Selective elimination of aphid endosymbionts: effects of antibiotic dose and host genotype, and fitness consequences

Abstract

Multiple endosymbionts commonly coexist in the same host insects. In order to gain an understanding of the biological roles of the individual symbionts in such complex systems, experimental techniques for enabling the selective removal of a specific symbiont from the host are of great importance. By using the pea aphid−Buchnera−Serratia endosymbiotic system as a model, the efficacy, generality, and fitness consequences of selective elimination techniques at various antibiotic doses and under a variety of host genotypes were investigated. In all the disymbiotic aphid strains examined, the facultative symbiont Serratia was selectively eliminated by ampicillin treatment in a dose-dependent manner, suggesting a generality of the elimination technique irrespective of host genotype. However, fitness consequences of the Serratia elimination differed between the aphid strains, indicating substantial effects of host genotype. In all the disymbiotic aphid strains, the obligate symbiont Buchnera was selectively eliminated by rifampicin treatment irrespective of the antibiotic dose. However, the survival and reproduction of the Buchnera-free aphids varied in a dose-dependent manner, and the dose dependence was strikingly different between the aphid genotypes. These results provide a basis for the development of new protocols for manipulating insect endosymbiotic microbiota.

pea aphid Acyrthosiphon pisum, essential primary endosymbiont Buchnera, facultative secondary endosymbiont Serratia symbiotica, selective elimination of endosymbionts, antibiotics, fitness consequences

Introduction

Bacterial endosymbionts are found in a diverse array of insects and other organisms. Some symbionts are obligate associates for their hosts and significantly contribute to the host fitness, other symbionts are facultative companions for their hosts and cause context-dependent effects, and the majority of the others are of unknown nature and are recognized by means of microscopy, PCR detection and/or 16S rRNA gene sequencing (Buchner, 1965; Baumann et al., 2000; Bourtzis & Miller, 2003).

Whether their effects are beneficial, detrimental or nearly neutral, many of these symbionts substantially affect the physiology, ecology, reproduction and behaviour of their hosts in a variety of ways (O'neill et al., 1997; Baumann et al., 2000; Bourtzis & Miller, 2003). Hence, in order to understand the biology of the many insects that are closely associated with microorganisms, it is of fundamental importance to evaluate experimentally the impact of the symbiont infection on the host phenotypes.

The phenotypic effects of a particular symbiont should be evaluated by a comparison between infected and uninfected host insects whose genetic backgrounds are ideally completely, but at least nearly, identical. Such insect strains can be generated by three techniques: (i) introgression (Bordenstein & Werren, 1998; Kondo et al., 2005), (ii) transfection (Boyle et al., 1993; Chen & Purcell, 1997; Sasaki & Ishikawa, 2000; Fukatsu et al., 2001), and (iii) curing (Wilkinson, 1998; Heddi et al., 1999). Each of these techniques has its own advantages and disadvantages. Introgression by repetitive backcrosses of infected females with uninfected males is, by virtue of the fact that it is mediated by a natural process of intraspecific mating, easily executable with sexually reproducing species, but it takes a considerable length of time (five or more generations) to homogenize the host genetic background sufficiently. Transfection of the symbiont from infected hosts into uninfected ones by microinjection can promptly establish infected and uninfected insect strains of the same genetic background. However, it requires special instruments, and such artificially generated host−symbiont combinations often lead to unstable infection and/or considerable deleterious effects on the host fitness (McGraw et al., 2002; Koga et al., 2003; Russell & Moran, 2005). Curing of the symbiont infection from infected hosts by antibiotic or heat treatment can also promptly establish infected and uninfected insect strains of the same genetic background. However, because of the possible side effects of the treatment, caution must be taken when interpreting the effects of the symbiont elimination on the host phenotypes. Note that these techniques are generally applicable to unculturable fastidious insect symbionts.

In cases in which the insect of interest is associated with a single symbiont species, these techniques are generally reasonably effective. However, multiple endosymbiotic bacteria frequently coexist in the same host organism. For example, many insects are coinfected with two or more different strains of Wolbachia endosymbionts (Kittayapong et al., 2000; Kikuchi & Fukatsu, 2003; Kyei-Poku et al., 2005). Many aphids harbour a variety of facultative endosymbiotic bacteria, such as Serratia, Hamiltonella, Regiella, Rickettsia, Spiroplasma, Arsenophonus and others, in addition to the obligate symbiont Buchnera (Tsuchida et al., 2002; Russell et al., 2003; Moran et al., 2005). In such ‘super-symbiotic systems’, the conventional introgression/transfection/curing approaches are usually not effective for evaluating the phenotypic effects of the respective symbiotic associates.

Experimental techniques for manipulating the endosymbiotic microbiota, by which a specific symbiont can be added to or eliminated from the host insect, enable us to evaluate the roles of the individual symbionts in the complex system. When the symbiont densities in super-infected host insects are experimentally suppressed (for example using antibiotics, heat stress or microinjection), the offspring sometimes include single-infected insects owing to stochastic symbiont sorting upon vertical transmission (Poinsot et al., 2000; Sasaki et al., 2002, 2005). In cases in which different symbionts are localized in different host tissues, selective symbiont transfection may be possible. In aphids, for example, the obligate symbiont Buchnera is exclusively endocellular, while a variety of facultative symbionts occur not only endocellularly but also freely in the hemolymph (Chen et al., 1996; Koga et al., 2003; Tsuchida et al., 2005). Thus, hemolymph injection can selectively transfer a facultative symbiont into recipient naive insects (Chen & Purcell, 1997; Chen et al., 2000; Fukatsu et al., 2001; Koga et al., 2003; Oliver et al., 2003, 2004; Tsuchida et al., 2004, 2006; Scarborough et al., 2005; Russell & Moran, 2006). Recently, novel antibiotic-based selective elimination techniques were devised for a super-symbiotic system in the pea aphid: ampicillin treatment selectively eliminated the facultative symbiont Serratia, while moderate rifampicin treatment selectively removed the obligate symbiont Buchnera from the host insect (Koga et al., 2003). The development of selective elimination techniques has proved to be a powerful tool in gaining an understanding of the biological roles played by individual symbionts in the complex system consisting of an insect and multiple endosymbiotic bacteria (Koga et al., 2003; Leonardo, 2004; Tsuchida et al., 2004; Sakurai et al., 2005; Leonardo & Mondor, 2006).

It should be noted, however, that these techniques have been developed and utilized in an empirical manner. There have been no studies on optimizing the selectivity and efficacy of the symbiont elimination techniques. No studies have evaluated the generality of the selective elimination techniques under variable parameters such as antibiotic dose and host genotype. In order to improve the utility of the techniques, a systematic survey of these aspects should be conducted.

In this study, using the pea aphid−Buchnera−Serratia endosymbiotic system as a model, the efficacy, generality, and fitness consequences of the selective elimination techniques at various antibiotic doses and under a variety of host genotypes were investigated.

Materials and methods

Materials

Naturally collected and experimentally generated strains of the pea aphid Acyrthosiphon pisum were established as clonal lineages, and were parthenogenetically maintained on seedlings of the broad bean Vicia faba at 20°C under a long-day photoperiod of 16 h light and 8 h dark (Table 1). These aphid strains were infected with either or both of the obligate symbiont Buchnera aphidicola (Munson et al., 1991) and the facultative symbiont Serratia symbiotica (Moran et al., 2005), but were not infected with any other secondary endosymbionts (Fukatsu et al., 2000; Tsuchida et al., 2002). Hereafter, these symbionts are referred to simply as Buchnera and Serratia, respectively. All insect materials were prefixed and preserved in acetone until use (Fukatsu, 1999).

1

Aphid strains used in this study

Strain name	Original locality/original strain and treatment	Buchnera*†	Serratia*†	Reference
AIST	Collected at Tsukuba, Ibaraki	+	−	Fukatsu et alet al. (2000)
IS	No data	+	+	Fukatsu et alet al. (2000)
KSM	Collected at Kasumigaura-machi, Ibaraki	+	+	Tsuchida et alet al. (2002)
NS	Collected at Nishine, Iwate	+	+	Tsuchida et alet al. (2002)
AISTIS	AIST, artificially Serratia-infected by injecting haemolymph of IS into AIST	+	+	Koga et alet al. (2003)
AISTAIST	AIST, injected with haemolymph of AIST	+	−	Koga et alet al. (2003)
ISdw	IS, injected with distilled water	+	+	This study
ISamp1+	IS, treated with ampicillin at the dose of 10 ng mg−1 of aphid	+	+	This study
ISamp2+	IS, treated with ampicillin at the dose of 100 ng mg−1 of aphid	+	+	This study
ISamp2−	IS, treated with ampicillin at the dose of 100 ng mg−1 of aphid	+	−	This study
ISamp3−	IS, treated with ampicillin at the dose of 1,000 ng mg−1 of aphid	+	−	This study
AISTIS/dw	AISTIS, injected with distilled water	+	+	This study
AISTIS/amp2+	AISTIS, treated with ampicillin at the dose of 100 ng mg−1 of aphid	+	+	This study
AISTIS/amp2−	AISTIS, treated with ampicillin at the dose of 100 ng mg−1 of aphid	+	−	This study
AISTAIST/dw	AISTAIST, injected with distilled water	+	−	This study
ISdw/rif	ISdw, treated with rifampicin‡	−	+	This study
ISamp2+/rif	ISamp2+, treated with rifampicin‡	−	+	This study
ISamp2−/rif	ISamp2−, treated with rifampicin‡	−	−	This study
AISTIS/dw/rif	AISTIS/dw, treated with rifampicin‡	−	+	This study
AISTIS/amp2+/rif	AISTIS/amp2+, treated with rifampicin‡	−	+	This study
AISTIS/amp2−/rif	AISTIS/amp2−, treated with rifampicin‡	−	−	This study
AISTAIST/dw/rif	AISTAIST/dw, treated with rifampicin‡	−	−	This study

*

The presence or absence of the symbionts was confirmed by diagnostic PCR.

†

+, presence; −, absence.

‡

Rifampicin was at the dose of either 2 ng mg⁻¹ or 20 ng mg⁻¹ of aphid (see text).

Diagnostic PCR

Total DNA was extracted from each of the insect materials using a conventional sodium dodecyl sulfate (SDS)/phenol method as described in Fukatsu (1999), and dissolved in 1000 μL of TE buffer [10 mM Tris-HCl (pH 8.0), 0.1 mM EDTA]. Diagnostic PCR detection of bacterial endosymbionts was performed as previously described (Koga et al., 2003). Buchnera was detected using the primers Buch16F (5′-GAGCTTGCTCTCTTTGTCGGCAA-3′) and Buch16R (5′-CTTCTGCGGGTAACGTCACGAA-3′) for the 16S rRNA gene (Tsuchida et al., 2002), and the primers BuchDnaK-AF1 (5′-ACAGAATTTAAAAAAGAACAAGGAATAGATT-3′) and BuchDnaK-AR1 (5′-ATTTTTGCTTTTTCCGCAGATT-3′) for the dnaK gene (Koga et al., 2003). Serratia was detected using the primers 16SA1 (5′-AGAGTTTGATCMTGGCTCAG-3′) and PASS5′cmp (5′-GCAATGTCTTATTAACAC AT-3′) for the 16S rRNA gene (Fukatsu et al., 2000), and the primers PASSGroE-AF1 (5′-CCTCAAGGCTGTGGCCG-3′) and PASSGroE-AR1 (5′-GAGTTTGCAGAGATGGTGCCTA -3′) for the groEL gene (Koga et al., 2003). To confirm the quality of extracted DNA samples, the elongation factor 1α gene of the host aphid was subjected to PCR detection using the primers ApisEF-AF1 (5′-CTGGAGAATTCGAAGCTGGTATTT-3′) and ApisEF-AR1 (5′-CACCCAAGGTGAAAGCCAATAG-3′) (Koga et al., 2003).

Ampicillin treatment

Using a fine glass needle, CO₂-anaesthetized 10-day-old adult aphids were injected with 0.1 μL of ampicillin solution per mg of body weight from the basement of a mid- or hind-leg. The antibiotic dosages were adjusted to 0 (distilled water), 1, 10, 100 or 1000 ng mg⁻¹ of body weight. The injected aphids were individually reared on broad bean seedlings, and their nymphs deposited from 24 to 48 h after injection were collected. These nymphs were defined as G1 of each of the isofemale lines. From the G1 nymphs, four insects were randomly chosen and reared on the plant until they became adult and produced a sufficient number of G2 offspring. The G1 mothers were then subjected to diagnostic PCR for their symbiont infection, and four G2 nymphs were randomly chosen and reared in the same manner to obtain G3 offspring. In this way, several Serratia-eliminated aphid strains that had been diagnosed as Serratia-free in both the G2 and G3 generations were established. During the maintenance of these strains, the infection status was periodically checked by diagnostic PCR.

Fitness measurement of ampicillin-treated aphids

In all the fitness experiments, insects at least 10 generations after the antibiotic injection were used in order to eliminate possible side effects of the treatment. Before the experiments, the infection status of each of the aphid strains was confirmed by diagnostic PCR. Ten-day-old aphids were allowed to deposit nymphs for 12 h. The newborn nymphs, defined as 0-day-old, were individually reared on the host plant at 20°C in the long-day regimen. Offspring production and survival of the insects were monitored every two or three days. Statistical analysis was performed using the Kruskal–Wallis test followed by Dunn's multiple comparison test.

Rifampicin treatment

Adult aphids were injected with rifampicin solution and their offspring were collected as described above for ampicillin treatment, except that different concentrations of the antibiotic were used. The antibiotic dosages were adjusted to 0, 2 and 20 ng mg⁻¹ of body weight. All G1 insects were reared on the host plant for a week, and subjected to diagnostic PCR for detection of the genes of Buchnera, Serratia and the host aphid.

Monitoring the reproduction of rifampicin-treated aphids

All newborn G1 offspring of each of the rifampicin-injected aphids were reared on the host plant, and their reproduction was monitored every two or three days until they reached 30 days old, because our preliminary observations revealed that the insects terminated their reproduction within 30 days after birth. Then, these G1 insects were subjected to diagnostic PCR as described above. Their offspring, if there were any, were transferred to a new host plant. To confirm the absence of Buchnera and the presence of Serratia, three or four insects per line were subjected to diagnostic PCR every one or two generations.

Whole mount FISH

Ovaries were dissected out of adult insects in 80% ethanol, and were fixed in Carnoy's solution (ethanol: chloroform: acetic acid=6:3:1 [v/v]) for several days. The fixed embryos were gently washed four times with absolute ethanol for 10 min each time, and were stored in ethanol at −20°C until use. The embryos were incubated three times with PBST (136.9 mM NaCl, 2.7 mM KCl, 1.5 mM KH₂PO₄, 8.1 mM Na₂HPO₄, 0.3% Triton X-100) for 5 min each time at room temperature, and were prehybridized three times with the hybridization buffer [20 mM Tris-HCl (pH 8.0), 0.9 M NaCl, 0.01% SDS, 30% formamide] (Fukatsu et al., 1998) for 5 min each time. Then, the embryos were hybridized with 10 nM each of the Serratia-targetting probe Cy3-PASSisR (5′-CCCGACTTTATCGCTGGC-3′) and the Buchnera-targetting probe Cy5-ApisP2a (5′-CCTCTTTTGGG TAGATCC-3′) (Koga et al., 2003) in the hybridization buffer overnight. Nuclei of the host cells were counterstained with Sytox Green (Molecular Probes). Then, the embryos were washed four times with PBSTx for 15 min each time, mounted in Slowfade antifade solution (Molecular Probes), and observed under an epifluorescence microscope (Axiophoto, Carl Zeiss) or a laser confocal microscope (Pascal 5, Carl Zeiss). To confirm specific detection of the symbionts, a series of control experiments were conducted as previously described (Sakurai et al., 2005).

Results and discussion

Selective elimination of Serratia from a naturally infected aphid strain by ampicillin treatment

The aphid strain IS was originally infected with Serratia in addition to Buchnera (see Table 1). Adult aphids of the strain IS were injected with ampicillin solution at various doses, and their parthenogenetic offspring were examined for symbiont infection by diagnostic PCR. In the G1 offspring of the aphids treated with 1000 ng of the antibiotic per mg of body weight, all insects were Buchnera-positive, while some insects were diagnosed as Serratia-negative (data not shown). In the G3 offspring at this dose, all aphid lines were Serratia-negative, although they still retained Buchnera infection. The selective elimination of Serratia was dependent on the antibiotic dose: administration of ampicillin at the doses of 1000, 100 and 10 ng mg⁻¹ resulted in 100%, 60% and 0% elimination of Serratia infection in the G3 offspring (Table 2). FISH analysis histologically confirmed the successful elimination of Serratia infection (Fig. 1).

2

Selective elimination of Serratia infection by ampicillin treatment

Aphid strain	Symbiont	Symbiont-eliminated lines/lines examined (% elimination)
		Dose of ampicillin (per mg of aphid)
		0 ng	10 ng	100 ng	1000 ng
IS	Serratia	0/12 (0%)	0/16 (0%)	12/20 (60%)	12/12 (100%)
	Buchnera	0/12 (0%)	0/16 (0%)	0/20 (0%)	0/12 (0%)
KSM	Serratia	NE	0/9 (0%)	7/9 (78%)	10/10 (100%)
	Buchnera	NE	0/9 (0%)	0/9 (0%)	0/10 (0%)
NS	Serratia	NE	NE	NE	13/13 (100%)
	Buchnera	NE	NE	NE	0/13 (0%)
AISTIS	Serratia	0/10 (0%)	0/20 (0%)	14/17 (82%)	18/18 (100%)
	Buchnera	0/10 (0%)	0/20 (0%)	0/17 (0%)	0/18 (0%)

Adult aphids were injected with the antibiotic solution at various concentrations, and successive parthenogenetic lines were examined for symbiont infection by diagnostic PCR at the 3rd generation after injection.

NE, not examined.

1

Whole-mount in situ hybridization of aphid embryos targetting Serratia. (a) Embryos of the Serratia-eliminated strain IS^amp2−. (b) Embryos of the original Serratia-infected strain IS. Orange signals indicate Serratia cells, while green signals are host nuclei visualized with Sytox Green. Bars, 200 μm.

Re-elimination of Serratia from an artificially infected aphid strain

The aphid strain AIST^IS was originally Serratia-free and was experimentally infected with Serratia by microinjection of hemolymph from the aphid strain IS (see Table 1). When adult aphids of the strain AIST^IS were treated with 1000 ng mg⁻¹ of ampicillin, their G3 offspring were all Serratia-negative while they retained Buchnera infection. The selective elimination of Serratia in the strain AIST^IS was also dependent on the antibiotic dose, as in the strain IS (Table 2).

Selective elimination of Serratia from naturally infected aphid strains of different genetic backgrounds

Because the aphid strains KSM and NS were collected at distant localities in Japan, they were probably genetically different from the aphid strains IS and AIST, and were naturally infected with Serratia in addition to Buchnera (see Table 1). When adult aphids of the strains KSM and NS were treated with 1000 ng mg⁻¹ of ampicillin, their G3 offspring were all Serratia-free while their Buchnera infection was not affected. In the strain KSM, the Serratia-specific elimination was, once again, dependent on the antibiotic dose (Table 2).

Stability of Serratia elimination in the ampicillin-treated aphid lines

Some of the G3 aphid lines were maintained, and their symbiont infection was inspected by diagnostic PCR in the G10 offspring. In all the aphid lines, the symbiont infection did not change through seven successive host generations (Table 3), indicating that the symbiont infection patterns manipulated by the ampicillin treatment were stable once established.

3

Stability of symbiont infection in the ampicillin-treated aphid lines

Aphid strain	Symbiont	Symbiont-eliminated lines/lines examined*
		Dose of ampicillin (per mg of aphid)
		10 ng	100 ng	1000 ng
IS	Serratia	0/5; 0/5*	4/4; 4/4	4/4; 4/4
	Buchnera	0/5; 0/5	0/4; 0/4	0/4; 0/4
KSM	Serratia	0/2; 0/2	2/2; 2/2	1/1; 1/1
	Buchnera	0/2; 0/2	0/2; 2/2	0/1; 0/1
NS	Serratia	NE	NE	12/12; 12/12
	Buchnera	NE	NE	0/12; 0/12
AISTIS	Serratia	0/2; 0/2	4/5; 4/5	4/4; 4/4
	Buchnera	0/2; 0/2	0/5; 0/5	0/4; 0/4

Some of the Serratia-eliminated/uneliminated lines at the 3rd generation after injection (see Table 2) were maintained, and were examined for symbiont infection at the 10th generation after injection.

*

For example, [0/5; 0/5] indicates [0 symbiont-eliminated lines/5 lines examined at the third generation; 0 symbiont-eliminated lines/5 lines examined at the tenth generation].

NE, not examined.

Selective elimination of Serratia infection is dependent on ampicillin dose and independent of host genotype.

From these results, it was concluded that the ampicillin treatment can efficiently eliminate the facultative symbiont Serratia from the disymbiotic aphid strains without affecting the essential symbiont Buchnera. The elimination efficacy was dose-dependent: 1000, 100 and 10 ng mg⁻¹ of the ampicillin treatment resulted in complete, partial and no elimination of Serratia, respectively. The dose dependence was consistently observed among aphid strains of different geographic origins, probably indicating a generality of the elimination technique irrespective of the host genotype. Once Serratia was eliminated by the ampicillin treatment, the manipulated status of endosymbiotic microbiota was stably maintained through host generations. On the basis of these findings, we suggest that the ampicillin treatment would be a general and useful technique for investigating the biological roles of the facultative endosymbiotic associate in the pea aphid. In fact, it has recently been shown that the ampicillin treatment is also effective for the selective elimination of other facultative symbionts of the pea aphid, such as Regiella (also called PAUS or U-type) (Leonardo, 2004; Tsuchida et al., 2004) and Rickettsia (also called PAR) (Sakurai et al., 2005). In this context, the applicability of the selective elimination technique might be wider than expected, possibly extendable to other insects and their endosymbionts.

Fitness effects of Serratia elimination in a naturally infected aphid strain

To gain insights into the biological consequences of Serratia infection for the host aphid, we measured several fitness parameters of the aphid strains IS^dw, IS^amp1+, IS^amp2+, IS^amp2− and IS^amp3−, which are ampicillin-treated derivatives of the naturally disymbiotic strain IS and are thus genetically identical to each other except for Serratia infection status (see Table 1). The time to reproduction and the total number of offspring did not differ between the Serratia-eliminated and Serratia-infected strains (Fig. 2a and b). Only the longevity was significantly improved in the Serratia-eliminated strains in comparison with the Serratia-infected strains (Fig. 2c). Note that the time to reproduction and the total number of offspring strongly affect the aphid fitness, while the longevity does so only weakly, and that few aphids die a natural death in the fields under the extremely intense predatory pressures they generally suffer (Minks & Harrewijn, 1998). These results suggest that, in the naturally disymbiotic aphid strain, Serratia infection exerts a slightly negative or nearly neutral fitness effect on the host aphid, which agrees with the results of a previous study (Koga et al., 2003).

2

Fitness effects of Serratia elimination in naturally disymbiotic aphid strains. (a) Time to reproduction; (b) total number of offspring; (c) longevity. Filled bars represent aphid strains retaining the Serratia infection after the treatment; open bars represent aphid strains cured of the Serratia infection after the treatment. Means and SDs are shown (n=20, 10, 27, 27 and 27 for IS^dw, IS^amp1+, IS^amp2+, IS^amp2− and IS^amp3−, respectively). Different alphabetical characters indicate statistically significant differences (P<0.05).

Fitness effects of Serratia elimination in an artificially infected aphid strain

In the same way, we measured the fitness parameters of the aphid strains AIST^IS/dw, AIST^IS/amp2+, AIST^IS/amp2− and AIST^AIST/dw, which are ampicillin-treated derivatives of the artificial disymbiotic strain AIST^IS and are thus genetically identical to each other except for Serratia infection status (see Table 1). The time to reproduction did not differ between the Serratia-eliminated and Serratia-infected strains (Fig. 3a). The total number of offspring was slightly higher in the Serratia-eliminated strains than in the Serratia-infected strains (Fig. 3b). The longevity was significantly improved in the Serratia-eliminated strains in comparison with the Serratia-infected strains (Fig. 3c). These results suggest that, in the artificial disymbiotic aphid strain, Serratia infection exerts a negative fitness effect by slightly reducing the host fecundity. In a previous study, it was reported that, when the naturally monosymbiotic aphid strain AIST was experimentally infected with Serratia by hemolymph injection, the resultant disymbiotic strain AIST^IS initially suffered drastic reduction in all the fitness parameters measured, including body weight, time to reproduction, total number of offspring, and longevity (Koga et al., 2003). However, the negative effects were considerably attenuated after 8 months of coadaptation (Koga et al., 2003). This study was performed over 8 months after the establishment of the strain AIST^IS, which may explain the relatively weak negative effects of Serratia infection detected in this study.

3

Fitness effects of Serratia elimination in artificial disymbiotic aphid strains. (a) Time to reproduction; (b) total number of offspring; (c) longevity. Filled bars represent aphid strains retaining the Serratia infection after the treatment; open bars represent aphid strains cured of the Serratia infection after the treatment. Means and SDs are shown (n=11, 10, 29 and 20 for AIST^IS/dw, AIST^IS/amp2+, AIST^IS/amp2− and AIST^AIST/dw, respectively). Different alphabetical characters indicate statistically significant differences (P<0.05).

Implication for the prevalence of Serratia infection in pea aphid populations

These experimental data consistently indicate that, at least after a long-term association, Serratia infection generally exhibits a slightly negative or nearly neutral effect on the host fitness. In natural populations of the pea aphid, the infection frequency of Serratia is generally high: for example, 88% (50/57) in one study in California (Chen & Purcell, 1997); 56% (43/77) in another survey in California (Leonardo & Muiru, 2003); 33% (286/855) in Japan (Tsuchida et al., 2002); 93% (67/72) in England (Haynes et al., 2003); and 90–100% on pea, 7–8% on alfalfa and 0–3% on red clover in France (Simon et al., 2003). The prevalence in natural populations must be relevant to the fact that Serratia infection scarcely impedes survival and reproduction of the host aphid.

Mechanism underlying the selective symbiont elimination by ampicillin

Why can the ampicillin treatment selectively eliminate the facultative symbiont Serratia without affecting the essential symbiont Buchnera although it coexists in the same host body? The antibiotic action of ampicillin, an inhibitor of bacterial cell-wall synthesis, must be the crucial factor responsible for the selectivity. The cell wall of Buchnera is reduced (Hinde, 1971), and the genome of Buchnera lacks some of the genes for the biosynthetic pathway of the cell wall (Shigenobu et al., 2000). In contrast, the presence of the cell wall has been shown by electron microscopy in Serratia, Regiella, Rickettsia and other facultative symbionts of the pea aphid (Fukatsu et al., 2000; Moran et al., 2005; Sakurai et al., 2005; Tsuchida et al., 2005). Another possible factor involved in the selectivity may be the different locations of the facultative and essential symbionts in vivo: Buchnera is exclusively endocellular in the primary bacteriocytes, whereas Serratia, Regiella and Rickettsia and other facultative symbionts occur not only in the secondary bacteriocytes and the sheath cells endocellularly but also in the hemolymph extracellularly (Chen et al., 1996; Chen & Purcell, 1997; Fukatsu et al., 2000; Koga et al., 2003; Moran et al., 2005; Sakurai et al., 2005; Tsuchida et al., 2005). Previous therapeutic studies have shown that penicillin and allied antibiotics, including ampicillin, are often not effective for endocellular pathogens such as Rickettsia owing to inefficient permeation into the host cells (Wisseman et al., 1974, 1982). It is plausible that these two factors synergistically contribute to the selective symbiont elimination by the antibiotic treatment.

Selective elimination of Buchnera by rifampicin treatment

Rifampicin, an antibiotic that inhibits prokaryotic transcription by binding to DNA-dependent RNA polymerase, is known to be effective for eliminating Buchnera infection from the pea aphid, and has been utilized for studies on the biological role of the essential symbiont (Houk & Griffiths, 1980; Ishikawa & Yamaji, 1985; Sasaki et al., 1990; Rahbe et al., 1994). It has recently been reported that rifampicin treatment at a low dose eliminated Buchnera infection but not Serratia infection (Koga et al., 2003). To corroborate this result, disymbiotic adult aphids were injected, which represented four strains under two distinct aphid genotypes, with 2 ng of the antibiotic per mg of body weight, and examined their offspring for symbiont infection by diagnostic PCR. In all of the strains, Buchnera infection was eliminated from all the G1 offspring while Serratia infection was retained (Table 4). Similarly, Buchnera infection was selectively and completely cured with a higher dose (20 ng mg⁻¹ body weight) of rifampicin (Table 4). The same treatments also eliminated Buchnera infection from the monosymbiotic aphid strains (data not shown).

4

Selective elimination of Buchnera infection by rifampicin treatment

Aphid strain	Symbiont	Symbiont-eliminated lines/lines examined
		Dose of rifampicin (per mg of aphid)
		0 ng	2 ng	20 ng
ISdw	Serratia	0/12 (0%)	0/12 (0%)	0/12 (0%)
	Buchnera	0/12 (0%)	12/12 (100%)	12/12 (100%)
ISamp2+	Serratia	0/12 (0%)	0/12 (0%)	0/12 (0%)
	Buchnera	0/12 (0%)	12/12 (100%)	12/12 (100%)
AISTIS/dw	Serratia	0/12 (0%)	0/12 (0%)	0/12 (0%)
	Buchnera	0/12 (0%)	12/12 (100%)	12/12 (100%)
AISTIS/amp2+	Serratia	0/12 (0%)	0/12 (0%)	0/12 (0%)
	Buchnera	0/12 (0%)	12/12 (100%)	12/12 (100%)

Adult aphids were injected with the antibiotic solution at various concentrations, and their offspring at the next generation were examined by diagnostic PCR.

Serratia-dependent survival and reproduction of Buchnera-eliminated aphid strains

Because Buchnera is involved in the production of essential amino acids and other nutrients, the pea aphid suffers retarded growth and sterility when deprived of the symbiont (Houk & Griffiths, 1980; Douglas, 1989). In fact, when the monosymbiotic aphid strains were treated with rifampicin, their aposymbiotic offspring became completely sterile (Fig. 4). On the other hand, when the disymbiotic aphid strains AIST^IS/amp2+, IS^amp2+ and IS^dw were treated with rifampicin at a dose of 2 ng mg⁻¹, their Serratia-infected offspring were, although free of Buchnera infection, able to produce offspring (Fig. 4a). These Buchnera-eliminated and Serratia-infected aphids were smaller in size and much less fecund than Buchnera-infected normal aphids (data not shown), and persisted up to the 8th generation after the antibiotic treatment (Fig. 4a). Exceptionally, an IS^amp2+ strain survived much longer, up to the 26th generation (Fig. 4a). FISH analysis confirmed that in these aphid strains Buchnera was completely removed while Serratia was localized in a number of roundish bacteriocytes where Buchnera would be harboured in normal aphids (Fig. 5). These results corroborate the previous report on Serratia-dependent survival and reproduction of Buchnera-eliminated aphid strains in A. pisum (Koga et al., 2003).

4

Reproduction of Buchnera-eliminated aphid strains in the presence of Serratia. (a) Survival of aphid strains after rifampicin treatment at the dose of 2 ng mg⁻¹ body weight; (b) survival of the same aphid strains after rifampicin treatment at the dose of 20 ng mg⁻¹ body weight. Each of the circles indicates an isoFEMale lines originating from Serratia-infected strains, while open circles represent those originating from Serratia-free strains.

5

Whole-mount FISH of aphid embryos targetting Buchnera (green) and Serratia (red). (a) A young embryo at stage 16 of the Buchnera-eliminated strain AIST^IS/dw/rif. (b) A mature embryo at stage 20 of the Buchnera-eliminated strain AIST^IS/dw/rif. (c) A mature embryo at stage 20 of the naturally Serratia-infected strain IS. Host nuclei are counterstained in blue. Bars, 50 μm. The developmental stages of the aphid embryos are according to Miura et al. (2003).

Effects of rifampicin dose on the reproductive ability of the Buchnera-eliminated aphid strains

When the disymbiotic aphid strains were treated with a higher dose of rifampicin (20 ng mg⁻¹), the effects on their reproductive ability differed between the aphid strains. When the strains AIST^IS/amp2+ and AIST^IS/dw were deprived of Buchnera infection by the antibiotic treatment, their Serratia-infected offspring persisted for up to 8th generation (Fig. 4b). In contrast, when the strains IS^amp2+ and IS^dw were similarly deprived of Buchnera infection, the offspring were unable to reproduce although they were Serratia-infected (Fig. 4b).

Possible factors relevant to the different reproductive ability in response to the rifampicin treatment: host genotype vs. symbiotic history

The rifampicin treatments at doses of 2 and 20 ng mg⁻¹ consistently and selectively eliminated Buchnera infection from all the disymbiotic aphid strains without loss of Serratia infection (Table 4). Interestingly, however, the reproductive ability of the Buchnera-eliminated, Serratia-infected aphids was dependent on the aphid strain and the antibiotic dose: the strains with AIST background (i.e. AIST^IS/amp2+ and AIST^IS/dw) were able to reproduce after the rifampicin treatment at a dose of either 2 or 20 ng mg⁻¹, whereas the strains with IS background (i.e. IS^amp2+ and IS^dw) were fertile after treatment at the dose of 2 ng mg⁻¹ but sterile at the dose of 20 ng mg⁻¹ (Fig. 4). Why did the effects of the higher dose of rifampicin (20 ng mg⁻¹) differ between the AIST aphids and the IS aphids? One possibility is that the aphid genotypes themselves may be responsible for the difference: the AIST genotype might perform better than the IS genotype under the Buchnera-free and Serratia-infected condition. Another possibility is that the symbiotic histories that these aphid strains have experienced may be relevant to the difference: the IS aphids were naturally infected with Serratia whereas the AIST aphids had originally been free of Serratia and were experimentally transfected with the symbiont from the IS aphids (see Table 1). Hence, it is expected that the IS aphids experienced coadaptation and coevolution with the facultative symbiont over evolutionary time, whereas the AIST aphids did not. In accordance with this expectation, it was reported that the proliferation and localization of Serratia were neatly regulated in the IS aphids whereas the AIST aphids often exhibited disordered localization and massive proliferation of the symbiont (Koga et al., 2003). It is plausible, although speculative, that the reproduction of the AIST aphids even after the higher dose of rifampicin treatment may be relevant to the higher infection density of Serratia in these aphids. Of course, these possibilities are not mutually exclusive: both the aphid genotype and the symbiotic history might contribute to the different reproductive abilities under the Buchnera-free and Serratia-infected condition. Although not examined in this study, symbiont genotype can also affect the phenotype (Oliver et al., 2005; Russell & Moran, 2006). To address these aspects in detail, further experimental studies will be needed.

Mechanism underlying the selective symbiont elimination by rifampicin

Why can the rifampicin treatment selectively eliminate the essential symbiont Buchnera without affecting the facultative symbiont Serratia? Although the mechanism underlying the selective symbiont elimination is totally unknown, we suggest here several putatively relevant factors. Owing to the reductive genome evolution through the long history of endosymbiosis (Mira et al., 2001; Wernegreen, 2002), Buchnera has lost many genes for outer membrane proteins and those for the biosynthesis of lipopolysaccharides (Shigenobu et al., 2000). Hence, the outer membrane of the symbiont is inferred to be structurally fragile, an inference corroborated by observations with an electron microscope (Hinde, 1971). Because the bacterial outer membrane can act as an effective barrier to antibiotics (Hancock, 1997), the reduced outer membrane may potentially facilitate the permeability of these drugs. In fact, it has been reported that bacterial mutants in biosynthetic genes for lipopolysaccharides exhibit increased sensitivity to hydrophobic antibiotics, including rifampicin (Vaara, 1993). These findings may add weight to the idea that Buchnera, with its reduced outer membrane, is highly sensitive to rifampicin. In contrast, electron microscopy has indicated that Serratia has a thick outer membrane (Fukatsu et al., 2000; Moran et al., 2005). In endocellular bacterial parasites of the genus Rickettsia, mutations in the gene for the DNA-dependent RNA polymerase β subunit (rpoB) have been identified as contributing to their naturally occurring resistance to rifampicin (Drancourt & Raoult, 1999). In this context, the rpoB structure of Serratia and its binding property to the antibiotic are of interest.

Conclusion and perspective

Recent studies have shown that not only obligate symbionts such as Buchnera but also facultative symbionts such as Serratia, Hamiltonella and Regiella play important biological roles for their host insects, particularly in specific ecological contexts. For example, tolerance to high temperature is conferred by Serratia and Hamiltonella (Montllor et al., 2002, Russell & Moran, 2006); resistance to parasitoid wasps is caused by Hamiltonella and Serratia (Oliver et al., 2003, 2005); broadening of host plant range is enabled by Regiella (Tsuchida et al., 2004); resistance to parasitic fungi is conferred by Regiella (Scarborough et al., 2005); and induction of winged morph and sexual generation is modified by Regiella (Leonardo & Mondor, 2006). Insects represent the majority of the biodiversity in the terrestrial ecosystem (Wilson, 1988), and more than half of all insect species are estimated to be in association with endosymbiotic microorganisms (Buchner, 1965, Jeyaprakash & Hoy, 2000; Werren & Windsor, 2000). We are only now becoming aware of the general importance of unrecognized microbial associates that may substantially affect various life-history traits of these insects. The results presented in this study may provide a basis for the development of new protocols for experimentally manipulating the endosymbiotic microbiota of aphids and other insects, thereby significantly enriching our understanding of insect ecology and evolution.

Acknowledgements

We thank S. Koike, J. Makino, K. Nomura and W. Kikuchi for technical and secretarial assistance. This study was financially supported by the National Institute of Advanced Industrial Science and Technology (AIST).

Author notes

Editor: Christoph Tebbe