Overview of three influenza seasons in Georgia, 2014–2017

Background Influenza epidemiological and virologic data from Georgia are limited. We aimed to present Influenza Like Illness (ILI) and Severe Acute Respiratory Infection (SARI) surveillance data and characterize influenza viruses circulating in the country over three influenza seasons. Methods We analyzed sentinel site ILI and SARI data for the 2014–2017 seasons in Georgia. Patients’ samples were screened by real-time RT-PCR and influenza viruses isolated were characterized antigenically by haemagglutination inhibition assay and genetically by sequencing of HA and NA genes. Results 32% (397/1248) of ILI and 29% (581/1997) of SARI patients tested were positive for influenza viruses. In 2014–2015 the median week of influenza detection was week 7/2015 with B/Yamagata lineage viruses dominating (79%); in 2015–2016—week 5/2016 was the median with A/H1N1pdm09 viruses prevailing (83%); and in 2016–2017 a bimodal distribution of influenza activity was observed—the first wave was caused by A/H3N2 (55%) with median week 51/2016 and the second by B/Victoria lineage viruses (45%) with median week 9/2017. For ILI, influenza virus detection was highest in children aged 5–14 years while for SARI patients most were aged >15 years and 27 (4.6%) of 581 SARI cases died during the three seasons. Persons aged 30–64 years had the highest risk of fatal outcome, notably those infected with A/H1N1pdm09 (OR 11.41, CI 3.94–33.04, p<0.001). A/H1N1pdm09 viruses analyzed by gene sequencing fell into genetic groups 6B and 6B.1; A/H3N2 viruses belonged to genetic subclades 3C.3b, 3C.3a, 3C.2a and 3C.2a1; B/Yamagata lineage viruses were of clade 3 and B/Victoria lineage viruses fell in clade1A. Conclusion In Georgia influenza virus activity occurred mainly from December through March in all seasons, with varying peak weeks and predominating viruses. Around one third of ILI/ SARI cases were associated with influenza caused by antigenically and genetically distinct influenza viruses over the course of the three seasons.


Introduction
Influenza viruses affect people of all ages and cause mild to severe disease sometimes leading to fatal outcomes [1][2][3][4]. Surveillance of Influenza Like Illness (ILI) and Severe Acute Respiratory Infection (SARI) represent an important tool for tracking trends of virus spread and changes in globally circulating influenza viruses [5]. The first pandemic of the 21 st century caused by influenza virus A/H1N1pdm09 demonstrated the importance of influenza surveillance worldwide. The monitoring of changes in seasonal influenza viruses plays a critical role in defining influenza vaccine composition [6,7]. Moreover, data obtained from surveillance systems enables healthcare authorities to better understand timings of influenza activity and consequent mobilization of resources (e.g. vaccines, antivirals), notably prioritized vaccination of risk groups among other measures. However, limited data are available on influenza morbidity and mortality in Georgia [8][9][10] and none of these publications describe antigenic and genetic characteristics of seasonal influenza viruses circulating in the country. Georgia is located in the Caucasus region and covers a territory of 69700 km 2 with a population of around 3.7 million people. Prior to 2006, a nationwide population-based surveillance system for influenza and upper respiratory tract infection provided influenza data based only on clinical diagnosis without laboratory confirmation. In 2007, in collaboration with US CDC and other international stakeholders, sentinel surveillance of ILI and SARI was initiated. The National Influenza Center (NIC) at the National Center for Disease Control and Public Health, Georgia (NCDC&PH) now screens specimens collected at ILI/SARI surveillance sites all year round to monitor activity of influenza viruses.
In this study we describe epidemiological and virologic data for three consecutive influenza seasons obtained from ILI/SARI surveillance systems. The objectives were to: (1) define periods of influenza activity in Georgia; (2) assess the proportions of influenza infections among ILI and SARI cases; (3) determine most affected age groups; (4) describe epidemiological characteristics of influenza-associated fatal cases; and (5) determine antigenic and genetic profiles of influenza viruses circulating in the country.

Setting and study design
ILI surveillance was carried out in one outpatient clinic with a predetermined catchment of approximately 60,000 residents in the capital city Tbilisi (5% of 1.2 million population). The Children's Central Hospital, the largest children's clinic of the country also in Tbilisi, and four hospitals in Kutaisi (the biggest city in Western Georgia) were selected as SARI surveillance sites (S1 Fig). Medical staff and epidemiologists at each site were trained on case definitions, specimen collection, storage and transportation, completion of individual patient questionnaires and ILI/SARI aggregated forms. and influenza viruses from various countries, were downloaded from the EpiFlu database of the Global Initiative on Sharing All Influenza Data (GISAID) (http://www.gisaid.org). Nucleotide and deduced amino acid sequences were aligned and analyzed using BioEdit (http://www. mbio.ncsu.edu/BioEdit/bioedit.html) and maximum likelihood phylogenetic trees were estimated using RaxML v8.2X with a GTRGAMMA substitution model (https://sco.h-its.org/ exelixis/software.html), followed by annotation with amino acid substitutions defining nodes and individual virus gene products using treesub (https://github.com/tamuri/treesub/blob/ master/README.md). Phylogenies were based on full-length open-reading-frames with the signal peptide component being removed from HA genes and stop codons removed from all genes: H1-HA 1647, H3-HA 1650, N1-NA 1407, N2-NA 1407, B/Vic-HA 1710, B/Yam-HA 1707, B/Vic-Na 1398 and B/Yam-NA 1398 nucleotides, respectively. Trees were visualized using FigTree (http://tree.bio.ed.ac.uk/software/figtree/) and highlighted using Adobe Illustrator CC 2015.3 (http://www.adobe.com/uk/products/illustrator/features.html). Sequence accession numbers for all reference and test viruses are given in S1 Table. Ethical considerations Written informed consent was not required for this study. Verbal consent from patients and guardians in the case of children was sufficient for sample and data collection as the study was carried out under Georgia's routine public health surveillance practice.

Statistical analysis
Fisher's exact test was used for statistical analysis of data with chi squared estimation of significance. P values of 0.05 were considered as statistically significant. Data analysis was performed using Epi Info (version 7). Median, range, and interquartile range (IQR) were calculated to assess continuous variables. As Georgia is in the Northern Hemisphere, an influenza season is defined as the time period from week 40 of each year to week 20 of the following year, when the vast majority of influenza detections are made, and only data for these periods were analyzed. ILI consultation rates per 100 000 population and SARI vs total hospital admissions were used for evaluating season dynamics. The seasonal threshold was calculated using WHO guidelines and methodology [5].

Influenza activity in Georgia
A total of 12,912 ILI and 10,218 SARI patients met the ILI/SARI case definitions at sentinel sites during three consecutive influenza seasons 2014-2017; of these, 1,248 (9.7%) ILI and 1,997 (19.5%) SARI patients were sampled and tested for influenza. Overall 30% (978/3,245) specimens were positive: 397 (32%) ILI and 581 (29%) SARI (including four specimens: three ILI and one SARI-positive for both influenza A and B viruses).

ILI surveillance
Of the 397 influenza virus-positive ILI samples over three seasons (2014-2017) 132 (33%) were A/H1N1pdm09, 121 (30%) were A/H3N2 and 147 (37%), influenza B (three were positive for both influenza A and B viruses). During influenza activity periods weekly outpatient visits for ILI rose in parallel with laboratory-confirmed influenza virus detection (Fig 2). Based on our data seasonal threshold was determined as an ILI referral of 231/100,000. The 2014-2015 influenza season had an unusually high referral of ILI cases which was considerably earlier than laboratory-confirmed influenza virus detection, presumably due to circulation of other respiratory viruses.  Only 18 of 1248 swabbed ILI participants had been vaccinated during the three seasons; of these, eight persons became ill with influenza A/H3N2 and two with A/H1N1pdm09.

Influenza-associated fatal cases
During the three influenza seasons, 27 (4.6%) patients out of 581 laboratory-confirmed influenza SARI cases died, all of whom were patients in Kutaisi SARI sentinel hospitals. Of the fatal cases 19 (70%) were male and 8 (30%) female. Influenza A/H1N1pdm09 was associated with  Overview of three influenza seasons in Georgia, 2014-2017 patients received antiviral treatment within 48 hours after the onset of clinical signs, the median time between disease onset and antiviral prescription was 6 days (range 1-13, IQR 4-7 days). The median time between symptomatic illness onset and influenza-associated death was 6 days (range 3-44, IQR 1-13 days). Only one patient was vaccinated against influenza, a 73 year old male resident of a disabled persons center suffering with mental and kidney health problems, who subsequently died.

Antigenic and genetic characterization of influenza viruses
A/H1N1pdm09. Eighteen (86%) of 21 viruses were successfully recovered by WHO CC London and studied by HI assay. All but two viruses were well recognized in HI assays, within 2-fold of the homologous titer, by antiserum raised against the vaccine virus, A/California/7/ 2009; A/Georgia/567/2015 and A/Georgia/791/2016 were recognized at titers 4-and 16-fold lower respectively. Both these viruses also showed reduced recognition, compared to the respective homologous titers, by antisera raised against several reference viruses. While A/ Georgia/567/2015 carried an HA1 substitution (N156S; Fig 4) in a position known to affect antigenicity [13]

Discussion
We present an analysis of data obtained from ILI and SARI sentinel surveillance sites and virus characterization to provide an overview of three consecutive influenza seasons in Georgia. The majority of laboratory-confirmed influenza detections were seen from December through to March, but seasonal peaks occurred in different weeks and coincided with those observed in the European region [14][15][16]. Influenza virus dominance was similar to that in Europe but for the 2014-2015 season when influenza B predominated in Georgia and the Ukraine, A/H3N2 viruses predominated in other European countries or co-dominated with influenza B [17][18][19][20].
During a season approximately one third of both ILI and SARI cases sampled by sentinel sites in Georgia were associated with influenza virus infection. Increased ILI consultation rates and vaccine viruses against which post-infection ferret antisera were raised for use in HI assays are in bold type. The scale bar represents nucleotide substitutions per site. https://doi.org/10.1371/journal.pone.0201207.g005 Overview of three influenza seasons in Georgia, 2014Georgia, -2017 at outpatient clinics correlated with raised numbers of laboratory-confirmed influenza infections. The same trend was observed for SARI hospital admissions: the proportion of SARI cases among total hospitalizations doubled during peak weeks of influenza activity compared to weeks with low or no influenza detections. The number of SARI hospitalizations was lowest in 2014-2015, an influenza B season, and highest in 2015-2016 when A/H1N1pdm09 viruses predominated.
Children under 5 years of age were influenza virus-positive less frequently compared to older patients, reflecting the reports of a recent Egyptian study [21]. In our study the influenza confirmation rate was highest among ILI patients aged 5-14 years while the proportions of influenza-associated SARI cases were highest in the age groups 15-29 and 30-64 years. Such an age distribution might be influenced by the fact that during weeks of no or low influenza activity referral and/or sampling of children, notably those under 5 years of age, was higher compared to older patients; the rate of influenza detections in children could be lowered by this increased referral compared to adults.
Among ILI patients those aged 5-14 years were most often infected with influenza A/H3N2 or B viruses, while A/H1N1pdm09 infection was most commonly observed in individuals aged 30-64 years; similar observations have been described in Bulgaria [22]. The proportion of SARI cases testing positive for influenza A/H3N2 was highest in the age group 30-64 years, influenza B related hospitalization was mainly observed in the age groups 5-14 and 15-29 years but the majority of viruses were not ascribed to a lineage, while A/H1N1pdm09 viruses were detected most frequently in adults aged 15-29 and 30-64 years.
Of influenza-confirmed SARI cases in Georgia approximately 5% died while the percentage of fatal outcomes among hospitalized influenza-confirmed cases has been reported to vary from 0 to 4% in other countries [21,[23][24][25]. A fatal outcome was predominantly associated with A/H1N1pdm09 infection and patients in the age group 30-64 years had the highest probability of a fatal outcome. Our data are consistent with findings from other countries showing that adults were at higher risk of death than younger patients when infected with A (H1N10pdm09 viruses [25][26][27][28]. The vast majority of fatal cases suffered with at least one chronic condition, with cardiovascular and neurological disorders being the leading comorbidities. In a previous study based on data collected during the 2009 pandemic and post-pandemic seasons in Georgia, the major risk factors for a fatal outcome were lung disease, heart disease and pregnancy [10]. Antigenic and genetic characterization of influenza viruses from Georgia mostly showed similarities to strains circulating worldwide. Since their emergence A/H1N1pdm09 viruses have evolved forming eight genetic groups [29]. Viruses of recent years mainly fall into genetic group 6 which has three sub-divisions 6A, 6B and 6C. As observed globally, Georgian viruses from 2014-2015 and the early part of 2015-2016 seasons belonged to the 6B clade while those later in the 2015-2016 season fell into subgroup 6B.1 [29]. The 6B and 6B.1 viruses carried several mutations in HA and NA genes compared to the A/California/7/2009 vaccine virus. Five HA amino acid substitutions were located in four antigenic sites (Sa, Sb, Ca1, Ca2) while substitution S185T (190 loop) and residue 222 polymorphism in one virus were in the vicinity of the receptor-binding site; such substitutions can affect the antigenic properties of A/ H1N1pdm09 viruses [30]. However, the majority of A/H1N1pdm09 viruses from Georgia, and those detected worldwide, remained antigenically similar to the A/California/7/2009 vaccine virus as assessed with post-infection ferret antisera.
Over the period of this study A/H3N2 viruses evolved rapidly genetically and showed some antigenic change, antigenic drift, such that three vaccine viruses were recommended: who.int/influenza/vaccines/virus/recommendations/en/). In 2014-2015, viruses from Georgia fell in genetic clades 3C.2a and 3C.3b while the majority of A/H3N2 viruses detected in the northern hemisphere belonged to clades 3C.2a and 3C.3a and the vaccine was reported to have low effectiveness [31,32]. Similar clade mismatching was observed in 2015-2016 (3C.2a1 and 3C.3a) and 2016-2017 (3C.2a1) in Georgia, with 3C.2a1 subclade viruses predominating in some other parts of the world during both seasons [33]. Worldwide multiple mutations were detected in A/H3N2 viruses with many encoding clade defining HA amino acid substitutions, some of which were located in antigenic sites A (7), B (7), D (2) and E (4) and/or altered HA glycosylation patterns; modifications known to cause antigenic drift [34,35]. However, the great majority of 3C.3a, 3C.2a and 3C.2a1 genetic group viruses yielded poor/no agglutination of RBCs, making antigenic characterization by HI impossible. Based on those that could be analyzed by HI  [37]. Genetic studies of the B/Victoria lineage viruses revealed several substitutions in the HA and NA. Circulating viruses did not share the egg-adaptive substitution N197K in HA1, observed in egg-propagated B/Brisbane/60/2008 vaccine virus, but had the substitutions I117V and N129D in HA1. There were a greater number of amino acid substitutions in the NA but the significance, if any, of these substitutions is unknown.
Our study had several limitations that could influence the results of analyses performed. False negative laboratory results may have been obtained due to the late referral of patients to clinics. Individual questionnaires were filled out during sampling and subsequent data regarding course of disease were not collected so we were unable to assess disease severity characteristics (developing pneumonia, need for ICU, etc.). For the same reason, we did not evaluate chronic conditions among non-fatal ILI/SARI cases to determine risk factors associated with any severe courses of disease.

Conclusions
Despite limitations, this study contributes to a much better understanding of the epidemiological and virologic characteristics of influenza in Georgia. Influenza virus activity in the country was mainly observed from December through March each season with varying peak weeks and predominant viruses. On average one third of patients with ILI/SARI screened in each season had influenza virus infection. Among ILI cases influenza detection was highest in patients aged 5-14 years while the highest proportions of influenza-confirmed SARI cases were among adults. Persons aged 30-64 years had a higher risk of a fatal outcome. The observed circulation of antigenically and genetically variable viruses in Georgia and selection of A/Georgia/532/ 2015 as a reference virus by WHO CC, London illustrates the need for routine surveillance to contribute to the work of the WHO Global Influenza Surveillance and Response System.