Comparing Whole-Genome Sequencing with Sanger Sequencing for spa Typing of Methicillin-Resistant Staphylococcus aureus

Abstract

spa typing of methicillin-resistant Staphylococcus aureus (MRSA) has traditionally been done by PCR amplification and Sanger sequencing of the spa repeat region. At Hvidovre Hospital, Denmark, whole-genome sequencing (WGS) of all MRSA isolates has been performed routinely since January 2013, and an in-house analysis pipeline determines the spa types. Due to national surveillance, all MRSA isolates are sent to Statens Serum Institut, where the spa type is determined by PCR and Sanger sequencing. The purpose of this study was to evaluate the reliability of the spa types obtained by 150-bp paired-end Illumina WGS. MRSA isolates from new MRSA patients in 2013 (n = 699) in the capital region of Denmark were included. We found a 97% agreement between spa types obtained by the two methods. All isolates achieved a spa type by both methods. Nineteen isolates differed in spa types by the two methods, in most cases due to the lack of 24-bp repeats in the whole-genome-sequenced isolates. These related but incorrect spa types should have no consequence in outbreak investigations, since all epidemiologically linked isolates, regardless of spa type, will be included in the single nucleotide polymorphism (SNP) analysis. This will reveal the close relatedness of the spa types. In conclusion, our data show that WGS is a reliable method to determine the spa type of MRSA.

INTRODUCTION

Methicillin-resistant Staphylococcus aureus (MRSA) remains endemic in many hospitals (1), and the prevalence in the community has been increasing worldwide for the last decades (2). spa typing has been widely used for typing of MRSA isolates to investigate outbreaks and local epidemiology. The introduction of the Ridom StaphType software program in 2003 (3) has made spa typing an excellent tool to compare MRSA types across regions and countries, since it has proven to be a typing scheme with high interlaboratory reproducibility (4). spa typing is the sequencing of the 24-bp (21 to 30 bp) repeat region of the Staphylococcus protein A gene (spa). Repeats are assigned a numerical code, and the spa type is determined from the order of these repeats. By 9 July 2014, 647 repeats and 13,857 different spa types were registered in the SpaServer database (http://spaserver.ridom.de). The traditional method of spa typing is by PCR amplification of the spa repeat region followed by Sanger sequencing (3).

At the Department of Clinical Microbiology at Hvidovre Hospital, Hvidovre, Denmark, whole-genome sequencing (WGS) has been performed since January 2013 on all MRSA isolates from the Capital Region of Denmark. WGS was introduced to improve outbreak investigations of MRSA and to gain a better and more thorough understanding of the spread and evolution of MRSA. MRSA genome sequences are analyzed for spa types, multilocus sequence types (MLST), and the presence or absence of PVL genes (Panton-Valentine leukocidin), and the findings are used as a common nomenclature to share with clinicians in patient records. For infection control, single nucleotide polymorphism (SNP) analysis is used routinely to compare MRSA isolates with identical or related spa types and isolates belonging to the same clonal complex (CC) from patients suspected to be part of an MRSA outbreak.

In Denmark, MRSA from infections and carriage has been notifiable since 2006. MRSA isolates from the departments of clinical microbiology in Denmark are sent for national surveillance at the Staphylococcal Laboratory at Statens Serum Institut, where isolates are spa typed by PCR amplification and Sanger sequencing.

The purpose of the present study was to evaluate the reliability of MRSA spa typing by WGS. This was done by comparing the spa typing results of WGS with Sanger sequencing results. As far as we know, this is the first study comparing spa typing by Sanger sequencing and WGS.

MATERIALS AND METHODS

The Capital Region of Denmark has three departments of clinical microbiology that perform microbiology analysis for 12 hospitals and the general-practice health care services of the region's 1.72 million inhabitants. One MRSA isolate per patient was confirmed, with an in-house multiplex real-time PCR assay detecting the presence of nuc, femA, mecA, and mecC (data not shown). All confirmed MRSA isolates from the Capital Region were whole genome sequenced on a MiSeq instrument (Illumina, San Diego, CA, USA) at the Department of Clinical Microbiology at Hvidovre Hospital. The current workload was a 4-day setup. DNA concentrations were normalized using a Qubit fluorometer (Invitrogen, United Kingdom). Libraries were made using a Nextera XT DNA sample preparation kit (Illumina, USA), and genomes were multiplexed to 24 isolates per run and sequenced using 2 × 150-bp paired-end reads. Genomes were assembled using the software programs Velvet v1.0.11 (5) and VelvetOptimiser v2.1.7, with hash (kmer) size and coverage parameters optimized to give the highest N₅₀ value. Variants were called using the SAMtools v0.1.12 mpileup command with options -M0-Q30 -q30-o40 -e20-h100 -m2 -D -S (6). An in-house analysis pipeline has been developed to analyze MRSA genomes for the mecA, mecC, nuc, and ccr genes, spa type, MLST, dru types, PVL, and ACME (the arginine catabolic mobile element). spa types were called from the assembled contigs by comparison to the published types on the SpaServer (http://spa.ridom.de/index.shtml). In cases where a spa type could not be determined, the WGS was routinely repeated. MLST was called from the assembled contigs by comparison to the published sequence types on http://saureus.mlst.net.

At Statens Serum Institut, all isolates were confirmed to be MRSA using a multiplex PCR detecting the presence of mecA, mecC, PVL, and spa (7). The spa type was determined by direct Sanger sequencing of the amplicons. The spa types were assigned using the software program BioNumerics v6.1 (Applied Maths, Sint-Martens-Latem, Belgium). In cases where the spa amplicons were undetected, alternative primers were used: spa_239f, 5′-ACTAGGTGTAGGTATTGCATCTGT-3′; spa_1717r, 5′-TCCAGCTAATAACGCTGCACCTAA-3′; spa_1084f, 5′-ACAACGTAACGGCTTCATCC-3′; spa_1618r, 5′-TTAGCATCTGCATGGTTTGC-3′.

The MRSA databases from Hvidovre Hospital and Statens Serum Institut were compared, resulting in a common database of 699 MRSA isolates from new patients in 2013.

We also evaluated whether the N₅₀, an indicator of assembly quality, could be used as a quality control for the likelihood of a correct spa type. To find N₅₀ of an assembly that contains N nucleotides, the contigs are sorted by size and the nucleotides in the sorted contigs are counted from one end. N₅₀ is the size of the contig containing nucleotide number N/2.

RESULTS

Of the 699 MRSA isolates, 680 (97%) had identical spa types by Sanger sequencing and WGS. These 680 isolates included 136 different spa types, the most common being t002 (n = 74), t008 (n = 63), t019 (n = 62), and t304 (n = 62). The number of spa repeats ranged from 3 to 17. Nineteen isolates had different spa types by Sanger sequencing and WGS.

The spa types of the 19 isolates with spa types that were different by the two methods are presented in Table 1. In 18 cases the spa type difference was caused either by the lack of 24-bp repeats (13 isolates) or by additional 24-bp repeats (5 isolates) by WGS. In one case, the spa types found by the two methods were unrelated based on spa repeats (t630/t304). In this case, the assembly was of low quality (N₅₀ = 2,248 bp) and only partial mecA and nuc genes were identified. Furthermore, the genome size was 3.4 Mbp, indicating an isolate contamination. The patient was related to a person with t304. We therefore resequenced the isolate and found both t304 and t630 in the same sample and a genome size of 3.35 Mbp.

TABLE 1.

Isolates with different spa types by WGS and Sanger sequencing^a

spa type/ST by WGS	spa type by Sanger sequencing	spa repeats by WGS	spa repeats by Sanger sequencing
t015/ST45	t026	08-16-02-16-34-13-17-34-16-34	08-16-34
t032/ST22	t1249	26-23-23-13-23-31-29-17-31-29-17-25-17-25-16-28	26-23-23-13-23-23-13-23-31-29-17-31-29-17-25-17-25-16-28
t086/ST88	t690	07-12-21-17-13-13-13-34-34-34-33-34	07-12-21-17-13-13-34-34-34-33-34
t186/ST88	t690	07-12-21-17-13-13-34-34-33-34	07-12-21-17-13-13-34-34-34-33-34
t304/ST6	t197	11-10-21-17-34-24-34-22-25	11-10-34-24-34-22-25
t355/ST152	t595	07-56-12-17-16-16-33-31-57-12	07-56-12-17-16-16-33-31-57-31-57-12
t359/ST97	t267	07-23-12-21-17-34-34-33-34	07-23-12-21-17-34-34-34-33-34
t630/NA	t304	08-16-02-16-34-17-34-16-34	11-10-21-17-34-24-34-22-25
t670/ST22	t5177	26-23-23-13-23-29-17-25-17-25-16-28	26-23-23-13-23-29-17-31-29-17-31-29-17-25-17-25-16-28
t718/ST22	t1249	26-23-23-23-13-23-31-29-17-31-29-17-25-17-25-16-28	26-23-23-13-23-23-13-23-31-29-17-31-29-17-25-17-25-16-28
t728/ST45b	t015	08-16-34-16-34	08-16-02-16-34-13-17-34-16-34
t790/ST22	t022	26-23-13-23-31-29-17-25-17-25-16-28	26-23-13-23-31-29-17-31-29-17-25-17-25-16-28
t934/ST80	t1198	07-23-12-34-34-34-34-33-34	07-23-12-34-34-34-34-34-33-34
t1028/ST78	t237	07-34-33-34	07-34-34-33-34
t4699/ST93	t3949	11-17-16-16-25	11-17-23-17-17-17-16-16-25
t5090/ST130	t843	04-82-17-16-17	04-82-17-25-17-25-25-16-17
t5608/ST5	t002	26-23-17-34-17-13-17-20-17-12-17-16	26-23-17-34-17-20-17-12-17-16
t3119/ST8	t1774	11-19-19-19-12-05-17-34-24-34-22-25	11-19-19-12-05-17-34-24-34-22-25

^aST, sequence type; NA, not available.

^bTwo isolates of t728. Both were t015 by Sanger sequencing.

In Table 2, the isolates are divided into two groups, either with an N₅₀ below or above 20,000. The number of isolates that obtained a spa type and the number of isolates with a correct spa type, using Sanger spa types as gold standards, are presented in the table, as is the percentage of a correct spa type in each group.

TABLE 2.

N₅₀ values and spa typing results^a

WGS run	N50 category	No. of isolates	No. of isolates with a spa type	No. of isolates with a correct spa type	% correct spa types (95% confidence interval)
First	Below 20,000	70	47	44	94 (82.4–98.7)
	Above 20,000	629	612	600	98 (96.9–99.1)
Second	Below 20,000	7	6	5	83
	Above 20,000	33	32	29	91
Third	Above 20,000	2	2	2	100

^aResequencing was performed only for isolates that lacked a spa type.

Forty isolates (5.7%) did not obtain a spa type by the first WGS due to either a poor sequencing result (n = 27) or the spa repeats being assembled on more than one contig (n = 13). These 40 isolates were resequenced, resulting in a spa type in all isolates (Table 2). A number of isolates lacked the spa amplicons using the multiplex PCR prior to Sanger sequencing, and this was resolved by the use of alternative spa primers. At Statens Serum Institut, repetition of Sanger sequencing due to an initial poor sequencing result is done for less than 5% of isolates.

DISCUSSION

With the rapidly increasing number of clinical bacterial isolates being whole genome sequenced, backward compatibility to sequence-based typing methods, such as spa typing and MLST, are very important. In this study, we focused on the commonly used S. aureus sequence-based typing method, spa typing, represented by more than 298.000 isolates in the SpaServer database. The spa gene is composed of a variable number of highly similar repeats and therefore should be more challenging for WGS, since repeated sequences can be misassembled (8). We used Illumina 150-bp paired-end sequencing, and thus spa types longer than six repeats are always on more than one read. When these reads are assembled, it can lead to a misinterpretation if identical repeats are in a series. The use of 250-bp paired-end sequencing might improve the spa typing reliability. However, with an agreement in this study of 97%, it would probably not give an add-on value compensating for increased expenses and longer run time.

In the 19 cases where the spa types differed by the two methods, 18 were caused by spa types differing by one or more repeats. In most cases, series of repeats were missed by the WGS assembly (Table 1). However, most of these spa types were generally so closely related that the BURP (based upon repeat pattern) algorithm (9) would have grouped them into the same spa complex. In one case, the spa types obtained by the two methods seemed unrelated (t630/t304). However, a further look at the WGS data revealed a low N₅₀, a larger genome size than expected, and only partial sequences of the mecA and nuc genes. Repeating the WGS resulted in the finding of both t304 and t630 in the same sample and again a genome size of 3.35 Mbp. This example illustrates the importance of having a thorough look at the WGS data before accepting a spa type. The reporting of incorrect spa types should not have any consequence in the case of a suspected MRSA outbreak, if MRSA isolates from all outbreak-related patients, irrespective of spa types, are included in the SNP analysis. The SNP analysis will reveal the close relatedness of the outbreak isolates and would therefore still confirm a connection between patients.

In our study, for 13 isolates, the spa repeats were located on 2 or 3 contigs after the first WGS, resulting in a nontypeable spa type. These isolates were all resequenced, and all isolates obtained a spa type by WGS after a second run (Table 2). The 13 isolates had spa types with between 4 and 16 repeats, and the assembly difficulties could therefore not be explained by long spa repeat regions. Furthermore, all isolates had different spa types, so the assembly difficulties could not be correlated to specific spa types.

No defined quality control for microbial WGS results exists. We look at the N₅₀, the genome size, and the complete detection of the mecA and nuc genes when evaluating the WGS data. We evaluated whether a low N₅₀ could be used as an indicator of when to repeat the WGS. As expected, a smaller number of isolates obtain a spa type when the N₅₀ is below 20,000 (Table 2). Although there is a higher percentage of correct spa types when the N₅₀ is above 20,000 than with an N₅₀ below 20,000 (94% versus 98%), the 95% confidence intervals overlap. Therefore, the N₅₀ is not a parameter for when to trust a spa type in our study. Our routine is therefore only to repeat the WGS when no spa type is obtained or when the genome size and/or the mecA and nuc genes are not of the expected length.

We believe that spa types are still important to obtain, since this information can easily be exchanged with and communicated to clinicians and distributed in the public domain. Furthermore, the spa typing scheme is well known and will be a natural part of an international WGS database. The advantage of WGS is that other genes of interest can easily be analyzed, including additional genes used for typing, for example, MLST, virulence genes, and resistance genes (10). In addition, the WGS data are stored and whenever needed can be reanalyzed for genes of interest or compared to other whole-genome-sequenced MRSA isolates. This is especially interesting for community isolates, where collection of epidemiological data can be difficult to obtain. The cost of WGS of one isolate is at the moment approximately €100, but obtaining all the data included in our analysis pipeline by Sanger sequencing would be more expensive and time-consuming.

In conclusion, our data show that spa typing by WGS can reliably replace Sanger sequencing. Virtually total agreement was found between the two methods. The few divergences had no true significance because in outbreak investigations, these isolates would cluster together with related spa types by SNP analysis. We believe that the advantage of the additional genomic information gained by WGS is higher than the disadvantage of a small number of inexact spa types.