Next-generation sequencing reveals viral aetiologies of encephalitis in Ghana: a prospective cross-sectional study

Abstract

In Ghana, cerebrospinal fluid polymerase chain reaction (PCR) is the primary diagnostic tool for viral encephalitis. However, its application remains limited, and diagnosis is predominantly syndromic. Current data on encephalitis in Ghana are sparse and often restricted to sporadic cases, with improved PCR diagnostics elucidating some aetiological agents but leaving approximately 60% of cases undiagnosed. This diagnostic gap arises because PCR targets specific pathogens, overlooking unexpected or novel agents. Consequently, tailored patient management is hindered, patient outcomes are adversely affected, and efforts to understand the true burden of viral encephalitis are impeded. To address these challenges, we conducted a cross-sectional study at the Komfo Anokye Teaching Hospital, a major tertiary referral centre in Ghana, from May 2019 to August 2022. Forty-three (43) cerebrospinal fluid samples from suspected encephalitis patients were analysed using Polymerase chain reaction (PCR) and high-throughput next-generation sequencing (NGS) to identify a broader range of viral pathogens and assess the role of co-infections in disease outcomes. Viral encephalitis was detected in 42% (18/43) of samples, revealing 11 viruses across six families. Herpesviruses (34%) and retroviruses (28%) were the most prevalent, with human immunodeficiency virus-1 (HIV-1) and enteroviruses identified as the primary causative agents. NGS identified a broader viral spectrum, detecting 6 additional viruses (HIV-1, EBV, Mumps, HHV-6, HRV, Rotavirus A) which were not targeted by PCR. While NGS and PCR demonstrated comparable sensitivity for certain pathogens, NGS identified additional co-infections (39%, 7/18) and rare viruses. The survival rate for patients with co-infections was 28.6% (2/7), with HIV-1 and herpesviruses implicated in 85.7% (6/7) of co-infected cases. This study highlights the potential of NGS to expand the diagnostic capabilities for viral encephalitis. By overcoming the limitations of PCR, NGS provides a more comprehensive approach, enhancing our understanding of the true burden of viral encephalitis in Ghana and informing better patient management strategies.

Background

Encephalitis is a serious inflammation of the brain parenchyma leading to high grade fever, altered mental status, and other neurological presentations. It is associated with high morbidity and mortality and constitutes a significant global public health challenge. With an estimated incidence of 3.5–7.4 cases per 100,000 people globally, encephalitis ranks among the top causes of paediatric mortality in sub-Saharan Africa and imposes substantial economic costs through hospitalizations, highlighting its severe public health burden [1].

The disease affects individuals across all age groups and gender, but younger patients and immunocompromised individuals are particularly vulnerable [2].

The actual global burden of encephalitis is very difficult to estimate. However, it is believed that the greatest burden of the disease is found in low-income African countries with under-resourced health care systems, poor immunisation plans, and data gaps [3].

In Ghana, and similarly in many low-income countries, the lack of appropriate diagnostic techniques has contributed to significant data gaps in understanding the aetiologies of encephalitis. Diagnosis has often been presumptive, leading to inaccurate or delayed identification of causative agents and, consequently, delayed initiation of appropriate treatment and longer hospital stays and associated healthcare costs. This has exacerbated the disease burden in these settings.

Encephalitis may be caused by a myriad of agents; however, the commonest pathogenic mechanism is infection with a heterogenous group of viruses. Although cerebrospinal fluid (CSF) Polymerase Chain Reaction (PCR) has been the mainstay for the diagnosis of viral causes of encephalitis, its wide application is limited in Ghana and diagnosis has predominantly been syndromic. The available reports on viral encephalitis are only focused on sporadic cases [4, 5]. To paint a better epidemiological picture, we previously conducted a PCR based study that identified specific viruses such as Cytomegalovirus (CMV), Enterovirus (ENTV), Herpes simplex virus (HSV), Varicella zoster virus (VZV), and Rabies virus (RABV) as causes of encephalitis among 40% (31/77) of our study patients [6]. While improved PCR diagnostics have shed light on some viral agents of encephalitis in Ghana, a significant proportion (about 60%) of cases remained undiagnosed. This “diagnostic gap” arises because PCR is designed to target specific, known viruses, leaving a vast array of potential pathogens undetected. This hampers accurate diagnosis, tailored patient management and, exacerbates antimicrobial resistance due to indiscriminate use of broad-spectrum antibiotics. Therefore, further calls for unbiased testing of patients’ samples for the identification of new or unexpected pathogens are warranted.

Next generation sequencing (NGS) is a revolutionary high throughput technique for determining the DNA/RNA sequences in a given sample for genetic variation determination and aetiological identification [7]. This approach overcomes the limitations of targeted diagnostic methods as it requires no prior knowledge or assumptions about the type of pathogen causing infection, therefore enabling detection of novel and unexpected pathogens [8], and possibly improving patient outcomes. Recently, the use of NGS has allowed the detection of known viruses and the discovery of novel viruses in clinical samples, and several studies have reported on its utility in aetiology identification [4, 5, 9, 10, 11]. However, most published work to date has been conducted in high-income settings, with limited prospective studies from sub-Saharan Africa assessing the comprehensive viral aetiology of encephalitis using NGS. We investigated CSF samples from suspected encephalitis patients using NGS to uncover a broader spectrum of viral pathogens associated with encephalitis and identified viral co-infections that may impact patient outcomes.

Methods

Study setting

This cross-sectional study investigated the diverse landscape of encephalitis in Ghana through a prospective design with short term outcome assessment. Patient recruitment was done at the Komfo Anokye Teaching Hospital (KATH). The hospital is located in Kumasi, the capital of Ghana’s Ashanti Region, which has a total population of 5,432,485. As a major referral centre, KATH serves not only the Ashanti Region but also parts of the northern, western, and central regions of Ghana, extending its catchment population well beyond the regional figure. We recruited patients who presented with signs and symptoms suggestive of encephalitis using a non-randomised strategy. Patient recruitment spanned a period of over 3 years between May 2019 and August 2022.

Inclusion criteria

Patients with the following conditions were included in the study:

• Altered mental status (e.g., personality change, altered consciousness, or lethargy) lasting > 24 h without an alternative cause,

• Documented fever within 72 h before or after neurological presentation,

• Partial or generalized seizures not linked to a preexisting seizure disorder,

• New-onset of focal neurological deficits.

Exclusion criteria

Patients with the following conditions were excluded from the study:

• Pre-existing neurological conditions,

• Stroke or cranial trauma,

• Neuroinvasive parasitic infections,

• Confirmed bacterial meningitis,

• Cerebral malaria.

Sampling

Cerebrospinal fluid collection was done by experienced clinicians, adhering to rigorous standard operating procedures (SOP). Prior to lumbar puncture, the patients were evaluated to ensure the absence of any contraindications [12]. Serial 1 ml aliquots of CSF were collected into sterile 2 ml cryotubes (Sarstedt, Germany) for viral testing. The CSF samples were immediately transported under cold chain conditions to the laboratory at the Kumasi Centre for Collaborative Research (KCCR), Kumasi, for initial specific viral screening for Herpes simplex viruses, varicella zoster virus, cytomegalovirus, enterovirus, rabies virus, and Kumasi Rhabdovirus. Aliquots of the CSF samples were batched and sent to the Institute of Virology, Charité, Universitätsmedizin Berlin, Germany, for genome sequencing and subsequent viral identification.

Laboratory analysis

Viral RNA was extracted from CSF samples using the spin protocol of Qiagen viral RNA mini kit (Qiagen, Hilden-Germany) according to manufacturer’s instructions. The extracted nucleic acid served as the template for quantitative PCR (qPCR) testing of a panel of six viruses following established protocols [6].

We utilized the KAPA RNA Hyper Prep Kit (Roche Molecular Diagnostics, Basel-Switzerland) for library preparation and performed high-throughput genome sequencing with the 150-cycle NextSeq reagent v3 cartridge (Illumina, San Diego, California, U.S.), following the manufacturer’s instructions. The obtained reads underwent quality trimming and virus identification in a bioinformatics pipeline available at the institute of virology, Charité. The results were independently confirmed by BLAST analysis of individual reads using DIAMOND against a curated NCBI GenBank viral protein database, thereby excluding non-viral sequences from the analysis.

Data management and statistical analysis

Patients’ information captured by the paper-based case report forms were inputted into the Research Electronic Data Capture (REDCap) web application (Vanderbilt University, TN-USA), hosted by the School of Medicine and Dentistry of KNUST. Subsequently, all data were exported to Microsoft Excel version 2019 (Microsoft Corporation, Washington, USA) and cleaned. Data analyses were carried out using the R statistical software version 4.2.1 (The R Foundation, Auckland, USA). Nominal characteristics, such as patients’ sex and symptoms, were described using summary frequencies and proportions. Continuous variables were described using median, mean, and interquartile range based on their distribution.

Results

The study analysed CSF samples from 43 patients, selected from an initial pool of 77 specimens received during the study period, after preliminary screening for sample quality, volume sufficiency, and completeness of clinical data. The study cohort primarily consisted of younger individuals (not more than 21 year) (67%, n = 29) and males (63%, n = 27), with an age range of 5 months to 68 years. Approximately 40% of participants (n = 17) had reported some form of animal contact, while only 2% (n = 1) had a family history of neurological infections. The most common symptoms were fever (86%, n = 37), seizures (67%, n = 29), lethargy (60%, n = 26), and headaches (60%, n = 26). Neurological signs included hydrophobia (9%, n = 4), cognitive dysfunction (16%, n = 7), behavioural changes (19%, n = 8), stiff neck (44%, n = 19), and neck pain (35%, n = 15). Notably, 8 patients (19%) were immunosuppressed due to HIV infection (Table 1).

Table 1.

Demographic and clinical characteristics of the patients

Characteristic	Frequency (%), N = 431
Age, median (IQR), years	15 (6, 36)*
Gender
Female	16 (37%)
Male	27 (63%)
Duration of admission (days)	10 (6, 15) *
Population
> 21 years	14 (33%)
≤ 21 years	29 (67%)
Animal contact	17 (40%)
Symptoms
Fever	37 (86%)
Seizures	29 (67%)
Headaches	26 (60%)
Lethargy	26 (60%)
Nuchal Rigidity	19 (44%)
Vomiting	18 (42%)
Neck Pain	15 (35%)
Sepsis Suspected	15 (35%)
Diarrhoea	11 (26%)
Behavioral Changes	8 (19%)
Cognitive Dysfunction	7 (16%)
Hydrophobia	4 (9%)
30 days mortality	19 (44%)
HIV infection	8 (19%)

^*Median (interquartile range), ¹n (%)

CSF analysis by NGS testing

Viral nucleic acid was detected in 42% (18/43) of the cerebrospinal fluid samples. We identified 11 viruses across six viral families from the 43 samples tested. Herpesviruses were the most prevalent, accounting for 35% (n = 10) of the encephalitis cases, followed by HIV-1 of the Retrovirus family at 28% (n = 8) and Picornaviruses at 21% (n = 6). Despite the high occurrence of Herpesviruses, the most frequently detected virus was human immunodeficiency virus-1 (HIV-1), responsible for 28% (n = 8) of cases, and enteroviruses, which contributed to 17% (n = 5). Herpes simplex virus, cytomegalovirus, and rabies virus each accounted for 10% (n = 3) of the total viruses identified. Other viruses detected included Epstein-Barr virus (7%, n = 2), Varicella Zoster virus (3%, n = 1), Human Herpesvirus 6 A (3%, n = 1), Human Rhinovirus (3%, n = 1), Mumps virus (3%, n = 1), and Rotavirus A (3%, n = 1) (Table 2). Of the eight cases confirmed as HIV-1 positive, only five (63%) had been previously identified through rapid diagnostic testing (RDT) (Table 2).

Table 2.

Distribution of viruses identified by NGS

Viral family	Virus	Frequency N = 29	Cumulative percentage
Herpesviridae	Herpes simplex virus (HSV)	3 (10%)	35
	Cytomegalovirus (CMV)	3 (10%)
	Epstein Barr virus (EBV)	2 (7%)
	Human herpes virus 6 (HHV-6)	1 (3%)
	Varicella zoster virus (VZV)	1 (3%)
Retroviridae	Human immunodeficiency virus 1 (HIV-1)	8 (28%)	28
Picornaviridae	Enterovirus (ENTV)	5 (17%)	21
	Human rhinovirus (HRV)	1 (3%)
Rhabdoviridae	Rabies virus (RABV)	3 (10%)	10
Paramyxoviridae	Mumps	1 (3%)	3
Reoviridae	Rotavirus A (RV)	1 (3%)	3

Comparative performance of NGS and PCR in viral detection

Out of the 43 cerebrospinal fluid (CSF) samples analyzed, 18 (41.9%) yielded at least one viral pathogen through PCR and/or NGS, while 25 samples (58.1%) remained negative for all viral targets tested by both methods. Compared with NGS, PCR detected 1 additional case of ENTV, 1 additional case of RABV, and 3 additional cases of CMV. Both methods detected HSV (10%) and VZV (3%) at similar rates. However, NGS identified a broader range of viruses not targeted by PCR, including HIV (28%), EBV (7%), HRV (3%), mumps virus (3%), rotavirus A (3%), and HHV-6 (3%). (Table 3).

Table 3.

Comparative detection rates of viruses using NGS and PCR

	NGS	PCR
Virus	N = 291	N = 221
CMV	3 (10%)	8 (36%)
HSV	3 (10%)	3 (14%)
ENTV	5 (17%)	6 (27%)
VZV	1 (3%)	1 (5%)
RABV	3 (10%)	4 (18%)
Viruses detected by only NGS
HIV	8 (28%)
EBV	2 (7%)
HRV	1 (3%)
Mumps	1 (3%)
Rotavirus A	1 (3%)
HHV6	1 (3%)

Type of infection and patient outcome

Seven distinct co-infections involving multiple viral pathogens were identified, with the lowest observed survival rate of 28.6% (2/7). HIV-1 was implicated in 85.7% (6/7) of these co-infections. In comparison, patients with single viral infections had a markedly higher survival rate of 72.7% (8/11). Among the 18 patients with any viral detection (combined PCR and NGS), the overall survival rate was 44.4% (8/18), while patients with no detectable viruses had a survival rate of 36.0% (9/25). The overall survival rate for all patients in the cohort was 55.8% (24/43) (Table 4).

Table 4.

Survival rates among encephalitis patients by infection type and viral detection status

Type of infection		Number of viruses	Alive, N = 241 55.8	Dead, N = 191 44.2
Co-infection
HIV/CMV/HSV/HRV		4	0 (0.0)	1 (100.0)
HIV/CMV/ENTV/EBV		4	1 (100.0)	0 (0.0)
HIV/CMV/Mumps		3	0 (0.0)	1 (100.0)
HIV/VZV		2	0 (0.0)	1 (100.0)
HIV/RV		2	0 (0.0)	1 (100.0)
HIV/EBV		2	0 (0.0)	1(100.0)
ENTV/HHV-6		2	1 (100.0)	0 (0.0)
Total Co-infections		–	2 (28.6)	5 (71.4)
Single Viral Infections		1	8 (72.3)	3 (27.7)
All Positive Patients		–	8 (44.4)	10 (55.6)
No Detectable Virus		–	9 (36.0)	16 (64.0)

¹n (%)

Discussion

This study examined CSF samples from suspected encephalitis patients using NGS to uncover a broader viral landscape, identify frequent co-infections contributing to disease complexity, and explore their potential impact on patient outcomes. A total of 77 samples were initially received, of which 43 met inclusion criteria for NGS analysis based on quality, volume, and clinical data completeness. A diverse array of 11 viruses from six families were identified, with Herpesviridae (35%), Retroviridae (28%), and Picornaviridae (21%) being the most prevalent. Notably, Human Immunodeficiency Virus-1 (HIV-1) was the most frequent pathogen, followed by enteroviruses. Co-infections involving HIV-1 and Cytomegalovirus (CMV) were also common.

The identification of a wide spectrum of viral pathogens underscores the complexity of encephalitis aetiology and the limitations of relying solely on targeted PCR diagnostics. While PCR remains the gold standard for rapid and specific pathogen detection [13], it is inherently limited to pre-selected targets. In contrast, NGS offers an unbiased diagnostic approach capable of detecting both expected and novel pathogens especially in cases of unexplained encephalitis [14, 15]. In our study, NGS served as a retrospective diagnostic tool to identify 11 viral targets compared to 5 by PCR, representing a 120% increase in pathogen detection. This difference does not reflect a direct comparison of sensitivity but rather the expanded scope enabled by NGS. These findings are consistent with previous studies that demonstrate the added value of NGS in pathogen discovery and epidemiological surveillance [16, 17, 18, 19].

NGS identified rare or unexpected pathogens such as Human Rhinovirus, Rotavirus, Epstein-Barr Virus, mumps virus, and Human Herpesvirus 6 A (HHV-6 A), which have not been widely reported in encephalitis studies [20, 21, 22]. The high prevalence of Herpesviruses including HSV, CMV, and VZV, also reinforces their established clinical relevance in CNS infections [23, 24, 25, 26]. Importantly, these findings support the potential of NGS to not only broaden diagnostic reach but also inform the expansion of region-specific PCR panels, improving future diagnostic strategies.

The detection of viral co-infections (n = 7) highlights the clinical complexity of encephalitis. These cases had a notably low survival rate (28.6%), suggesting more severe disease progression. HIV-1 was implicated in 85.7% (n = 6/7) of co-infected cases, reinforcing its role as a critical modulator of immune response and disease severity. While some of the detected viruses such as Epstein-Barr Virus and HHV-6 A are known to persist latently or be reactivated under immunocompromised conditions, others like CMV and HSV may reflect active neuroinvasive disease. Given that metagenomic sequencing does not differentiate between latent, active, or incidental viral presence, we cannot assert that all co-detections represent true co-infections. However, the clinical context, poor outcomes, and high burden of immunosuppression in our cohort support the likelihood that at least a subset of these co-infections were indeed clinically significant.

Patients with single viral infections had the highest survival rate (72.7%), whereas those with co-infections had the poorest (28.6%). These findings are consistent with prior research indicating worse outcomes in the context of immunomodulating viruses such as HIV-1 and CMV [27, 28]. Interestingly, patients with no viral pathogen detected had a survival rate of only 36.0%, which may reflect the presence of non-viral causes such as bacterial, fungal, or parasitic infections, autoimmune encephalitis, or late-stage disease masking pathogen detection. These undiagnosed cases highlight the need for future studies incorporating broader diagnostics.

In settings like Ghana where HIV rapid diagnostic tests (RDTs) are widely used, our finding that 37% of HIV-1 positive cases were missed by RDTs underscores a critical diagnostic gap. HIV-1 remains one of the few viral causes of encephalitis for which effective antiviral treatment exists [29, 30, 31] and timely initiation of treatment is crucial. NGS detected HIV-1 in several RDT-negative patients, which may have implications not only for clinical outcomes, but also for public health through earlier diagnosis and reduced transmission. In high-prevalence regions, more sensitive diagnostic methods could help close detection gaps and enhance treatment access. Our findings align with prior reports on the central role of HIV-1 in encephalitis morbidity, particularly in immunocompromised populations [32, 33]. Co-infections involving HIV-1 and other viruses, such as Cytomegalovirus, often result in worsened clinical outcomes [34], emphasizing the need for early detection and comprehensive management strategies. Enhancing diagnostic sensitivity through methods like NGS may thus play a pivotal role in improving both patient outcomes and broader public health efforts.

In contrast to studies that focus on single pathogens [35, 36, 37], our findings reflect a more diverse viral landscape, including the identification of less frequently implicated pathogens. For example, Liu et al. reported HSV and HIV as frequent causes of encephalitis [26], however, our study extends these findings by detecting a broader range of pathogens, including Human Rhinovirus, Rotavirus, and Epstein-Barr Virus, mumps virus, and human herpes virus 6; pathogens that have not been widely associated with encephalitis in earlier studies. These findings reinforce the need to expand existing diagnostic panels to accommodate regionally relevant viruses and co-infection profiles. The potential of NGS to inform this expansion is especially valuable in resource-limited settings.

Clinically, these results call for more comprehensive diagnostic protocols for encephalitis, particularly in high HIV burden areas. Although the integration of NGS into clinical workflows holds promise for improved diagnosis and treatment guidance, practical challenges remain. These include high costs, the need for technical expertise, and limited access to bioinformatics infrastructure. Addressing these barriers will require investment in simplified sequencing technologies, capacity building, and the establishment of regulatory and quality assurance frameworks for clinical metagenomics [38, 39].

This study was limited by the batching of samples prior to NGS processing, which meant results were not available in real time and could not inform immediate clinical decision-making. This delay may have contributed to missed opportunities for timely intervention, as supported by previous studies [40, 41]. Additionally, our exclusion of bacterial, fungal, parasitic, and autoimmune causes restricted our scope to viral pathogens. Finally, while viral nucleic acid detection in CSF supports potential involvement in disease, it does not alone confirm causation especially for viruses such as HIV, CMV, EBV, and HHV-6, which can be latent or reactivated. Future research should aim for real-time diagnostics and incorporate larger, more diverse cohorts along with broader pathogen detection strategies.

Conclusion

Our study represents the first comprehensive investigation of the viral landscape of encephalitis in Africa using next-generation sequencing (NGS). The findings underscore the complexity of encephalitis aetiology and the need to account for co-infections and rare pathogens in diagnostic protocols. The ability of NGS to detect a broader spectrum of viruses, including those overlooked by conventional PCR, demonstrates its potential to revolutionize diagnostic approaches. Notably, the data generated by NGS can be used to inform and expand existing PCR panels, enabling future diagnostics to be more inclusive of regionally relevant or previously unconsidered viral pathogens. This comprehensive diagnostic capability can guide more targeted treatment strategies, ultimately improving clinical outcomes for patients.

Despite the advantages of NGS, PCR remains indispensable for targeted, rapid, and sensitive detection of specific viral infections. The complementary strengths of NGS and PCR highlight the importance of integrating both methods into diagnostic workflows to address the diverse challenges of viral encephalitis diagnosis.

Abbreviations

PCR

Polymerase chain reaction

NGS

Next generation sequencing

HIV

Human immunodeficiency virus

EBV

Epstein barr virus

HHV-6

Human herpes virus-6

HRV

Human rhinovirus

CSF

Cerebrospinal fluid

CMV

Cytomegalovirus

ENTV

Enterovirus

HSV

Herpes simplex virus

VZV

Varicella zoster virus

DNA

Deoxyribonucleic acid

RNA

Ribonucleic acid

SOP

Standard operating procedures

BLAST

Basic local alignment search tool

NCBI

National center for biotechnology information

RV

Rotavirus

RABV

Rabies virus

RDT

Rapid diagnostic test

GLP

Good laboratory practice

GCP

Good clinical practice

CHRPE

Committee for human research, publication, and ethics

Author contributions

RY: Investigation, data curation, formal analysis, writing– original draft, and writing– review & editing. RG: Investigation, Data curation, formal analysis, writing– review & editing. PED: Investigation, data curation, formal analysis, writing– review & editing. JOM: Data curation, formal analysis, writing– review & editing. HKA: Investigation, Data curation, formal analysis, writing– review & editing. ENA: Investigation, writing– review & editing. MO: Supervision, formal analysis, writing– review & editing. YAA: Investigation, methodology, writing– review & editing. KOD: Supervision, writing– review & editing. CD: Conceptualisation, funding acquisition, writing– review & editing VMC: Conceptualisation, writing– review & editing. ROP, AAS: Conceptualisation, funding acquisition, project administration, writing– review & editing. All authors approved the final version of the manuscript.

Funding

This publication is part of the PANDORA-ID-NET (EDCTP Reg/Grant RIA2016E-1609), funded by the European and Developing Countries Clinical Trials Partnership (EDCTP2) programme, which is supported under Horizon 2020, the European Union’s Framework Programme for Research, and Innovation. The views and opinions of authors expressed herein do not necessarily state or reflect those of EDCTP.

Data availability

The clinical and experimental datasets that support the findings of this study are publicly available in Zenodo with the identifier https://doi.org/10.5281/zenodo.14633638. However, the viral genome sequences are currently being prepared for deposition in an appropriate public repository and will be made publicly available upon completion. Accession numbers will be provided in the final published version of this article.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the ethical principles as outlined in the Helsinki declaration [42] and consistent with Good Laboratory Practice (GLP) and Good Clinical Practice (GCP). The study protocols underwent thorough review and approval by the Committee for Human Research, Publication, and Ethics (CHRPE) at the School of Medicine and Dentistry of KNUST (CHRPE/AP/231/18). All participating patients provided written informed consent. In cases involving individuals below the legal consenting age (18 years) or incapacitated patients, informed consent was obtained from their parents, guardians, or legal representatives. Authors had no access to information that could identify individual participants during or after data collection.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.