A comparison of the clinical, laboratory and epidemiological features of two divergent subpopulations of Plasmodium knowlesi

Plasmodium knowlesi, a simian malaria parasite responsible for all recent indigenous cases of malaria in Malaysia, infects humans throughout Southeast Asia. There are two genetically distinct subpopulations of Plasmodium knowlesi in Malaysian Borneo, one associated with long-tailed macaques (termed cluster 1) and the other with pig-tailed macaques (cluster 2). A prospective study was conducted to determine whether there were any between-subpopulation differences in clinical and laboratory features, as well as in epidemiological characteristics. Over 2 years, 420 adults admitted to Kapit Hospital, Malaysian Borneo with knowlesi malaria were studied. Infections with each subpopulation resulted in mostly uncomplicated malaria. Severe disease was observed in 35/298 (11.7%) of single cluster 1 and 8/115 (7.0%) of single cluster 2 infections (p = 0.208). There was no clinically significant difference in outcome between the two subpopulations. Cluster 1 infections were more likely to be associated with peri-domestic activities while cluster 2 were associated with interior forest activities consistent with the preferred habitats of the respective macaque hosts. Infections with both P. knowlesi subpopulations cause a wide spectrum of disease including potentially life-threatening complications, with no implications for differential patient management.


Results
Baseline characteristics. Of 722 admissions with a diagnosis of malaria, there were 420 eligible patients with PCR confirmed P. knowlesi monoinfection (Fig. 1). These patients had a median [IQR] age of 42  years, most were male (61.4%) and Iban was the dominant ethnicity (86.2%) ( Table 1). Single P. knowlesi subpopulations were identified in 413/420 patients (98.3%), of which 298/413 (72.2%) cases were cluster 1 and 115/413 (27.8%) cluster 2. While there was no significant age difference between subpopulations, patients with cluster 2 infections were more likely to be male (p = 0.043) and to have reported recent jungle activity (p = 0.023) than those with cluster 1. After adjusting for age, types of activities remained associated with subpopulations, with patients with cluster 1 infections 1.8-fold (95% CI 1.14-2.95, p = 0.013) more likely to report peri-domestic activity than those with cluster 2 ( Table 2). The types of activities were sex-and age-specific, with females being fivefold (OR 5.09 (95% CI 3.09-8.62), p < 0.001) more likely to report peri-domestic activity compared to males and each year increase in age increasing the odds of reported peri-domestic activity (OR 1.04 (95% CI 1.03-1.06), p < 0.01) (see Supplementary Table S1 online).
The median duration of illness before presentation and route of hospital referral in the two groups were similar (Table 1). Of the 27.6% patients who received treatment before recruitment, intravenous artesunate was given to one patient with a cluster 1 infection (0.3%) and two patients with cluster 2 infections (1.7%). Most cases were uncomplicated with 44/420 (10.5%) defined as severe according to WHO criteria 27 . There were seven patients with mixed cluster 1 and 2 infections and all were men. Their median [IQR] age was 42 [36][37][38][39][40][41][42][43][44][45][46][47][48] years and one had severe disease characterised by hyperparasitaemia, jaundice, acute kidney injury (AKI) and severe anaemia. Laboratory findings. Haematological and biochemical findings at presentation were largely similar in the subpopulation groups, with thrombocytopenia being the commonest and almost universal abnormality (Table 3). Patients with cluster 1 infections had lower median platelet counts (p < 0.001) and serum sodium concentrations (p = 0.031), and had higher serum total bilirubin (p = 0.049) and blood urea concentrations (p = 0.002), than those with cluster 2 infections (Table 3). These differences persisted after adjustment for age, sex, and ln (parasitaemia) (p < 0.05; Supplementary Patients not recruited since study clinicians not working and/or not informed (33), b Only index cases were recruited, c Other concurrent infections such as active tuberculosis (n = 1), leptospirosis (n = 2) and dengue positive by rapid diagnostic test (n = 1) and d Not recruited; PCR analysis not done for one patient who was identified posthumously as P. knowlesi by microscopy while another was identified as P. knowlesi after transfer to a tertiary hospital (patient's sample obtained and subsequently identified by PCR). e Patients with both cluster 1 and cluster 2 infections. Pk = P. knowlesi; Pm = P. malariae; Pf = P. falciparum, Pv = P. vivax, Po = P. ovale.

Predictors of type of P. knowlesi subpopulation.
Optimal thresholds for platelet count, serum concentrations of sodium, total bilirubin and urea controlled for age, sex and ln (parasitaemia) in multivariable logistic regression remained independently significant for differences between subpopulations (see Supplementary Table S3 online). However, the AUCs for all variables were < 0.7.
Severe malaria. On the basis of WHO research criteria 27 , we identified 44 cases of severe malaria, with AKI being the commonest complication. Of these, 22 (50%) met 2 or more severe criteria (Fig. 2). The commonest criterion to develop after admission was severe anaemia, occurring in eight patients including three cases that were uncomplicated at presentation ( Table 4). Out of these eight patients, 5 had hyperparasitaemia, 2 had a parasite count > 20,000/µL and one patient had a parasite count of 243/µL but with abnormal bleeding 28 . No patients presented with, or developed, hypoglycaemia or met the criteria for cerebral malaria. Blood cultures were taken from six severely ill patients with clinical features of sepsis. All these patients were culture negative but received empirical intravenous antibiotic therapy (ceftriazone in three cases, metronidazole plus cefoperazone in two with clinical features of intra-abdominal infection, and amoxycillin/clavulanate in one). A higher incidence of severe disease was present in patients with cluster 1 infections (11.7% [95% CI: 8.3-16%] versus 7.0% [95% CI 3.1-13.3%] in cluster 2) but this did not reach statistical significance (p = 0.208; Table 1). The severe cases with cluster 1 tended to have lower parasite counts compared to severe cluster 2 infections (geometric mean, 24,052  www.nature.com/scientificreports/ vs 89,322/µL, p = 0.098), with 50% (4/8) of severe cluster 2 cases having hyperparasitaemia (> 100,000/ µL) compared to 34% (12/35) in the severe cluster 1 group (p = 0.443) ( Table 4). There was no significant difference observed between the proportions of severity criteria met by cases of severe malaria in the two P. knowlesi subpopulations (Table 4). Among all knowlesi malaria cases, AKI was the commonest complication in cluster 1 patients, occurring in 5.6% (16/298) of cases, while for cluster 2 infections the commonest severe criteria were hyperparasitaemia and jaundice, both of which occurred in 3.5% (4/115) of cases (Table 5). Multivariable logistic regression using severe criteria as outcome among all knowlesi malaria cases showed no significant difference between the two clusters for any severe criteria (Supplementary Table S4 online). Subpopulation was not independently associated with severe disease (p = 0.427, see Supplementary  Table S5 online).
Ordinal regression was used to assess associations between cluster types and markers of end organ disease, and identified a significant association between clusters for renal dysfunction (Fig. 3). Cluster 1 infections showed 2.4-fold higher odds of step wise change from normal to abnormal and to severe renal dysfunction compared to cluster 2 infections (p = 0.036) after adjustment for age, sex and ln (parasitaemia). Clinical course. Overall 12 patients developed one or more severe criteria, with 4 developing anaemia (two from each cluster respectively with one cluster 1 uncomplicated upon presentation), 2 developed ARDS (cluster 1 and mixed cluster), 1 developed ARDS with anaemia (cluster 2), 1 developed anaemia and acidosis (cluster 1), 1 developed anaemia and abnormal bleeding (cluster 1, uncomplicated at presentation), 1 developed AKI and anaemia (cluster 1), 1 developed AKI and 1 developed hypotension respectively (both were cluster 1, the latter was uncomplicated at presentation) (Fig. 2).
There were no deaths but two patients who were not recruited to the present study died during the study period; one identified posthumously as P. knowlesi by microscopy and another having been initially missed at presentation to Kapit Hospital but who was subsequently diagnosed at a tertiary hospital 29 . The latter case was confirmed as P. knowlesi by PCR and subsequently identified as cluster 1. Taking these two fatal cases into account, the overall fatality rate was 0.35% (2/572) (95% CI 0.04-1.3%).

Discussion
In the present study, the clinical and laboratory features of patients infected with one of two divergent subpopulations of P. knowlesi in Kapit arising from different zoonotic hosts, one associated with long-tailed macaques (cluster 1) and the other with pig-tailed macaques (cluster 2) 9 , were prospectively characterised. Infections with each subpopulation resulted in a wide spectrum of disease but the clinical presentation was predominantly with . The scale on the far right x axis shows the total number of patients with each severe criteria. The y-axis of the graph shows number of patients with same combination/s of severe criteria. The criteria were : hyperparasitaemia (> 100,000 parasites/μL); severe anaemia (haemoglobin concentration < 7 g/dL); hypotension (systolic blood pressure < 80 mmHg); acute kidney injury (serum creatinine > 265 μmol/L or blood urea > 20 mmol/L); jaundice (serum total bilirubin > 50 μmol/L) with parasite count > 20,000/μL; acidosis (base deficit > 8 mEq/L or plasma bicarbonate < 15 mmol/L or venous plasma lactate ≥ 5 mmol/L); acute respiratory distress syndrome (ARDS) (respiratory rate > 30 breaths/minute plus oxygen saturation < 92% on air and/or pulmonary infiltrates on chest radiograph).   Table 5. Overall distribution of severe criteria among patients infected with P. knowlesi. Data are presented as no. (%). Hyperparasitaemia is defined as > 100,000 parasites/μL; severe anaemia defined as haemoglobin level < 7 g/dL; hypotension-systolic blood pressure < 80 mmHg; acute kidney injury-serum creatinine > 265 μmol/l or blood urea > 20 mmol/l; jaundice-serum bilirubin > 50 μmol/l with parasite count > 20,000/μl; hypoglycaemia-serum glucose < 2.2 mmol/l; acidosis-base deficit > 8 meq/l or plasma bicarbonate < 15 mmol/l or venous plasma lactate ≥ 5 mmol/l; ARDS (acute respiratory distress syndrome)respiratory rate > 30 breaths/minute plus oxygen saturation < 92% on air and/or pulmonary infiltrates on chest radiograph. p value is between cluster 1 and cluster 2. Abnormal bleeding involved splenic bleeding.  www.nature.com/scientificreports/ uncomplicated malaria. Furthermore, the incidence of severe disease, the proportions of patients with individual criteria for severity, and the distribution of parasitaemia did not differ significantly between subpopulations. Haematological and biochemical findings revealed minor, clinically insignificant differences between subpopulations with cluster 1 infections having lower platelet counts and sodium serum concentrations, together with higher serum total bilirubin and urea concentrations, compared with cluster 2.
An ordinal logistic regression model was used to explore markers of end organ disease and found that, after adjusting for age, sex and ln (parasitaemia), cluster 1 had a greater association with renal dysfunction. A previous study in patients at Sibu and Sarikei in Sarawak has shown increased end organ disease with selected alleles in P. knowlesi subpopulations, including renal dysfunction 26 . In that study, the driver of a severe phenotype was an association with the P. knowlesi normocyte binding protein (PkNBP) xa gene, which along with the Duffy binding protein (DBP) gene, has been shown to be highly differentiated between the subpopulations 30 . PkNBP belongs to a family of invasion ligands proteins known as reticulocyte binding-like proteins that are associated with recognition of erythrocyte receptors upon which DBP proteins are released to begin invasion 31,32 . The dimorphism observed in these two genes do suggest plausible differences in efficiency of host cell selection and invasion between the two subpopulations. Disease severity in malaria is a complex interplay between the host's immune system, parasite virulence and also entomological inoculation rate which contribute to the wide clinical spectrum commonly seen in malaria 33,34 . Therefore, clinical disease severity and virulence extend beyond heterogeneity in these genes which may explain the lack of distinct clinical phenotypes observed.
Consistent with previous prospective studies in Sarawak and Sabah 18,20 , severe disease was present in 10.5% of all knowlesi malaria cases regardless of subpopulation. Also consistent with previous reports 5,20,35 , AKI was the commonest severity criteria observed, while two or more severity criteria were present in half of our patients with severe disease. Anaemia following treatment occurred in eight of the patients. Post-treatment anaemia in uncomplicated malaria is well recognised and multifactorial, but has been attributed mainly to haemolysis of both parasitized and non-parasitized erythrocytes 36 . Delayed haemolysis has also been reported with artemisinin derivatives which could be a possible contributing factor in the present series 37,38 . This is thought to result from shorter lifespan in pitted erythrocytes following treatment and this has been found to correlate with hyperparasitaemia in falciparum malaria 37 .
The incidence of patients meeting severe respiratory criteria in the current study (n = 14/44, 32%) was higher than seen in a recent study from Sabah (n = 2/28, 7%) 20 . This may reflect between-study differences in diagnostic criteria. Overall there were no differences in severe phenotype between subpopulations but, since there were only eight severely ill patients with cluster 2 infection, more cases will need to be studied before definitive conclusions can be reached. The clinical characteristics of the seven patients infected with both subpopulations were not compared to the single cluster infections due to their small number. However, it is of interest that all were males with one severe case having 4 severe criteria including hyperparasitaemia (167,434/µL).
None of the patients recruited into the study died but, if two fatal knowlesi malaria cases that were known to have occurred during the study period were included, this would represent a mortality rate of 0.35%. This is lower, but not statistically different to, the rate of 1.8% observed during our first prospective study conducted 10 years ago at Kapit Hospital (p = 0.13) 18 . Despite an increased number of health staff, improved awareness of knowlesi malaria as a potential diagnosis, and prompt use of artemisinin-based combination therapies at Kapit Hospital compared with a decade ago, the two deaths that occurred were associated with delays in diagnosis 29 . Although late presentation associated with limited local transportation and health infrastructure may have been  www.nature.com/scientificreports/ unavoidable, it was also frequently observed in the early descriptions of fatal cases 39 . As thrombocytopenia is almost universally detected at presentation in knowlesi malaria cases, it remains one of the most useful prompts for including P. knowlesi infection in the differential diagnosis of illness in endemic areas and in travellers returning from malaria-endemic countries 18 . The age and sex distributions of the patients in the present study were similar to those previously described in the Kapit Division 18 , and the proportions of patients with cluster 1 and 2 infections were consistent with that of our previous study at approximately 70% and 30%, respectively 13 . Patients who had undertaken recent peri-domestic or forest fringe activities were more likely to have cluster 1 infections while those with cluster 2 infections reported recent jungle or interior forest exposure such as hunting and logging. The higher proportion of cluster 1 infections and the different risk factors of acquiring the two subpopulations could be due to the differences between the habitats preferred by the macaque hosts. Long-tailed macaques are adaptable to environmental changes and have a wide range of habitat, including being found in secondary forests close to human habitation 14,15 . In contrast pig-tail macaques prefer to live in primary forests away from human habitation 16 . It is also possible that there are differences in the mosquito species responsible for transmission of each of the two subpopulations. Detailed studies on the bionomics of vectors and the distribution and relative abundance of the two different macaque hosts need to be undertaken for a comprehensive understanding of the transmission dynamics of the two P. knowlesi subpopulations in Malaysian Borneo.
In conclusion, infections with the two divergent subpopulations of P. knowlesi present in Kapit resulted in similar disease phenotypes, and so testing for cluster type at presentation is unlikely to alter clinical management. Although they result in predominantly uncomplicated malaria, both clusters can lead to potentially life threatening complications. The population risk profile differs between the two subpopulations, and cluster 1 infections may cause severe disease across a wider range of parasitaemia and are more likely associated with renal dysfunction. Despite increased awareness and improved healthcare compared to the first prospective study 10 years ago, the knowlesi malaria fatality rate has not changed. Clinicians working in Southeast Asia and those managing travellers returning from P. knowlesi endemic areas thus need to be aware of the wide spectrum of disease and risk of potentially life-threatening complications in knowlesi malaria.

Study site and subjects. A prospective observational study was conducted at Kapit Hospital, in the Kapit
Division of Sarawak, Malaysian Borneo. Kapit Division covers an area of 38,934 km 2 with a total population of 112,762 and mainly of Iban ethnicity (67.4%) 40 . All individuals who are blood slide microscopy-positive for malaria are required to be hospitalised in Malaysia and discharged when clinically well and have had two malaria-negative blood films on two consecutive days.
All non-pregnant patients aged ≥ 15 years admitted to Kapit Hospital between September 2016 and October 2018 with a positive blood film for either P. knowlesi or P. malariae and no travel history outside of Malaysia within the previous 28 days were eligible for enrolment. Patients with significant co-morbidities (established end organ disease or co-infection) or who had received antimalarial treatment within the previous 14 days were excluded. Patients diagnosed at a peripheral health care facility were included if they arrived at Kapit Hospital within one hour of starting antimalarial treatment. Only patients with a PCR confirmed P. knowlesi monoinfection were subsequently included in the study. All patients provided informed written consent to the study procedures which were approved by the Medical Research Ethics Committee of Universiti Malaysia Sarawak and the Medical Research and Ethics Committee of the Ministry of Health Malaysia (NMRR-16-943-31224 [IIR]). All procedures on patients were performed in accordance with relevant guidelines and regulations.
Epidemiology data and clinical procedures. Patient demographics, recent travel history including activities within 14 days prior to hospital admission, clinical details and baseline laboratory data were recorded on a standardised case report form by study clinicians. Recent peri-domestic activities were defined as any outdoor activities that were conducted around the perimeter of residential households and towards the town centre, while recent jungle activities refer to any outdoor activities inside deep jungle including logging, activities around the logging camps, jungle trekking and hunting. Ten mL of venous blood was collected from each patient for biochemical and haematologic tests. Severe knowlesi malaria disease was defined according to the WHO research criteria 2014 27 at the time of, or during course of admission. The criteria were: hyperparasitaemia (> 100,000 parasites/μL); severe anaemia (haemoglobin concentration < 7 g/dL); hypotension (systolic blood pressure < 80 mmHg); AKI (serum creatinine > 265 μmol/L or blood urea > 20 mmol/L); jaundice (serum bilirubin > 50 μmol/L) with parasite count > 20,000/μL; hypoglycaemia (serum glucose < 2.2 mmol/L); acidosis (base deficit > 8 mEq/L or plasma bicarbonate < 15 mmol/L or venous plasma lactate ≥ 5 mmol/L); acute pulmonary oedema/acute respiratory distress syndrome (ARDS) (respiratory rate > 30 breaths/minute plus oxygen saturation < 92% on air and/or pulmonary infiltrates on chest radiograph); cerebral malaria (unarousable coma). In assessing the presence of acidosis, serum bicarbonate or base excess was used when plasma lactate was unavailable, or vice versa, as per WHO 2014 research criteria.
Treatment followed the Malaysian Ministry of Health (MMoH) guidelines, including the diagnosis of severe malaria in patients with > 20,000 parasites/µL 41 . All patients received artemether-lumefantrine and/or intravenous artesunate depending on disease severity. Clinical management, including the investigation and treatment of bacterial infection, was otherwise at the discretion of the attending clinician. Similarly, each patient was assessed at least daily by the treating team and baseline investigations only repeated if clinically indicated.
Laboratory procedures. Blood  www.nature.com/scientificreports/ extracted from the blood spots using InstaGene™ 42 and nested PCR assays for identification of malaria species with primers for P. falciparum, P. malariae, P. vivax, P. ovale, P. knowlesi, P. cynomolgi and P. inui were utilised as described previously 1,43,44 . Each P. knowlesi subpopulation was identified using cluster-specific PCR assays 13 . Parasite counts were determined using the average readings of two independent experienced laboratory technicians derived from the number of parasites per 500 white cells, adjusted by the total white cell count for each patient.
Haematological profiles on-site at the hospital laboratory were determined using semi-automated methods (Mek-6410 K, Nihon Kohden Corporation, Japan). Citrated blood samples were analysed for coagulation studies with Sysmex® Ca 500v (Kobe, Japan). Biochemistry tests were conducted using Beckman-Coulter™ AU480 and a point-of-care meter was used for determination of lactate levels (StatStrip Xpress® Lactate Hospital Meter, Nova Biomedical, USA).

Statistical analyses.
All analyses were carried out in R Studio version 1.2.1335 45 using arsenal, finalfit, pROC, cutpointr, ggplot2 and ggupset packages. Pairwise deletion was used in analysing the data with number of observations for each variable listed in Supplementary Table S6 online. Normally and non-normally distributed continuous data were compared using independent t-test and Mann-Whitney test respectively. Parasite densities were natural logarithm (ln) transformed before analysis. Proportions were compared using Fisher's exact test with post hoc pairwise comparison using Bonferroni correction. Ordinal logistic regression was used to detect differences between subpopulations with normal, abnormal and defining criteria for severe disease severity, and adjustment for confounders. The categorisation of variables used in ordinal logistic regression was normal (1), abnormal but not severe (2) and severe (3) as summarised in Supplementary Table S7 online. Logistic and continuous linear regression were used to determine associations with outcome after adjusting for confounders. In continuous linear regression, non-normal data was transformed with natural logarithm (ln). Receiver-operating characteristic (ROC) and area under the ROC curve (AUC) were used to determine reliability of predictors and Youden's index (equal weight was given to sensitivity and specificity) was used to determine optimal cut-off for variables. The optimal cut-off for these variables was analysed in multivariable logistic regression controlled for age, sex and ln (parasitaemia). All analyses considered a p value of < 0.05 (two-tailed) as significant.