Haemophilus influenzae is a gram-negative coccobacillus that can cause respiratory infections as well as invasive infections such as septicemia and meningitis (Nørskov-Lauritsen et al., 2021). It is categorized into six serotypes (i.e., Hia, Hib, Hic, Hid, Hie, and Hif) according to the polysaccharide capsule, as well as non-encapsulated strains that are referred to as non-typeable H. influenzae (non-cap) (Potts et al., 2019). In the pre-vaccine era, H. influenzae serotype b (Hib) was a frequent cause of bacterial meningitis among young children (Bijlmer, 1991; Nørskov-Lauritsen, 2014; Jalalvand and Riesbeck, 2018), but after the introduction of Hib vaccines in the 1980s, the numbers diminished dramatically worldwide (Morris et al., 2008; Nørskov-Lauritsen, 2014). The Hib vaccine was introduced into the Danish immunization program in 1993 as a part of the diphtheria-tetanus-pertussis-polio combination vaccine recommended at 3, 5, and 12 months of age, with high vaccination coverage of 96% (12-month vaccination; ¹ accessed 17 December 2021), and it has been estimated to have an effectiveness of more than 97% for the prevention of Hib-related meningitis (² accessed 29 November 2021).

Haemophilus influenzae has traditionally been serotyped by phenotypic methods and PCR (Potts et al., 2019; Watts and Holt, 2019). Recently, two in silico approaches using whole-genome sequencing (WGS), “hinfluenzae_capsule_characterization” (Potts et al., 2019) and “hicap” (Watts and Holt, 2019), have been described that allow easier serotyping.

Haemophilus influenzae can furthermore be characterized based on eight biotypes defined by the presence and absence of the three enzymes, namely, tryptophanase (indole production), urease, and ornithine decarboxylase (Nørskov-Lauritsen, 2014).

In this study, we evaluated the use of two different in silico serotyping programs based on WGS for national surveillance of invasive clinical H. influenzae isolates from Denmark received between 2014 and 2021. Additionally, we investigated the clonal relatedness to serotype and biotype.

The characterization of H. influenzae isolates has traditionally been based on phenotypic methods (Kilian, 1976; Kilian et al., 1979; Nørskov-Lauritsen, 2014). With the introduction of the Hib vaccine four decades ago, the importance of H. influenzae capsular serotype distribution has increased due to continued monitoring of vaccine efficacy (Satola et al., 2007).

Serotyping of H. influenzae isolates has for many years mainly been based on slide agglutination serotyping that detects the expressed capsule (Satola et al., 2007; Potts et al., 2019); however, the slide agglutination serotyping is not always considered reliable, and false-positive Hib designations have been reported (Maaroufi et al., 2007; Satola et al., 2007; Nørskov-Lauritsen, 2014; Potts et al., 2019). With the description of the capsular gene sequences and the development of molecular typing methods, new PCR and WGS-based procedures for H. influenzae serotyping have been presented (Maaroufi et al., 2007; Satola et al., 2007; Nørskov-Lauritsen, 2014; Potts et al., 2019; Watts and Holt, 2019). In the study by Potts et al. (2019), 675 invasive isolates were collected from surveillance programs from 27 states in the United States, and 13 isolates of other origins were compared between the in silico program “hinfluenzae_capsule_characterization” described serotypes and the serotype results obtained by both PCR and slide agglutination tests. They found that the WGS serotyping method was 99.9% concordant with the slide agglutination tests and completely concordant with RT-PCR. Watts and Holt (2019) tested the in silico program “hicap” using 41 publicly available isolate WGS sequences. In 40 of the 41 isolates (98%), they obtained the correct serotype. The difference between the two in silico approaches is that the in silico program “hicap” uses Prodigal (Hyatt et al., 2010) to identify open reading frames (ORFs) in genomic assemblies and BLAST (Camacho et al., 2009) to identify capsule genes based on a custom reference database. The serotype is then predicted based on the detection of capsule genes. The approach by Potts et al. (2019) performs BLAST directly on the genome assemblies and then parses the identified cap genes for truncations and internal stop codons before predicting a serotype, rather than using external software for gene prediction.

This study, representing 73% of all clinical isolates from 2014 to 2021 from Denmark, found 100% concordance between the capsular genotype and the phenotype (slide agglutination test). This is comparable to the observations by Potts et al. (2019) and Watts and Holt (2019). This is the first study to show that in silico serotyping programs are feasible in a national H. influenzae surveillance program with the notion that the Hib vaccine efficacy depends on the phenotypic expression of the Hib polysaccharide and not the presence of the capsular sequences.

As a complementary approach to capsular serotyping, a biotyping setup has been described by Kilian in 1976 (Kilian, 1976; Kilian et al., 1979; Nørskov-Lauritsen, 2014). The system is based on eight biotypes defined by variations in the production of tryptophanase (indole production), urease, and ornithine decarboxylase (ODC) (Nørskov-Lauritsen, 2014). However, because the biotype is not related to the capsular serotype and therefore not considered an important part of the efficacy monitoring of the Hib vaccine in Denmark, phenotypic biotyping has not been performed since 2019. However, comparing the phenotypic biotype with the molecular biotype obtained from isolates sampled from 2014 to 2019 did show a high concordance between phenotypic and molecular biotyping. In total, 27 isolates (5.1%) showed diverging biotypes. Adjusting the percentage of coverage and identity could improve the concordance between phenotypic and molecular biotyping (Table 1). However, there was a mixture of capsular H. influenzae isolates and non-cap H. influenzae isolates among each of the specifically detected biotypes (Figure 2).

While the isolates expressing a capsule could be linked to a specific CC, this was not the case for the non-cap H. influenzae isolates, which exhibited a broad diversity of ST types (Tables 4, 5, and Figure 2). Although the ST types in this study could be used to differentiate the capsular H. influenzae isolates from the non-cap isolates, it has previously been shown that MLST is not optimal for the identification of capsular H. influenzae isolates (Potts et al., 2019). However, we found that the MLST type can indicate the correct identification of the genotype because the capsular isolates appear to be part of only six clonal complexes (Table 4). Therefore, this can, in a national routine H. influenzae surveillance program for monitoring the effect of the Hib vaccine, be used as an additional confirmation of correct capsule identification, in that non-cap ST types belonging within these six CC or capsular isolates not belonging to these six CC, will need further evaluation before final capsule reporting.

Based on the results of this study, we recommend a workflow using genotyping either based on PCR (Maaroufi et al., 2007; Satola et al., 2007) or WGS (Potts et al., 2019; Watts and Holt, 2019). Because of the use of the Hib vaccine, we additionally suggest confirming the expression of the Hib capsule using a latex agglutination test for vaccine surveillance. As an additional verification of the serotype, we propose MLST typing as clinical capsular isolates appear to cluster in specific clonal complexes (Tables 4, 5), although the MLST itself cannot be used for serotype identification (Potts et al., 2019).

The weakness of this study is that by only looking at clinical isolates, the tested isolates representing all six serotypes are not equally distributed but dominated by serotypes f, b, and e (Table 2), while only a limited number of isolates with serotype a were detected, and none with serotype c and serotype d. The strength of the study is that the evaluation is based on clinical isolates received from Danish regional laboratories of clinical microbiology, and all the tested isolates showed a strong correlation between the results obtained with genotyping vs. phenotyping, in line with observations in other studies (Potts et al., 2019; Watts and Holt, 2019).

In conclusion, our study showed that there is a complete concordance between phenotypic serotyping and molecular-based methods for Danish clinical H. influenzae isolates. The concordance is 100% for both the described in silico serotyping methods, replacing laborious phenotypic methods with WGS-based in silico approaches in a clinical setting. In addition, biotyping shows a high concordance between molecular and phenotypic methods. Furthermore, in our Danish routine laboratory where epidemiological surveillance of H. influenzae for monitoring the effect of the Hib vaccine is an important routine task, the MLST type can be used as an indication of whether an isolated expresses a capsule or not. In that, all capsular-defined isolates in this study belonged to one of six clonal complexes depending on their capsule.

The datasets presented in this study can be found in the online repository at the European Nucleotide Archive (ENA) under project no. PRJEB56415.

H-CS and KF designed the study, analyzed the data, and drafted the manuscript. All authors performed the genomic analyses and reviewed the data, contributed to the manuscript, and critically revised the manuscript, and have approved the final manuscript.