A comparison of two PCR protocols for the differentiation of Plasmodium ovale species and implications for clinical management in travellers returning to Germany: a 10-year cross-sectional study

Background To assess the occurrence of Plasmodium ovale wallikeri and Plasmodium ovale curtisi species in travellers returning to Germany, two real-time PCR protocols for the detection and differentiation of the two P. ovale species were compared. Results of parasite differentiation were correlated with patient data. Methods Residual nucleic acid extractions from EDTA blood samples of patients with P. ovale spp. malaria, collected between 2010 and 2019 at the National Reference Centre for Tropical Pathogens in Germany, were subjected to further parasite discrimination in a retrospective assessment. All samples had been analysed by microscopy and by P. ovale spp.-specific real-time PCR without discrimination on species level. Two different real-time PCR protocols for species discrimination of P. o. curtisi and P. o. wallikeri were carried out. Results were correlated with patient data on gender, age, travel destination, thrombocyte count, and duration of parasite latency. Results Samples from 77 P. ovale spp. malaria patients were assessed, with a male:female ratio of about 2:1 and a median age of 30 years. Parasitaemia was low, ranging from few visible parasites up to 1% infected erythrocytes. Discriminative real-time PCRs revealed 41 cases of P. o. curtisi and 36 cases of P. o. wallikeri infections. Concordance of results by the two PCR approaches was 100%. Assessment of travel destinations confirmed co-existence of P. o. curtisi and P. o. wallikeri over a wide range of countries in sub-Saharan Africa. Latency periods for the two P. ovale species were similar, with median values of 56.0 days for P. o. curtisi and 58.0 days for P. o. wallikeri; likewise, there was no statistically significant difference in thrombocyte count with median values of 138.5/µL for patients with P. o. curtisi and 152.0/µL for P. o. wallikeri-infected patients. Conclusions Two different real-time PCR protocols were found to be suitable for the discrimination of P. o. curtisi and P. o. wallikeri with only minor differences in sensitivity. Due to the overall low parasitaemia and the lack of differences in severity-related aspects like parasite latency periods or thrombocyte counts, this study supports the use of P. ovale spp. PCR without discrimination on species level to confirm the diagnosis and to inform clinical management of malaria in these patients. Electronic supplementary material The online version of this article (10.1186/s12936-019-2901-0) contains supplementary material, which is available to authorized users.

Background Human Plasmodium ovale spp. malaria is caused by two different sympatric Plasmodium species, namely P. ovale curtisi, which was considered the "classic" human pathogen, and P. ovale wallikeri, which has been described as a genetically distinguishable variant. Parasitological features such as parasitaemia, 18S rRNA gene copy number, or pan-aldolase activity are similar, with an overrepresentation of infected erythrocytes lacking Schüffner's stippling among P. ovale wallikeri isolates [1]. Although the organisms are morphologically virtually indistinguishable and are associated with sympatric ranges through tropical Africa, Asia, and Oceania, considerable genetic differences have been observed between P. ovale curtisi and P. ovale wallikeri. They do not recombine and are associated with substantial genetic divergence between sexual stage proteins, such as 6-cysteine and proteins containing Limulus coagulation factor C domains, probably resulting in sexual incompatibility [2]. Phylogenetic separation and diversity of P. o. curtisi and P. o. wallikeri surface antigens support species status [3,4].
Potential differences in pathogenicity of P. o. curtisi and P. o. wallikeri are of ongoing interest. An early Spanish study assessing patients from 2005 to 2011 indicated significantly more severe thrombocytopenia in patients with P. o. wallikeri malaria as well as non-significant trends for shorter time periods between infection and onset of clinical symptoms (latency period), lower albumin levels, higher temperature, and more haemolysis markers compared to patients with P. o. curtisi infections [13]. A British study showed geometric mean latency periods of 40.6 days for P. o. wallikeri and of 85.7 days for P. o. curtisi, with both parasites frequently occurring in patients despite their using chemoprophylaxis during travel compared with Plasmodium vivax or Plasmodium falciparum malaria [15]. A multi-centre study conducted in France, Italy, and Spain recently also suggested P. o. wallikeri to be more pathogenic, as defined by associations with more severe thrombocytopenia, coagulopathy as expressed by higher international normalized ratios (INR), and shorter latency in Caucasians.
Further, P. o. wallikeri infections were more frequent in males and particularly in Caucasians. However, severe cases were observed for P. o. curtisi as well, and infections with both P. o. spp. suggested an association with diabetes mellitus and-again-decreased effectiveness of antimalarial drugs as chemoprophylaxis compared with P. falciparum [14]. The latter phenomenon might be associated with the geometric mean latency periods and the formation of hypnozoites. Recently, a well-characterized case of a hypnozoite-associated relapse of P. o. wallikeri malaria with a strain of identical genotype has been reported [29]. However, few well-documented cases of P. ovale spp. malaria relapses have been published so far [29][30][31] with a strong dominance of P. o. curtisi [31].
In this study, samples from 77 patients with P. ovale spp. malaria returning to Germany between 2010 and 2019 were further differentiated in a retrospective assessment based on two different real-time PCR approaches. Results were compared with patient characteristics to contribute to the scant information available on the possibly differences in aetiological relevance of P. o. curtisi and P. o. wallikeri.

Aim, design, and setting
The epidemiology of P. o. curtisi and P. o. wallikeri in travellers returning to Germany and the available patientrelated information were assessed in a retrospective cross-sectional study over 10 years using clinical information and residual sample materials collected at the Keywords: Malaria, Plasmodium ovale curtisi, Plasmodium ovale wallikeri, Parasite latency, Epidemiology, Test comparison Bernhard Nocht Institute for Tropical Medicine Hamburg, the German National Reference for Tropical Pathogens, between 2010 and 2019.

Samples
Residual nucleic acid extractions from EDTA blood samples of all 77 patients with confirmed P. ovale spp. infections within the study interval were used, which had been collected and stored at − 80 °C at the Bernhard Nocht Institute for Tropical Medicine, Hamburg, Germany. The samples were collected between 2010 and 2019. Extractions had been performed using the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany). All samples had been assessed by species-specific real-time PCR before storage (RealStar Malaria S&T PCR Kit 1.0, altona Diagnostics, Hamburg, Germany), and a subset also by microscopy (Additional file 1).

PCR-based discrimination of P. ovale curtisi and P. ovale wallikeri
All DNA samples were analysed using two different published real-time PCR protocols suitable for the discrimination between P. o. curtisi and P. o. wallikeri. PCR protocol 1 was applied as a duplex real-time PCR using the primers for P. o. curtisi and P. o. wallikeri as described by Bauffe et al. [18] (Additional file 2). The reaction was run on a Corbett RotorGene 6000 or a Corbett Rotor-Gene Q cycler using 20 µL volumes with a HotStarTaq Mastermix (Qiagen). The final MgCl 2 concentration was adjusted to 3 mmol/L; the applied primer concentrations were 1.6 pmol/µL and the probe concentration 0.1 pmol/µL each. A 5 µL volume of extracted DNA was used. Reaction conditions comprised 10 min at 95 °C followed by 45 cycles of 10 s denaturation at 95 °C as well as annealing and amplification for 30 s at 60 °C with subsequent cooling to 40 °C for 30 s.
PCR protocol 2 consisted of two separate real-time PCRs, one specific for P. o. curtisi and the other specific for P. o. wallikeri as described by Calderaro et al. [32,33] (Additional file 2). Again, 20 µL volumes were assessed on a Corbett RotorGene 6000 or a Corbett RotorGene Q cycler using the HotStarTaq Mastermix (Qiagen). The optimal final MgCl 2 concentration was 3 mmol/L; the primer concentrations were adjusted to 1.2 pmol/µL and the probe concentration to 0.25 pmol/µL each. The amount of extracted DNA used was 2 µL. Reaction conditions comprised 10 min at 95 °C followed by 45 cycles of 15 s denaturation at 95 °C as well as annealing and amplification for 60 s at 60 °C with subsequent cooling to 40 °C for 30 s.
The oligonucleotide sequences of the primers, probes, and positive control plasmids are shown in Additional file 2. The positive control plasmids based on pEX-A128 vector backbones (eurofins Genomics, Ebersberg, Germany) were used to prepare rows of 10-fold dilutions in order to calculate detection limits using the program SciencePrimer (http://scien cepri mer.com/copy-numbe r-calcu lator -for-realt ime-pcr, last accessed March 8, 2019). For PCR protocol 1, calculated detection limits were < 1 copy/µL for P. o. wallikeri and 52 copies/µL for P. o. curtisi; for PCR protocol 2, calculated detection limits were 8 copies/µL for P. o. wallikeri and 9 copies/µL for P. o. curtisi.
Inhibition control PCRs were not included in these typing PCRs, because inhibition had already been excluded during the initial diagnostic PCR that led to the diagnosis of P. o. spp. infection.

Patient characteristics
All patients were anonymized and the respective clinical data were computer-coded. As far as available, the following data were assessed: gender, age, region of presumed infection, number of days between leaving the endemic country and onset of malaria symptoms, and thrombocyte counts.

Statistics
Due to the low number of P. ovale spp. positive samples, only descriptive analysis was performed. The Mann-Whitney test was applied to independent parameters as age, latency time, and thrombocyte count using the GraphPad Instat version 3.06 software (GraphPad Software, Inc., San Diego, CA, USA). Matched parameters as Ct-values while comparing both PCR assays were assessed using Wilcoxon's signed ranks test with the help of the open-source software R Commander version 2.5-3, which was also used to calculate descriptive parameters like interquartile ranges (IQR).

Ethics
Ethical clearance was provided by the Ethics Committee of the Medical Association of Hamburg, Germany (registration number WF-012/19), allowing the anonymized assessments to be made without informed consent.

Samples
Residual DNA from 82 diagnostic EDTA blood samples containing parasites of P. ovale spp. according to the results of microscopy and PCR was stored at the German National Reference Center for Tropical Pathogens Bernhard Nocht Institute for Tropical Medicine Hamburg (BNITM). The samples had been sent to BNITM from clinicians and laboratories all over Germany for diagnostic verification. Of the 82 samples, 5 represented follow-up samples. Accordingly, first diagnostic samples from 77 independent patients were included in this study.
Parasitaemia was generally low and ranged from a few visible parasites up to 1% infected erythrocytes. Initial microscopic identification of P. ovale spp. had been successful in 12 samples, while initial misidentification as P. vivax had occurred in 3 cases. In the remaining 67 samples, either parasitaemia was insufficient for a qualified microscopic parasite differentiation (n = 40) or microscopy had not been performed at the Bernhard Nocht Institute because PCR was requested by the sender of the samples only and the initial microscopic results from the senders' laboratories were unknown (n = 17) ( Table 1).

Discussion
The study compared two recently published real-time PCR protocols for the identification and differentiation of P. o. curtisi and P. o. wallikeri in clinical samples. Results of parasite subtyping were correlated with patient characteristics. All patients were travellers returning to Germany from countries known to be endemic for malaria. The results further confirmed previous findings about the high specificity of PCR for the diagnosis of non-falciparum malaria at species level, as several misidentifications during the initial microscopic assessment were corrected. In addition, both real-time PCR protocols that were compared in this study proved to be suitable for the identification of P. o. curtisi and P. o. wallikeri with 100% concordance. The protocol of Bauffe et al. [18] was slightly more sensitive than that of Calderaro et al. [32,33], as suggested by lower Ct values and by a case of failed detection of P. o. wallikeri in a post-treatment control sample. Another advantage of the Bauffe protocol was that the assay can be run as a one-well-duplex PCR, while the Calderaro protocol requires two separate PCR assays to distinguish P. o. curtisi and P. o. wallikeri. Accordingly, the relatively more sensitive and convenient duplex PCR by Bauffe et al. [18] is the preferred test in this patient population.
Of note, P. ovale spp. infections frequently occur in asymptomatic individuals in high endemicity settings [40]. Underlying acquired resistance phenomena have been described as early as in the late 1930s [41,42]. In the here-described study, only patients were included whose clinical symptoms had triggered diagnostic testing for malaria. Accordingly, no conclusions can be drawn on the diagnostic performance of the applied PCR approaches in patients with completely asymptomatic P. ovale spp. infections. From previous studies, however, generally superior sensitivity of PCR compared to microscopy [43][44][45] and alternative modern approaches like mass spectrometry [46] for malaria screening are well documented. Especially, species-specific malaria PCR is considered as more reliable in detecting mixed plasmodial infections in particular in cases with significant differences in density of the different parasite species [44,47,48]. Further, PCR is a reliable tool if suboptimal pre-analytic conditions do not allow parasite microscopy, e.g. in case of hemolytic blood [49].
With regard to P. ovale spp. epidemiology, the heredescribed assessment clearly confirmed the co-occurrence of P. o. curtisi and P. o. wallikeri in a wide range of malaria-endemic countries in Sub-Saharan Africa, as already reported [3,4]. In contrast to previous reports [13][14][15], the findings shown here did not support the suggestion of earlier onsets of clinical disease due to P. o. wallikeri compared with P. o. curtisi [13][14][15]. Of note, latency could only be assessed for 52 out of 77 patients (67.5%). Likewise, significant differences in thrombocyte counts were not confirmed. Differences in thrombocyte counts have been reported earlier and were used as an indicator for differences in disease severity. Admittedly, thrombocyte count was available for 19 patients (24.7%) only. Altogether, the here-described study comprised only half as many samples as the assessment by Nolder et al. [15] but it was of similar size to the multi-national study of Rojo-Marcos et al. [14] comprising 79 patients. However, the latter multicentre study [14] was prospective and thus more robust, while the interpretation of the here-described study is limited by its retrospective design. As further limitations, patient characteristics and travel information were documented only for subsets of patients and information on the proportions of travellers visiting friend and relatives as well as of immigrants were no available. Next to this, details on ethnicity, prophylaxis and therapy could not be provided.