the
rBBN-Penicillin G complex is more stable than rBBN alone, demonstrating
the specific binding between the nanobody and the ligand.
The selectivity of rBBN for H1-PG and Omalizumab was also studied by
Blue Native-PAGE, as rBBN has a high isoelectric point (pI = 8.5) which
prevented the protein from entering the gel in a regular Native-PAGE. In
this way, tested proteins were mixed with a loading buffer containing
Coomassie Blue G (Serva, GmbH) which adds negative charges to the
proteins to force them entering the gel and running according to their
molecular weight (MW) and hydrodynamic conformation. In Figure 2C, the
electrophoretic mobility of the proteins alone and in complex is shown.
The rBBN at 1 and 3 µM run as a single band with high MW
(~245 kDa), possibly due to the high hydrodynamic volume
caused by the conformational flexibility of the loop connecting the two
nanobodies. In the case of the H1-PG conjugate, three bands were shown,
one at ~55 kDa, which we identified as the H1-PG
conjugate monomer, a faint band around ~100 kDa, which
could be ascribed to a dimer and another at high MW (~
245 kDa) ascribed to multimers. Then, rBBN and H1-PG were mixed at a
molar ratio of 1:1 (lane 4) and 1:3 (lane 5) during 15 minutes and were
run in the gel. A decrease in the intensity for the 55 kDa band of H1-PG
was observed as the amount of rBBN increased, while strong accumulation
around the higher MW band was shown, thus, ascribing the electrophoretic
mobility changes to the complex formation. When Omalizumab was assayed,
it barely entered the gel producing two faint bands at
~150 kDa and at ~245 kDa, which could be
ascribed to the monomeric and dimeric antibody respectively, and a
strong accumulation of protein at the top of the gel that was ascribed
to multimers. The addition of rBBN to Omalizumab was concomitant with
the disappearance of the 150 kDa band and a decreased intensity for the
accumulated band which indicated complex formation. Finally, incubation
of the three samples, H1-PG/rBBN/Omalizumab was performed at molar ratio
of 1:1:1 and 1:3:1. As observed before, the band of 150 kDa
corresponding to H1-PG disappeared and a strong accumulation for the
high MW band (245 kDa) was shown. Interestingly, the Omalizumab band at
the top of the gel dramatically decreased intensity and became diffused.
These changes in the electrophoretic mobility are indicative of complex
formation.
Analysis of serum samples.
The produced rBBN was used as a standard to determine the concentration
of sIgE to penicillin G present in a representative serum sample,
following the homologous calibration scheme (Figure 3A) in the hmCLIA
method. The current method for the determination of sIgE follows a
heterologous scheme and defines 1.0 IU/mL as equivalent to 2.4 ng/ml
IgE. Due to the remarkable difference in molecular weight between the
international standard IgE (190 kDa) and rBBN (35 kDa), a new unit
(Penicillin Standardised Unit, PSU), equivalent to 0.44 ng/mL of rBBN
was established on the basis of the IgE-rBBN molar ratio that is 5.4. To
build the calibration curve, rBBN was diluted in sIgE-free serum at
concentrations of 95, 54, 19, 5.4, 3.8, 1.90 and 0.5 PSU/mL. The
stability of rBBN in sIgE-free human was studied for a period of 8 h.
The results revealed that there were no statistically significant
differences (p < 0.0001) between the signals obtained over the
studied period of time (1-8 h), which guarantees its good performance
during the total assay time.
Figure 3B shows the calibration curve. The signals (n = 10) were fitted
to a four-parameter logistic curve, giving a dynamic response ranging
from 0.1 to 95 PSU/mL, and a LOD of 0.02 PSU/mL (8.8 pg/mL), a
concentration below the currently internationally accepted cut-off for
allergy diagnosis (0.1 IU/mL or 240 pg/mL). In the hmCLIA accuracy
assessment, the CVs range from 2 to 9%, resulted in a linear regression
equation with a correlation coefficient (r) of 1. Therefore, the results
reveal that the hmCLIA assay, using the new standards under a homologous
calibration, has good analytical performances.
A cohort of 65 sera collected from patients allergic to penicillin G
(35) and non-allergic subjects (30) was analysed in triplicate by hmCLIA
assay. The signals obtained for each of the samples were interpolated in
both the homologous (synthetic standard) and heterologous (WHO
international standard) calibration curves. Table 1 shows the results.
The Pearson’s correlation coefficient statistical study between the
results obtained using both calibration methods (Figure 4A) displays
that there is a direct positive correlation (r = 0.995, p <
0.0001) between the two interpolation schemes. Furthermore, the results
obtained under the homologous calibration relate well with those
obtained by ImmunoCAP, showing a Pearson correlation close to 1 (r =
0.999, p < 0.0001) (Figure 4B). In addition, as it is shown in
Figure 4B, the slope of the curve is 3, representing the relationship
between the value given by ImmunoCAP and that obtained using rBBN as
standard.
To normalize the results, the concentrations given by the hmCLIA assay
are multiplied by a factor of 1/3 to get those obtained by ImmunoCAP
(see results in Table 1). The use of this factor allows the
classification of patients in the same RAST group. In comparison with
the reference approach, the results using the new standard also showed
recovery values ranging from 91 to 135%, revealing a reliable
quantification.
The linearity of dilution test was performed using a pooled positive
serum sample (sample 36 in Table 1). The linearity dilution study was
carried out using 4-fold serial dilutions (1/4-1/4096) prepared in the
pooled control serum. The experimental concentrations were calculated by
ImmunoCAP and hmCLIA. As is shown in Figure 4C, the linearity of the
hmCLIA assay was good over a wide range of dilution, detecting sIgE up
to the dilution 1/1024 (0.04 PSU/mL) what revealed good flexibility to
quantify low levels of sIgE.
On the other hand, we compared the ability of our assay to identify
allergic and non-allergic individuals. Analysis of the controls
(non-allergic patients) shows that hmCLIA, using the standards
identifies all the control samples as negative, as the results obtained
are below the LOD. The absence of false positives reveals the excellent
specificity (100%) of the hmCLIA method.
Regarding the sensitivity of the in vitro test, the hmCLIA method
identified as positive 23 out of the 35 positive samples (66%), while
ImmunoCAP detected 10 (28%). In order to compare hmCLIA with the
reference method at the diagnostic level, the correlation between them
was made using the samples detected as positive by both assays. It is
worth mentioning that all the positive results given by the ImmunoCAP
were also positive by hmCLIA. Receiver Operating Characteristic (ROC)
(Figure 4D) analysis showed good area under the curve for hmCLIA.
Indeed, both in vitro tests showed high diagnostic specificity
since no false positives were detected. In summary, the clinical
sensitivity of the developed assay was significantly better, doubling
that of the ImmunoCAP assay, what allowed the identification of more
positive allergic patients.
Discussion
Here we showed that the use of selective bispecific binanobodies as
synthetic standards is a relevant alternative to quantify specific IgE,
improving the performances of in vitro tests for penicillin G
allergy testing. This might contribute to the harmonization of this
diagnosis and facilitate the appropriate delabeling of patients
categorized as allergic to penicillin, thereby reducing the spread of
antibiotic resistance and healthcare costs and optimizing patient
outcomes. These standards are robust and highly standardized calibrators
that are easy and inexpensive to produce. Indeed, rBBNs can be
reproduced indefinitely in any laboratory with identical characteristics
requiring only their sequence. This circumvents the issues of
availability and reproducibility related to the use of pooled human sera
and defibrinated plasma, in addition to avoiding potential risks of
transmission of infectious diseases. This positions rBBNs as
advantageous standard material for inter-laboratory comparison studies,
allowing verification of whether the methods produce results that agree
with the reference one, providing the basis for certificates of
quantitative analysis on a given material. In summary, we believe that
the work presented here represents a major step towards the
standardization of in vitro tests for allergen-specific IgE
antibody. This might support the reliable quantification of the level of
sIgE directly to establish a more precise RAST classification. In
addition, the strategy of producing this generation of calibrators with
different specificities would allow more precise diagnostics to
successfully tackle drug allergy delabeling to other antibiotics.