Both Coinfection and Superinfection Drive Complex Anaplasma marginale Strain Structure in a Natural Transmission Setting

ABSTRACT

Vector-borne pathogens commonly establish multistrain infections, also called complex infections. How complex infections are established, either before or after the development of an adaptive immune response, termed coinfection or superinfection, respectively, has broad implications for the maintenance of genetic diversity, pathogen phenotype, epidemiology, and disease control strategies. Anaplasma marginale, a genetically diverse, obligate, intracellular, tick-borne bacterial pathogen of cattle, commonly establishes complex infections, particularly in regions with high transmission rates. Both coinfection and superinfection can be established experimentally; however, it is unknown how complex infections develop in a natural transmission setting. To address this question, we introduced naive animals into a herd in southern Ghana with a high infection prevalence and high transmission pressure and tracked the strain acquisition of A. marginale through time using multilocus sequence typing. As expected, the genetic diversity among strains was high, and 97% of animals in the herd harbored multiple strains. All the introduced naive animals became infected, and three to four strains were typically detected in an individual animal prior to seroconversion, while one to two new strains were detected in an individual animal following seroconversion. On average, the number of strains acquired via superinfection was 16% lower than the number acquired via coinfection. Thus, while complex infections develop via both coinfection and superinfection, coinfection predominates in this setting. These findings have broad implications for the development of control strategies in high-transmission settings.

INTRODUCTION

Complex infections, the presence of at least two genetic variants of a pathogen within a host at a given time, are increasingly recognized as common, particularly in association with a high pathogen prevalence. For example, >85% of humans infected with Plasmodium falciparum have complex infections (1). Theoretical and experimental studies have demonstrated that complex infections can drive altered disease dynamics, microorganism community structure, and the evolution of virulence (2–4). As the ability to accurately differentiate genetic variants of one pathogen within individual hosts increases, the biological and epidemiological relevance of complex infections under natural conditions is starting to be understood (5, 6). However, many fundamental knowledge gaps remain, including how genetic variants accumulate in hosts through time under natural transmission conditions.

The accumulation of multiple strains in a host can be through coinfection, superinfection, or, in the case of some RNA viruses, rapid mutation (7). The focus of this paper is on strain accumulation through coinfection or superinfection. Coinfection occurs when different strains of a pathogen establish an infection, either simultaneously or sequentially, before the development of an adaptive immune response. Superinfection occurs when a second strain establishes an infection in a host that is infected with a primary strain and has mounted an adaptive immune response to the primary strain.

Coinfection and superinfection lead to different types of within-host competition, and thus, each may influence the population structure and pathogen phenotypes differently (8, 9). Coinfection results in direct competition among strains for limited resources such as nutrients and receptors for host cell entry. In Plasmodium chabaudi, more-virulent strains are consistently present at higher levels in the blood than less-virulent strains when inoculated concurrently. Because higher parasite levels are associated with increased transmission, coinfection, through time, is thought to select for more-virulent parasites (3, 10).

In contrast, superinfection requires the incoming strain to overcome the existing immune response in order to establish infection. Therefore, for successful superinfection, the second strain must be antigenically and, thus, genetically distinct from the initial strain. Hence, superinfection is vital for the maintenance of genetic diversity. Superinfection can alter clinical outcomes in a variety of ways. For example, individuals with HIV superinfection experience a decrease in their CD4⁺ cell count, which leads to faster disease progression (11–14). Alternatively, with P. chabaudi, superinfecting strains have lower growth rates, likely due to the presence of a preexisting immune response to the initial strain and possibly a reduced number of erythrocytes available for a superinfecting strain to invade (9).

Pathogens in the genus Anaplasma, including A. marginale and A. phagocytophilum, are obligate, intracellular, tick-borne bacteria (15). Anaplasma spp. have marked genetic diversity, especially relative to their small genome sizes, and typically establish long-term infection in the mammalian host (16–18). Because A. marginale is endemic in cattle throughout the world, and the entire transmission cycle in the natural hosts can be readily replicated under experimental conditions, complex infections are well described under both field and laboratory conditions. For example, in a region of Mexico with an A. marginale prevalence of 100%, more than 95% of the animals had >6 strains (19, 20). In contrast, in a temperate region of the United States with an infection prevalence of 29%, approximately 4% of the animals had 2 to 3 strains (21). Experimentally, both superinfection and coinfection with A. marginale have been demonstrated (22, 23).

How complex infections develop in animals over time in a natural transmission setting is unknown. This study addresses this knowledge gap by determining if multiple strains of A. marginale are acquired through coinfection and/or superinfection by tracking naive animals introduced into an endemically infected herd in Ghana. Here, we present and discuss our results in the context of genetic diversity under conditions of natural transmission.

RESULTS

Multistrain A. marginale infections in the study population.

We used multilocus sequence typing (MLST) similar to that described previously by Esquerra et al. to estimate the number of A. marginale strains per animal within a persistently infected herd of cattle in southern Ghana (20). This approach used alleles from five outer membrane protein-encoding genes (omp genes) to differentiate between strains in animals in a region of Mexico with a high A. marginale prevalence (20). These omp genes were previously shown to be invariant during long-term persistent infection within a single strain and, in combination, differentiate between strain pairs based on single nucleotide polymorphisms (SNPs), substitutions, insertions, and deletions in each locus (16, 24, 25).

Anaplasma marginale strains from 30 animals that had been in the herd for 3 to 5 years were genotyped. The minimum number of strains per animal was estimated based on the maximum number of unique alleles of any given omp. Of the 30 animals, only 1 was infected with a single strain, and 29 (97%) harbored multiple strains. Of these 29 animals, 2 harbored six strains, 7 animals harbored five strains, 6 animals harbored four strains, 7 animals harbored three strains, and a different 7 animals harbored two strains (see Table S1 in the supplemental material).

Overall, there was a great deal of genetic diversity of A. marginale within these animals. Including all loci, a total of 105 alleles were identified, with more alleles in some loci than in others (Table 1). Specifically, omp5 had 9 alleles and omp9 had 16 alleles identified in the infected animals (Fig. S1 and S2). omp12 and omp13 had 21 and 22 alleles, respectively (Fig. S3 and S4). There were 37 alleles for omp14 (Fig. S5). The greatest numbers of substitutions and indels were in omp5 and omp13. When considering all animals, the mean number of alleles for a given locus varied from 1.4 for omp9 to 3.3 for omp14 (Table 1). The total numbers of animals that had two or more alleles at a locus were 20 for omp5, 11 for omp9, 20 for omp12, 21 for omp13, and 25 for omp14. Only omp13 and omp14 had more than three alleles in 6 and 13 animals, respectively.

TABLE 1.

Allelic diversity by locus in persistently infected animals

Locus	Total no. of alleles	No. of substitutions/indelsa		Mean no. of alleles detected/infection ± SD (max no. of alleles)	No. of infections/total with:
		Min	Max		2 or more alleles at a locus	>3 alleles at a locus
omp5	9	1/0	48/21	1.8 ± 0.6 (3)	20/30	0/30
omp9	16	8/0	18/0	1.4 ± 0.5 (2)	11/30	0/30
omp12	21	1/0	11/0	1.8 ± 0.7 (3)	20/30	0/30
omp13	22	1/0	14/79	2.5 ± 1.3 (5)	21/30	6/30
omp14	37	1/0	34/3	3.3 ± 1.6 (6)	25/30	13/30

^aMinimum and maximum numbers of substitutions, insertions, and deletions at each locus in comparison to the reference, which is the allele for each locus that occurred most frequently in the persistently infected animals.

Acquisition of complex infections during natural transmission of A. marginale to naive animals.

To determine if strains accumulate in animals by coinfection or superinfection during natural transmission, naive calves were introduced into the herd, and the number of new strains per animal was determined through an 18-week period (Fig. 1). The detection of two or more different alleles at the same locus before seroconversion was defined as coinfection. The detection of additional new alleles following seroconversion was defined as superinfection.

FIG 1.

Experimental design for tracking strain acquisition through time. Shown is the timeline for sampling individual naive animals that were introduced into an A. marginale-infected herd. The specific timeline varied for each animal, depending on the week at which seroconversion occurred. PCR was done weekly to detect A. marginale infection. Once an animal became PCR positive, MLST was done weekly to determine the number of A. marginale strains in each animal. Serum was collected every 2 weeks to detect seroconversion via a cELISA. Prior to seroconversion, all detected strains were defined as coinfecting strains. Following seroconversion, MLST was conducted every 2 weeks to determine the number of A. marginale strains per animal. All strains detected following seroconversion that were not previously detected were defined as superinfecting strains. The horizontal arrow represents the time period following seroconversion when samples were collected and PCR and MLST were performed to detect additional new strains. (−) indicates negative PCR. (+) indicates positive results by PCR or an msp5 cELISA.

All 16 naive animals introduced into the herd became infected with A. marginale during the study period. At the initial detection of infection, two to four strains were present in all animals. Prior to seroconversion, 6% of animals had acquired seven strains, 31% had acquired six strains, 31% had acquired five strains, 19% had acquired four strains, and 13% had acquired three strains. All the calves developed anti-A. marginale antibodies approximately 14 days following the detection of the initial infection. All animals, except those that died, acquired between one and three additional strains following seroconversion (Table 2; Table S2).

TABLE 2.

Number of new strains detected through time in individual naive introduced animals

Animal ID	No. of new strains detected at time (wk)a
	0	1	2	3	4	6	8	10	12	14	16	18
7030	—	—	—	4	1	Nc	N	Died
7031	—	—	—	2	3	2c	1	1	N	1	N	1
7032	—	3b	1	Nc	N	N	N	N	Died
7033	—	—	—	—	—	4	2c	1	1	1	N	N
7034	—	3	1	2c	3	Died
7035	—	3	2	1c	2	Died
7036	—	—	—	—	—	—	3	Nc	1	2	N	N
7037	—	—	—	—	—	4	1c	1	N	N	N	N
7038	—	4	1	1c	1	2	N	N	N	N	N	N
7039	—	—	—	—	4	1c	1	N	1	N	N	N
7040	—	—	—	—	3	Nc	N	1	N	N	N	N
7041	—	—	—	3	1	Nc	1	N	1	N	N	2
7042	—	—	—	—	—	4	1c	1	N	1	1	N
7043	—	—	—	3	2	1c	N	N	1	N	N	N
7044	—	—	—	—	2	3c	2	N	N	N	N	N
7045	—	—	—	—	3	1c	1	1	N	N	N	N

^aDashes indicate the absence of A. marginale based on PCR. Gray shading indicates all sampling time points following seroconversion. N indicates time points at which no new strains were detected.

^bNumber of new strains detected at each sampling period.

^cSeroconversion following the detection of initial infection.

Overall, more strains were acquired through coinfection than through superinfection. A mixed-effects Poisson regression model estimated that the fold change in the average number of strains acquired via superinfection was 0.168 times smaller than the average number of strains acquired via coinfection (P < 2 × 10⁻¹⁶) after controlling for the animal effect and the time of infection, with a 95% confidence interval of 0.11 to 0.25. In other words, on average, the number of strains acquired via superinfection is 16% lower than the number acquired via coinfection. Table 3 provides more detailed information about the test for each covariate considered.

TABLE 3.

Mixed-effects log-linear regression model on strain distribution

Parameter	Estimate	SE	Z value	P value	95% confidence interval
Intercept	0.38	0.11	3.51	8.0 × 10−4	0.15	0.59
Superinfection	−1.78	0.21	−8.71	<2.0 × 10−16	−2.19	−1.39

Chi-square proportion tests were used to compare the numbers of new strains detected before and after seroconversion relative to the number of times that the animals were sampled before and after seroconversion, termed observations. Including all animals, there were totals of 40 observations before seroconversion and 76 observations after seroconversion. The proportion of observations in which three (P = 1.0 × 10⁻⁴) or four (P = 2.5 × 10⁻³) new strains were detected at one time point was higher before seroconversion than after seroconversion. In contrast, there was no difference in the proportion of observations in which one (P = 9.6 × 10⁻¹) or two (P = 1.2 × 10⁻¹) new strains were detected before or after seroconversion. The proportion of observations in which no additional strains were detected after seroconversion was higher than that before seroconversion (P = 2.8 × 10⁻³) (Table 4).

TABLE 4.

Number and proportion of new strains detected at each observation before and after seroconversion

No. of new strains detected	No. (proportiona) of observations relative to seroconversion		P value
	Before	After
None	5 (0.13)	49 (0.65)	2.77 × 10−7
1	12 (0.30)	21 (0.28)	0.96
2	7 (0.18)	5 (0.07)	0.12
3	10 (0.30)	1 (0.01)	1.41 × 10−4
4	6 (0.15)	0 (0.00)	2.48 × 10−3

^aValues in parentheses are the proportions of observations in which a new strain or strains were detected.

Amino acid diversity of alleles able to establish superinfection.

We then asked if the alleles detected as new, incoming alleles following seroconversion, indicating the ability to establish superinfection, had more amino acid diversity than those detected only prior to seroconversion. In Omp5, three alleles (Omp5-1, Omp5-9, and Omp5-13) were detected as new alleles only prior to seroconversion. Thus, they were associated with the establishment of coinfection but not superinfection. The median percent identities of all pairwise comparisons for these three alleles were 97.2% for both Omp5-9 and Omp5-13 and 98.3% for Omp5-1 (Table 5; Tables S3 and S4 and Fig. S6). For the alleles able to establish superinfection (Omp5-2, Omp5-3, Omp5-4, Omp5-5, and Omp5-6), the median pairwise percent identity varied from 78.7% to 98.9%. While there was increased amino acid variation in two alleles able to establish superinfection (Omp5-3 and Omp5-5), this was not consistent among all alleles able to establish superinfection. For all loci, the findings were similar such that the amount of amino acid variation did not segregate with the ability to establish superinfection (Table 5; Tables S3 to S8 and Fig. S6 to S10).

TABLE 5.

Median percent amino acid identities of alleles able to establish coinfection and superinfection

Locus	Before seroconversion (coinfection)			After seroconversion (superinfection)
	Most divergent allele(s)a	Least divergent allele(s)b	Range of median % identitiesc	Most divergent allele(s)d	Least divergent allele(s)e	Range of median % identitiesf
Omp5	5-13, 5-9	5-1	97.2–98.3	5-3, 5-5	5-6	78.7–98.9
Omp9	9-40	9-1, 9-6	71.3–97.1	9-3	9-30	72.4–98.7
Omp12	12-4	12-2	96.0–99.0	12-39	12-29	98.0–99.0
Omp13	13-2	13-45	71.2–95.4	13-43, 13-24	13-1, 13-14	88.9–95.4
Omp14	14-45	14-1, 14-9	77.4–92.5	14-3	14-10, 14-14	78.5–92.5

^aThe most divergent allele from each locus that is able to establish coinfection but not superinfection.

^bThe least divergent allele from each locus that is able to establish coinfection but not superinfection.

^cThe median percent identity of pairwise comparisons of the most and least divergent alleles able to establish coinfection but not superinfection.

^dThe most divergent allele from each locus that is able to establish superinfection.

^eThe least divergent allele from each locus that is able to establish superinfection.

^fThe median percent identity of pairwise comparisons of the most and least divergent alleles able to establish superinfection.

DISCUSSION

Ninety-seven percent of animals from this herd in southern Ghana had complex infections, with two to six strains per animal. This is similar to previous findings from a high-prevalence, tropical region of Mexico where individual animals harbored up to six strains of A. marginale. In the southern Ghana herd, coinfection contributed more than superinfection to the development of complex infections in individual naive animals introduced into the herd. Similarly, patients residing in areas with high rates of P. falciparum transmission harbored up to 16 closely related genotypes, indicating that coinfection by parasites from a single mosquito was more common than superinfection acquired by repeated mosquito bites (5).

The numbers of strains acquired through coinfection and superinfection are likely minimal estimates. In the case of coinfection, the anti-Msp5 antibody response was used as a measure of the development of strain-specific immunity. Anti-Msp5 antibodies develop relatively early in infection and are easy to detect. Although a robust marker of seroconversion, anti-Msp5 antibodies may develop earlier than a strain-specific immune response. Consequently, we may have underestimated the number of new strains acquired via coinfection. All animals that were coinfected acquired additional strains following seroconversion. However, following seroconversion, A. marginale levels are typically maintained at relatively low levels through time compared to those prior to seroconversion (26). Additionally, the existing immune response will prevent new strains from replicating to high levels. Consequently, it is likely that some new alleles were not detected following seroconversion, leading to an underestimation of the number of new strains acquired by superinfection.

It is possible, although unlikely, that alleles arose via mutation following infection in individual animals. The spontaneous mutation rate, which varies among bacterial species, has not been determined for A. marginale (27). In Escherichia coli, the spontaneous mutation rate is approximately 2 × 10⁻¹⁰ mutations/nucleotide/generation (28). If we assume a similar spontaneous mutation rate for A. marginale and adjust for a genome size of 1.26 Mb, the mutation rate would be 2.4 × 10⁻⁴ mutations/genome/generation. This is likely an overestimate because A. marginale, as an obligate intracellular bacterium, overall has poor tolerance for mutations, with the exception of the hypervariable surface protein Msp2.

There was no association between amino acid diversity and the ability of an allele to establish superinfection. This suggests that there is little selective pressure exerted on these proteins by the immune system and is consistent with the current hypothesis that immune evasion is predominantly due to variation in the hypervariable proteins Msp2 and Msp3.

Some alleles were identified only before seroconversion, while others were identified only after seroconversion. The Omps are surface-exposed members of the Msp2 superfamily (24). It is possible that this segregation of alleles was random. Alternatively, as surface proteins, they may serve as porins, nutrient transporters, or adhesins and invasins. Thus, particular variants may provide some advantage in either the tick or bovine host under conditions of intense, within-host competition. However, until we understand gene function, the mechanisms that potentially confer a competitive advantage to particular variants will remain unknown.

The genetic diversity of a pathogen allows the host to be infected with multiple genotypes, thus effectively maintaining a susceptible host population even under conditions of high infection and transmission rates (22). High A. marginale prevalences combined with heavy tick burdens are likely required for the high levels of coinfection observed in this study. These findings are relevant under similar conditions that are common throughout the tropics and subtropics (29–31). However, in temperate regions, even in herds with a high A. marginale prevalence, tick feeding is seasonal, and burdens tend to be lower than those in tropical and subtropical regions (32). Thus, there are potentially fewer opportunities for coinfection. This may account for the observation that complex infections are less common and have fewer strains per animal in herds in temperate regions, although it remains to be determined how these complex infections develop in temperate regions (21).

The impacts of heavy rates of coinfection on morbidity and mortality are unknown. Unpredictable outbreaks with moderate to high levels of morbidity and mortality are a prominent but poorly understood aspect of bovine anaplasmosis. It is possible that intense coinfection by multiple strains in a short time frame could play a role in the development of these outbreaks. However, epidemiological data are required to correlate high levels of coinfection with disease outbreaks or increased morbidity or mortality.

Typically, persistently infected animals, even under conditions of high transmission rates, do not develop disease upon superinfection because the bovine immune system can prevent high levels of replication of the incoming strain and, thus, disease. Additionally, there may be fitness costs to the genetic variants, thus limiting the ability of the variants to replicate to high levels (33). These findings suggest that vaccination could limit coinfection and, thus, the number of strains per animal in a herd.

Ultimately, within-host competition in the tick, including the ability of a strain or multiple strains to be transmitted simultaneously from an individual tick, will dictate the success of particular variants and may help guide intervention strategies focused on tick control. In experimental infections, ticks can acquire at least two strains of A. marginale, although the number of strains transmitted by a single tick during a blood meal is unknown. In the context of within-host competition between two A. marginale strains in a tick, there are differences in the abilities of individual strains to invade and replicate in tick tissues (23, 34).

Under conditions of high genetic diversity, the system is more complex. With Francisella novicida, also transmitted by ticks, under laboratory conditions using up to 94 closely related strains, diversity was limited in the tick compared to the mammalian host due to selective forces and stochastic factors (35). How these factors influence transmission and trait selection in a natural transmission setting remains unknown. Thus, understanding the role of the tick in the establishment of complex infections is the next step in building a more complete understanding of the disease ecology of tick-borne bacterial pathogens.

MATERIALS AND METHODS

Ethical statement.

Cattle used in this study were treated in strict accordance with guidelines set by the University of Ghana Institutional Animal Care and Use Committee (Noguchi Memorial Institute for Medical Research NIACUC protocol number 2015-01-5X) and the Institutional Animal Care and Use Committee at Washington State University (IACUC-04326).

Experimental design.

This study was conducted in the Coastal Savannah of the Ghana Livestock and Poultry Research Centre (LIPREC) using Sanga Friesian cross-bred cattle maintained on pasture for 3 to 5 years. The A. marginale prevalence is approximately 75% in this herd (36). There are multiple tick species within this herd. Efficient vectors of A. marginale include Rhipicephalus decoloratus, R. annulatus, and R. evertsi evertsi. As part of general herd management, animals are treated with acaricides as needed.

The LIPREC research herd has 40 cattle. From this herd, 30 adult A. marginale-infected animals, as determined by PCR, were enrolled in this study. The number of A. marginale strains within these 30 animals was determined using MLST.

In order to track the accumulation of strains during natural transmission, 16 naive calves, confirmed by negative msp5 PCR and a competitive enzyme-linked immunosorbent assay (cELISA), maintained under tick-free conditions were recruited from LIPREC, Accra, Ghana. These calves were >6 months of age in order to avoid false-positive results of the cELISA due to residual colostral antibody. The animals were introduced into the A. marginale-infected herd, and strain acquisition was measured through time.

The naive calves were bled weekly and tested for A. marginale using msp5 PCR. Positive samples were further characterized by MLST. Following the detection of A. marginale, serum samples were collected weekly to detect seroconversion. Following seroconversion, biweekly blood samples were collected for MLST (Fig. 1).

PCR to detect A. marginale infection.

Genomic DNA was extracted using the Qiagen DNeasy blood and tissue isolation kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s instructions. A. marginale was detected using msp5 PCR, as previously described (37). Dream Taq green polymerase (Thermo Fisher Scientific, Waltham, MA, USA), published msp5 primers, and 200 ng of DNA were used for each reaction (37). msp5 was amplified using the following cycling parameters: 94°C for 3 min; 35 cycles of 95°C for 30 s, 60°C for 30 s, and a final extension step at 72°C for 1 min; and a 5-min final extension step at 72°C.

cELISA to detect seroconversion.

Blood was collected from each calf at 14 to 21 days postinfection, and serum was harvested. Sera were then tested for antibodies against A. marginale using an Msp5 cELISA (VMRD Inc., Pullman, WA, USA) according to the manufacturer’s protocol.

MLST for target loci and primer design.

MLST was performed on msp5 PCR-positive samples. Five genes, omp5, omp9, omp12, omp13, and omp14, were previously validated for MLST (20). Library construction consisted of two rounds of PCR. The primer sets and corresponding amplicon sizes are listed in Table 6. Round 1 PCR (PCR1) amplified the target genes with primers that included universal short consensus (CS) tags (CS1 and CS2 tags), which were added to the 5′ end of each locus-specific primer. These universal CS tags and barcode primers used in our study were prepared by the University of Idaho Genomics Resources Core (IBEST GRC). Round 2 PCR (PCR2) extended the universal sequences with adapters and barcodes.

TABLE 6.

Genes and primers used for MLST

Gene	Primer name	CS tag sequence	Tag name	Spacera	Target-specific sequence (5′–3′)	Size (bp)
omp5	CS1-omp5F	ACACTGACGACATGGTTCTACA	CS1		CTGGAAAGCTGCACAGGATG	310
	CS1-omp5F.T	ACACTGACGACATGGTTCTACA	CS1	T	CTGGAAAGCTGCACAGGATG
	CS2-omp5R	TACGGTAGCAGAGACTTGGTCT	CS2		CGACGCTTCCGCAAACATAC
	CS2-omp5R.T	TACGGTAGCAGAGACTTGGTCT	CS2	T	CGACGCTTCCGCAAACATAC
omp9	CS1-omp9F	ACACTGACGACATGGTTCTACA	CS1		GGCAATTCCAATCATGTGCG	318
	CS2-omp9R	TACGGTAGCAGAGACTTGGTCT	CS2		CAAGCTGTGAAGTCACTACACG
	CS2-omp9R.T	TACGGTAGCAGAGACTTGGTCT	CS2		CAAGCTGTGAAGTCACTACACG
omp12	CS1-omp12F	ACACTGACGACATGGTTCTACA	CS1		CTAGCGCTATGTTGCATGCATC	347
	CS1-omp12F.T	ACACTGACGACATGGTTCTACA	CS1	T	CTAGCGCTATGTTGCATGCATC
	CS2-omp12R	TACGGTAGCAGAGACTTGGTCT	CS2		ACGCAAATTCAGATCACAGGG
	CS2-omp12R.T	TACGGTAGCAGAGACTTGGTCT	CS2	T	ACGCAAATTCAGATCACAGGG
omp13	CS1-omp13F	ACACTGACGACATGGTTCTACA	CS1		CAAGCAGATCCACAGCATCAATTC	306
	CS2-omp13R	TACGGTAGCAGAGACTTGGTCT	CS2		GTGACGCCCTCATTGACC
	CS2-omp13R.C	TACGGTAGCAGAGACTTGGTCT	CS2	C	GTGACGCCCTCATTGACC
omp14	CS1-omp14F	ACACTGACGACATGGTTCTACA	CS1		GCAGAAGGAGTTGTCCAAGC	327
	CS2-omp14R	TACGGTAGCAGAGACTTGGTCT	CS2		CCACTTATTTCCACAATCTCCATGC

^aThe spacer nucleotide generates synthetic diversity to improve read quality.

PCR1 was performed in a 25-μl reaction mixture containing Kapa HiFi HotStart ready mix 2× master mix polymerase (Roche, San Francisco, CA, USA), forward and reverse target-specific primers (Table 6), the A. marginale DNA template, and water. Targets were amplified using the following cycling parameters: 94°C for 3 min; 35 cycles of 95°C for 30 s, with different temperatures for each gene (omp5, -9, and -14 at 60°C; omp12 at 58°C; and omp13 at 59°C) for 30 s; 72°C for 1 min; and a 5-min final extension step at 72°C. Amplicons from PCR1 were used as a template for PCR2.

PCR2 was performed in a 20-μl reaction mixture containing Kapa HiFi HotStart ready mix 2× master mix polymerase (Roche), unique barcoded primers, 3 to 5 μl of PCR1 products as the template, and water. The target was amplified under the following conditions: 94°C for 30 s; 10 cycles of 95°C for 30 s, 60°C for 30 s, and 72°C for 1 min; and a 5-min final extension step at 72°C.

The amplicons were visualized on a 1% agarose gel, and a 1-kb low-DNA-mass ladder (Invitrogen, Carlsbad, CA) was used to estimate the concentration of each amplicon. Amplicons were pooled into 5 groups based on the gene locus after barcode ligation. Each group had 30 samples with similar concentrations of DNA. Each of the pools was cleaned using a 1× AMPure XP cleanup kit (Beckman Coulter, Indianapolis, IN). The purified products were run on a 1% gel, excised, and column purified using a Qiagen MinElute PCR purification kit according to the manufacturer’s protocol.

To determine the quality of the resulting amplicon pools, samples were PCR purified with Illumina adapter-specific primers and analyzed with a Bioanalyzer 2100 system (Agilent Technologies Inc., Santa Clara, CA, USA) according to the manufacturer’s protocol. Quantitative PCR was performed on the final library pools to determine the amount of sequenceable library and predict cluster numbers more accurately. The library was then sequenced at the University of Idaho IBEST GRC using 50% of the 2-by-300 MiSeq run.

Data analysis.

Sequencing reads were demultiplexed to dually barcoded pairs and PCR1 primers using dbcAmplicons applications in R (38). After quality trimming, paired reads were overlapped and joined into a single continuous sequence using FLASH version 2.2.00, with minimum and maximum overlaps set to 30 and 600. FastQ sequence reads were loaded into Divisive Amplicon Denoising Algorithm 2 (DADA 2) (version 1.4.0, R version 4.0) for all analyses. The parametric error model in DADA 2 was used to set an error rate for the data set. DADA 2 took raw amplicon sequence data in FastQ files as an input and produced an error-correctable table of abundances of amplicon sequence variants (ASVs) in each sample. This was done using an amplicon variant tool (ASV), which produced an output ASV table of reads.

Technical error determination.

After DADA 2 processing, a technical error rate according to a method established previously by Esquerra et al. was calculated (20). To determine the technical error rate, the target loci were cloned and sequenced from genomic DNA extracted from blood from an animal experimentally infected with the St. Maries strain of A. marginale. The cutoff for determining the technical error rate was the mean plus 3 standard deviations of the percentage of sequenced clones not identical to the reference clone. Three to five clones per locus were used to calculate the cutoff. For all clones, a single genotype comprised >92% of all the sequences. The percent cutoff for each of the genes ranged from 2% to 8%.

Nucleotide sequence analysis.

Multiple-sequence alignment was performed using the Clustal W algorithm in Sequencher (Gene Codes Corporation, Ann Arbor, MI, USA). Identification of DNA sequence polymorphisms, insertions, and deletions was done using Sequencher (version 5.4.6) and Geneious Prime (version 2020.2.3) software packages.

Statistical analysis.

We fit a mixed-effects Poisson regression model with random effects of animals and the offset being the log number of weeks. This model estimates the distribution of strains before seroconversion (coinfection) and after seroconversion (superinfection). The response variable was the number of new strains detected either before or after seroconversion, and the explanatory variable was the type of infection (coinfection or superinfection), with the effects of animals and the number of weeks controlled for. All analysis was done in R studio version 4.0.3. P values of ≤0.05 were considered statistically significant. A chi-square proportion test was used to analyze the proportion of new strains detected in each animal before (coinfection) and after (superinfection) seroconversion.