Molecular phylogeny of a novel human adenovirus type 8 strain causing a prolonged, multi-state keratoconjunctivitis epidemic in Germany

The German infectious disease surveillance system revealed an increase of epidemic keratoconjunctivitis (EKC) from an average of 320 cases/year (2001 to 2010) up to 2146 and 1986 cases in 2012 and 2013, respectively. From November 2011 until December 2013 (epidemic period) 85% of typed isolates were human adenovirus type 8 (HAdV-D8), whereas only low level circulation (19%) of HAdV-D8 was observed outside the epidemic period. In order to investigate whether a novel monophyletic HAdV-D8 strain prevailed during the epidemic period, complete genomic sequences of 23 HAdV-D8 isolates were generated by deep sequencing and analyzed phylogenetically. For comparison, eight HAdV-D8 isolates from outside the epidemic period were sequenced. HAdV-D8 isolates of the epidemic period had a very high sequence identity of at least 99.9% and formed a monophyletic cluster with two subclusters. A single outlier was closely related to HAdV-D8 strains isolated prior to the epidemic period. Circulation of the epidemic strain was detected as early as 2010 but not after the epidemic period in 2014. In conclusion, molecular phylogeny of complete genomic sequences proved a monophyletic HAdV-D8 epidemic. However, co-circulation of other HAdV types as well as better reporting may have contributed to the huge increase of reported cases.

Because of the severity and public health impact of EKC, notification to local health authorities is mandatory for laboratories that detect adenoviruses from conjunctival swabs according to the German Protection against Infections Act (IfSG, §7.1) 10  nosocomial outbreaks with the same reporting path is mandatory (IfSG §6.3). The German notification system has been described previously in detail 12 .
Starting with a big nosocomial EKC outbreak caused by HAdV-D8 in November 2011 13 , adenovirus (kerato-) conjunctivitis case numbers reported to RKI increased from an average of 320 cases/year in the years 2001 until 2010 to 2146 cases in 2012 ( Table 1). Circulation of HAdV-D8 had not been observed in Germany from 2006 to 2009. Epidemiological data collected according to the IfSG were analyzed together with results of the reference laboratory (KLA, Konsiliarlabor für Adenoviren at the Hanover Medical School). Complete adenovirus genomic sequences of 31 EKC cases were generated in order to investigate that the circulation of a monophyletic HAdV-D8 strain caused the prolonged multi-state EKC epidemic observed in 2012 and 2013 (analyzed epidemic period from November 2011 to December 2013).
Patients infected with HAdV-D8 were significantly older and more often female than patients infected with other types (Table 3). Information on the presumed place of infection (setting) was provided only for 236 HAdV-D8 outbreak-cases and for 1187 other outbreak cases. Infections were considered healthcare-associated when the assumed place of infection was practice, medical treatment center or hospital. In contrast to outbreak cases caused by other or unknown types, most HAdV-D8 outbreak cases (233 of 236 (98.7%) with information, including all 209 cases in the large outbreak in North Rhine-Westphalia, November 2011) were healthcare associated (Table 3).
Of all HAdV-D8 outbreak cases reported to RKI from 2001 until 2015, 86.1% were reported during the epidemic period.
The geographical distribution of HAdV-D8 cases available at the KLA and the incidence of EKC in the federal states (RKI data) are depicted in Fig. 1. Typed cases at the KLA originated from eight of the sixteen German states (Bavaria, Berlin, Brandenburg, Baden-Wurttemberg, Lower Saxony, Mecklenburg-Vorpommern, North Rhine-Westphalia and Rhineland-Palatinate). No samples were received at the KLA from two states with a high EKC incidence (Saxony-Anhalt and Thuringia) but also no typing results were reported from these two states to the RKI.
HAdV DNA quantification. Adenovirus DNA concentrations (median 3.9 × 10 7 copies (genome equivalents)/ ml, 25th percentile 5.4 × 10 6 copies/ml, 75th percentile 2.5 × 10 8 copies/ml) of HAdV-D8 positive samples were significantly higher (p = 0.0065, Mann-Whitney test) compared to samples containing other types (median 2.3 × 10 6 copies/ml, 25th percentile 4.6 × 10 5 copies/ml, 75th percentile 2.6 × 10 7 copies/ml). All values relate to eye swab specimens resuspended in 1 ml sterile 0.9% NaCl.   Complete Genomic Sequences and Phylogenetic Analysis. The complete adenovirus genomic sequences of 31 HAdV-D8 cases were determined using next generation sequencing (NGS) including 23 cases from the epidemic period, 6 cases from 2003 to 2010, and two cases from 2014. All sequences from the epidemic period (with exception of a single outlier sequence) formed a monophyletic cluster, supported by a bootstrap value of 92% (Fig. 2). This monophyletic cluster consisted of two subclusters (A and B) and both were supported by bootstrap values of 100%. Subcluster A included sequences from the early epidemic period, with the last sequence from December 2012. In addition, subcluster A included all sequences from the November/December 2011 outbreak in North Rhine-Westphalia and the last pre-epidemic German sequences from Lower Saxony, 2010. Sequence diversity was higher in subcluster A (maximum diversity 0.07%) than in subcluster B (maximum diversity 0.02%). Circulation of subcluster B HAdV-D8 strains (first sample isolated in July 2012) overlapped with circulation of subcluster A HAdV-D8, both temporally and geographically. Two American HAdV-D8 strains isolated in 2012 (#KT340070 and #KT340056) clustered with the German sequences of the 2012/2013 epidemic but were neither associated with subcluster A nor with subcluster B. All these sequences were most closely related to the sequence of a clinical HAdV-D8 isolate from Japan (#AB448769.1), labelled as genome type 8e (Fig. 2).
A single outlier from Rhineland-Palatinate, December 2013 (#KP016741) did not cluster with the other sequences of the epidemic period but formed a second monophyletic cluster with German samples isolated in In silico restriction enzyme analysis (REA). Subcluster A and the two American sequences (#KT340070 and #KT340056) that clustered with German epidemic sequences showed an identical REA pattern to genome type 8e, whereas one additional SacI cleavage site was observed in subcluster B sequences (Fig. 3). Thus, the monophyletic cluster of sequences from the epidemic period comprised two different REA patterns. On the other hand, several other sequences which formed a second monophyletic cluster with the above mentioned outlier from the epidemic period (#KP016741) had an in silico REA pattern identical to genome type 8E and the subcluster A sequences.
Common and unique features of the HAdV-D8 sequences from the epidemic. Genomic sequences of the monophyletic cluster from the epidemic period had 40 single nucleotide polymorphisms in common compared to the HAdV-D8 prototype (Trim strain), and revealed a very high nucleotide sequence identity (≥ 99.85%) to each other.
Nucleic acid sequences and deduced amino acid sequences of the major capsid proteins (penton base, hexon, fiber) were found to be highly conserved between the German epidemic period samples. Compared to the HAdV-D8 prototype, all samples had only 3 amino acid substitutions (Leu809Val, Phe811Leu, Thr819Ala) in the hexon, which were not located in the hypervariable loops of the neutralization determinant. Only a few point mutations were noted: A single sample of the November/December 2011 outbreak in Northrhine-Westphalia (# KP016722) had a non synonymous point mutation (Asp270Asn) in the loop1 of the neutralization determinant. Another sample (Brandenburg, 2012/09, #KP016730) had a single amino acid substitution (Met494Ile) in the hexon but not located in its neutralization determinant. This sample also had a non synonymous fiber mutation (Asp102His). The only other fiber mutation (Glu350Gln) was found in a sample from Lower Saxony, 2012/03 (#KP016727). Two samples originating from Northrhine-Westphalia isolated in 2013/01 (#KP016738) and 2013/02 (#KP016740) lacked the CR1-alpha open reading frame due to a frameshift mutation caused by a 5 bp deletion, in addition to a single nucleotide insertion in the CR1-gamma (29.1 kDa protein) ORF causing a frameshift mutation and thus truncation of the protein. As these gene products are not essential for HAdV replication in cell cultures, it should be pointed out that these deletions were found in genomic sequences derived directly from eye swabs without virus isolation on cell culture.
On the other hand, the outlier sequence (#KP016741), which was not part of the monophyletic cluster of the epidemic period, had a 9 bp deletion within the coding region of the DNA binding protein (E2A genome region).  A and B) together with HAdV-D8e, Trim and the American sequences (#KT340070 and #KT340056). One additional cut site was observed in subcluster B sequences using SacI digestion. All sequences showed an identical band pattern with both HindIII and SmaI digestions (not depicted). In silico REA analysis was performed using CLC Genomics Workbench v7.

Discussion
Although adenovirus EKC is a notifiable disease according to the German Protection against Infection Act (IfSG), underreporting is presumed as doctors often diagnose EKC solely based on clinical symptoms. However, the RKI reference definition for reporting requires additional laboratory evidence or an epidemiological link to a laboratory confirmed case.
From November 2011 to December 2013, a huge increase in notified adenovirus eye infections was observed in Germany starting with a big nosocomial outbreak in Northrhine-Westphalia affecting 209 patients 13 . This outbreak received a lot of attention in mass media. In (or after) outbreak situations, when public awareness is high, it is likely that more cases get a laboratory confirmation or will be epidemiologically linked to these cases. Preliminary investigations on the epidemic period had a confusing result, as several types of HAdV were detected (types 3, 4, 8, 64 (previously known as genome type 19a), 37, 53). This finding raised suspicion that increased laboratory testing of eye swabs picked up low level circulation of multiple HAdV types because in the previous years (2001-2010) merely one or two different HAdV types were detected per year. Therefore, increased reporting due to increased laboratory testing has to be considered when comparing case numbers of 2012 and 2013 to previous years.
However, HAdV-D8, which was only occasionally isolated in the previous years (in 2004 and 2005 but not in 2006 to 2009), predominated in about 85% of the typing results in the epidemic period from November 2011 to end of 2013. HAdV-D8 seemed to re-emerge in June 2010 but it was found to be a novel strain phylogenetically linked to the later epidemic HAdV-D8 strain. Molecular phylogeny strongly supported the epidemiological hypothesis of a prolonged, multi-state epidemic caused by a monophyletic HAdV-D8 strain between November 2011 and December 2013 in spite of co-circulation of other HAdV types. Moreover, there is a hint that HAdV-D8 strains closely related to the German epidemic HAdV-D8 strain circulated world-wide because two American sequences clustered with the German 2012/13 epidemic samples, although these American sequences were not assigned to either subcluster A or B. All these sequences were most closely related to the sequence of a Japanese isolate of the genome type 8e (#AB448769) 4,14 . This is the first study on molecular epidemiology of EKC analyzing complete genomic sequences.
Classical adenovirus typing approaches such as epitope sequences and restriction enzyme analysis (REA) patterns can be deduced from complete genomic sequence data sets [14][15][16][17] . However, REA would lead to misleading results and conclusions compared to phylogenetic analysis of complete genomic sequences. REA separated the closely related monophyletic subclusters A and B into two different genome types, and on the other hand associated the outlier sequence (#KP016741) with subcluster A. Previously, an analysis of the molecular phylogeny of HAdV-B21 also showed that very closely related isolates of a monophyletic subtype were resolved as two genome types by REA, thus erroneously obscuring a monophyletic origin 18 . Epitope sequences of the hexon (neutralization epitope) and the fiber (hemagglutination inhibition epitope) of this study were highly conserved and thus also not helpful for molecular epidemiology, as already reported by a previous review 14 .
A few German epidemic HAdV-D8 strains had additional deletions in genes coding for proteins (CR1α , CR1γ ) not essential for virus replication. These may either impede the immune response (CR1α gene) 19 or may be associated with an EKC phenotype (CR1γ gene) 20 . However, the significance of these mutations for virulence and disease severity of the epidemic HAdV-D8 strain cannot yet be estimated, mostly for the reason that adequate data for other HAdV-D8 strains were not yet available. Viral loads in eye swabs of the 2012/2013 epidemic HAdV-D8 samples were significantly higher compared to virus loads in eye infections with other HAdV types, indicating higher levels of cytolytic virus replication and infectious virus progeny. This suggested a high virulence of the epidemic HAdV-D8 strain. However, this result should be considered cautiously because the adequate control group, virus loads from non-epidemic HAdV-D8 eye swabs, was not available.
Low population immunity may have promoted the re-emergence of HAdV-D8, leading to a prolonged multi-state epidemic, but data on the immune status against HAdV-D8 were not available for the German population. Low prevalence of HAdV-D8 in the years prior to the epidemic may have resulted in low population immunity. However, this has to be interpreted with caution, as HAdV-D8 infections of children and adolescents, the presumed non-immune age groups, were not reported. Moreover, HAdV-D8 patients were rather old (median 69 years) and patients of these age group may have been exposed to HAdV-D8 many years ago but immunity may have waned. Therefore, the high age of the HAdV-D8 patients during the epidemic period can rather lead to the speculation that nosocomial transmission at routine visits to ophthalmologists' offices (e.g. for glaucoma screening, presbyopia) may have contributed to the epidemic. However, this hypothesis is only supported by data on the first healthcare associated outbreak of the epidemic in Northrhine-Westphalia in late 2011 13 .
In conclusion, a prolonged multi-state HAdV-D8 epidemic was demonstrated by phylogenetic analysis of complete genomic sequences. Emergence of a novel monophyletic HAdV-D8 strain was observed in Germany starting 2010. However, co-circulation of other HAdV types as well as better reporting due to increased public awareness most likely have contributed to the huge increase of reported cases. (b) Cases with the clinical picture defined as redness of the conjunctiva AND with an epidemiological link to a laboratory-proven infection in humans by human-to-human transmission or common source of exposure (e.g. ophthalmological examination devices). The incubation period of 5-12 days, occasionally longer, has to be taken into account.

Description of cases reported to Robert
Thus, in an outbreak at least one case has to be laboratory confirmed.
We also describe subgroups of cases with typing results reported to RKI in the period from November 2011 until December 2013 (epidemic period) as well as cases outside the epidemic period (2001 to October  Laboratory Methods used in the national reference laboratory for HAdV (KLA). HAdV DNA quantification and molecular typing. HAdV DNA detection and molecular typing was performed on all eye swabs sent to the KLA. Viral DNA was extracted from eye swabs with the Qiagen blood kit (Qiagen, Hilden, Germany). A generic, real-time PCR for HAdV was performed on the DNA extracted directly from eye swabs as described previously 21,22 . Molecular typing by sequencing of the loop 2 of the major neutralization determinant ε was performed as described previously 23 . Additional virus isolation was performed using A549 (ATCC, CCL-185) cell cultures.
Next Generation Sequencing (NGS). Complete genomic sequencing was performed from both cell culture isolates and DNA extracted directly from eye swabs as described previously 1,18 . Depending on the concentration of the input sample, library preparation was performed using either the Nextera or NexteraXT DNA Sample Prep Kits (Illumina, San Diego, CA) according to manufacturer's protocol. Briefly, a total input of 50ng (Nextera kit) or 1ng (NexteraXT) from each DNA sample was fragmented and tagged with adaptors in the presence of transposomes, purified and enriched using a 5-cycle PCR (or 12-cycles in case of NexteraXT). Size selection and quality control was performed as described previously 18 . Samples were sequenced with an Illumina MiSeq generating paired-end 300 bp reads. The minimum achieved average coverage was 61 ± 13 fold and the highest average coverage achieved was 14206 ± 3495 fold. The data sets were assembled de novo and also using a genomic HAdV-D8 GenBank sequence (#AB746853.1) as a backbone. All data analysis was performed using the CLC Genomics Workbench version 7 (Aarhus, Denmark).
Phylogenetic Analysis. Whole genomic nucleotide sequences were aligned using the MAFFT online server 24 . The multiple alignment was visualized and edited using the BioEdit package (version 7.2). Phylogenetic analysis was performed using the MEGA software package (version 6). The phylogenetic tree from whole genomic nucleotide sequences was constructed using the Neighbor-joining algorithm. Kimura 2-parameter was used as a substitution model with 1000 bootstrap replicates.
Restriction enzyme analysis (REA) was performed in silico for the enzymes BamHI, HindIII, PstI, SalI, SmaI, and SacI cutting sites using the CLC Genomics Workbench version 7 (Aarhus, Denmark).
Ethical statement. All methods were carried out in accordance with relevant guidelines and regulations. For scientific use of routine anonymized data, ethical approval is not required in Germany (confirmed by the Ethikkommission of the Medizinische Hochschule Hannover (2586-2015)). Informed consent of patients is not required for this type of study.