Facilitating EMA binding test performance using fluorescent beads combined with next‐generation sequencing

Abstract The eosin‐5′‐maleimide (EMA) binding test is widely used as diagnostic test for hereditary spherocytosis (HS), one of the most common haemolytic disorders in Caucasian populations. We recently described the advantages of replacing the use of healthy control blood samples with fluorescent beads in a modified EMA binding assay. In this study we further explore this novel approach. We performed targeted next‐generation sequencing, modified EMA binding test and osmotic gradient ektacytometry on consecutive individuals referred to our laboratory on the suspicion of HS. In total, 33 of 95 carried a (likely) pathogenic variant, and 24 had variants of uncertain significance (VUS). We identified a total 79 different (likely) pathogenic variants and VUS, including 43 novel mutations. Discarding VUS and recessive mutations in STPA1, we used the occurrence of (likely) pathogenic variants to generate a diagnostic threshold for our modified EMA binding test. Twenty‐one of 23 individuals with non‐SPTA1 (likely) pathogenic variants had EMA ≥ 43.6 AU, which was the optimal threshold in receiver operating characteristic (ROC) analysis. Accuracy was excellent at 93.4% and close to that of osmotic gradient ektacytometry (98.7%). In conclusion, we were able to simplify the EMA‐binding test by using rainbow beads as reference and (likely) pathogenic variants to define an accurate cut‐off value.

EMA fluorescence can be detected by flowcytometry and mainly reflects decreased RBC Band 3, which in HS is reduced compared to healthy controls [2]. Given its simplicity and the wide availability of flow cytometers, this test can be employed in most laboratories at a low cost. Often, results are reported as a ratio of the individual's mean fluorescent intensity (MFI) to that of healthy controls, making the test somewhat comparable across laboratories [5]. This approach does, however, require blood samples from up to six healthy -and ideally age matched -controls, which can be challenging to locate [5][6][7]. We recently described a modified version of the EMA binding test, in which we substituted healthy control samples with fluorescent beads. [8]. Although healthy controls were still utilised for calibration, the number of control samples needed was reduced significantly.
Performance of this modified EMA binding test was compared to that of the traditional method, using osmotic gradient ektacytometry as validation. We found that accuracy was not compromised, making this approach an attractive and simple alternative [8].
Osmotic gradient ektacytometry is a method for determining RBC deformability and is increasingly used due to the advent of a new generation of ektacytometers [9,10]. Although this test reliably identifies the RBC characteristics associated with HS, it is incapable of discriminating spherocytes in HS from autoimmune haemolytic anaemia [10].
Results from the EMA binding test and osmotic gradient ektacytometry are often sufficient to diagnose HS, but both tests have limitations and may produce equivocal results [33][34][35]. Many previous studies have evaluated these tests mainly using clinical features of HS as proof of disease, hereby creating an inherent risk of confirmation bias [36].
In this study, we wish to further investigate the modified EMA binding test using rainbow beads instead of healthy control samples.
By defining HS as the presence of diagnostic cytoskeleton protein gene mutations identified using tNGS and validating results using osmotic gradient ektacytometry, we provide a reproducible way of estimating a cut-off value for the modified EMA-binding test. Finally, we briefly describe the identified underlying pathogenic mutations.

Population
We included samples from all individuals referred to our laboratory with suspected HS between 1st May 2017 and 1st July 2018 ( Figure 1).
As samples were shipped from other institutions, clinical data were not available. Samples have previously been used to test the performance of the EMA binding test using fluorescent beads versus healthy controls [8].

Ethics
Data were stored and handled in accordance with permission from the Danish Data Protection Agency (10122009 HEH-L.HB). All participants or a parent/guardian consented to diagnostic tests for haemolytic anaemia including tests for HS.

tNGS
Genomic DNA was extracted from peripheral blood using the QIAamp

Osmotic gradient ektacytometry
Osmotic gradient ektacytometry was performed on EDTA stabilised blood within 48 h of sampling, using a LoRRca ektacytometer (RR Mechatronics, Zwaag, Netherlands) as previously described [9]. Two parameters were evaluated on the ektacytometry curve: O min and EI max . O min reflects the minimal RBC surface/volume ratio, increasing in conditions with reduced surface/volume ratio such as HS [39]. EI max reflects the maximal deformability of the RBCs. Reduction of EI max typically represents a reduced RBC surface area, as is seen in HS [39].
O hyper , which reflects hydration status, was not used in this setting as this has been found either high or low in HS [40].

Population
A total of 99 individuals were included in the study. Fifty-six (56%) were female and the mean age was 30.7 years (SD 28.  pathogenic EPB41 mutation was excluded from further analyses, as examination of a peripheral blood smear confirmed the diagnosis of hereditary elliptocytosis (Figure 1). Forty-one patients harbored the common SPTA1 mutation c.6531-12C > T (α-spectrin LELY ), which is considered benign in itself but may cause overt HS, hereditary elliptocytosis or hereditary pyropoikilocytosis in trans to SPTA1 mutations [42,43].

Mutations identified
Seventeen of the 58 individuals harbored more than one mutation.
In total, 34 individuals harbored one or more (likely) pathogenic variants and 24 individuals harbored one or more VUS as the only mutations ( Figure 1). Forty-one individuals had no proven mutations. One

EMA binding test and osmotic gradient ektacytometry as predictors of mutation status
We demonstrated significantly higher EMA (ΔMFI%) values in individuals with (likely) pathogenic variants compared to individuals without mutations. Similarly, EMA values were significantly higher in individuals with VUS compared to individuals with no mutations (Figure 2A; p = 0.00044). In ROC curve analysis, we found that a threshold of Applying these thresholds, we subsequently calculated: sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy for the individual parameters ( Table 2). All demonstrated sensitivities, specificities, NPV and PPV above 87%. As a single parameter, EI max , yielded the best results with an accuracy of 95.1%.

Combining osmotic gradient ektacytometry and the EMA binding test to predict mutation status
The distribution of all 95 samples, based on the EMA binding test, O min and EI max values, is illustrated in Figure 3. Adjusting the O min ≥ 166 mOsm/kg while keeping EI max < 0.579 improved the sensitivity to 100%, while keeping an excellent specificity of 97.4% (Table 2). Subsequently, we calculated the sensitivity, specificity, PPV, NPV and accuracy using the obtained EMA, O min , and EI max thresholds in combination. However, this approach resulted in a marked reduction of sensitivity and NPV without improving other measures ( Table 2). Figure 3B shows an excellent relationship between the modified EMA binding test with fluorescent beads and EI max in individuals with (likely) pathogenic variants and VUS, regardless of the type of the mutated gene.

DISCUSSION
In this study, we assessed the number of RBC cytoskeleton protein gene mutations in a population of individuals with suspected HS, using pathogenic mutations as the gold standard. This enabled us to set a diagnostic cut-off value for our newly described modified EMA binding test with fluorescent beads (Figure 2 and Table 2), thereby alleviating it from the otherwise obligatory use of up to six healthy control samples [5]. Using the EMA binding test alone, we obtained a diagnostic accuracy ( Table 2) comparable to those previously reported using healthy controls [44][45][46][47]. In many settings, obtaining suitable control samples can be challenging [8]. Furthermore, the inherent variation in control samples complicates interlaboratory comparisons and quality assessment schemes [24]. Our approach has demonstrated a robust performance, comparable to that of the traditional EMA-binding test with healthy controls [8] and osmotic gradient ektacytometry (the gold standard of membranopathy diagnostics) across a range of causative genes ( Figure 4).  Table 1). c.2909 C > A was originally classified as pathogenic (autosomal recessive) [52], but this is likely due to frequent co-occurrence of c.4339-99C > T in cis [51].

F I G U R E 3 Genetic variants and functional testing. (A) Distribution
In contrast, all VUS were missense mutations, except three intron mutations (two in the SPTA1 gene and one in the SLC4A1 gene).
As α-spectrin is synthesised in excess [53], heterozygous SPTA1 pathogenic mutations are considered clinically benign but may be pathogenic in homozygous and compound heterozygous state. Accordingly, individuals heterozygous for (likely) pathogenic STPA1 mutations were not used in ROC analysis but several had borderline ΔMFI% changes (Table 1). This is in line with some degree of RBC surface area loss and even mild clinical haemolysis as previously described [28]. Four individuals in our study only harbored a heterozygous SPTA1 mutation (HS29, HS72, HS78, HS94 in Table 1