Parthenogenetic A. pisum females were collected from across the United States between 1998 and 2007 (Table S1). For each clone, a single female was used to establish clonal lineages maintained continuously under long day (16:8 L:D) conditions. Experimental aphid lines were kept in a growth chamber at 20 °C on Vicia faba seedlings in cup cages [15]. Short day conditions were used to induce sexual forms for clones 8–10–1 and 5A, which were reciprocally mated to yield F₁ aphid clones, as described by Moran and Dunbar [16]. Individuals hatching from the sexually produced eggs were isolated and allowed to establish full sib clonal lineages under long day conditions. For all experiments, clones were divided into 3 sub-clones and allowed to reproduce for 3 generations prior to collection to control for maternal and environmental effects. Each experiment was replicated twice, beginning with the establishment of new subclones.

Adult viviparous females were placed on fresh V. faba seedlings and allowed to deposit nymphs for 12 hours, after which the adults were removed and the nymphs allowed to develop for 6 days, to their 4^th instar. Fourth instar aphids were dissected in buffer A (250 mM sucrose, 35 mM Tris-HCl, 25 mM KCl, 10 mM MgCl₂) in a watch glass. All bacteriocytes were identified, separated and counted under 6× magnification.

Adult viviparous A. pisum were placed on fresh V. faba seedlings and allowed to deposit nymphs for 12 hours. Nymphs were either allowed to develop for six days to their fourth instar or immediately collected at their first instar. Individual nymphs were collected in pestle tubes, frozen in liquid nitrogen, and crushed with a pestle. The resulting homogenized tissue was treated according to the Qiagen DNEasy kit.

To isolate DNA from individual bacteriocytes, single bacteriocytes from the aphids used for bacteriocyte counts were collected in pestle tubes, frozen in liquid nitrogen and crushed. Due to the small amount of starting material, the resulting homogenate was treated with lysis buffer [17] and then washed twice with phenol:chloroform:isoamyl alcohol 25:24:1, then once with chloroform. The DNA was then precipitated with sodium acetate and ethanol and resuspended in low TE (10 mM EDTA, 100 mM Tris-HCl). All DNA was treated with RNAse I at 37 °C for 30 minutes. Three aphids from each of the 3 subclones from each clone were used for the experiments.

Buchnera titer was measured by comparing the number of Buchnera genomes to the number of aphid genomes using a single copy gene from both the aphid and the symbiont. This provided a rough correction for size differences between aphid clones, though some aphid cells are polyploid [18]. Aphid genomes were counted by assessing copy number of the gene encoding elongation factor 1-alpha (ef1α), while Buchnera genomes were counted by using the gene encoding adenosylmethionine-8-amino-7-oxononanoate aminotransferase (bioA). Primers used were ApEF1-alpha 107F 5′ - CTGATTGTGCCGTGCTTATTG - 3′, ApEF1-alpha 246R 5′ - TATGGTGGTTCAGTAGAGTCC - 3′, BuchAPS bioA 374F 5′ - AGTATTGGCAAGCATTAGGGC - 3′, BuchAPS bioA 526R 5′ - AAAAGAAGAAACTGGTCGTC - 3′. Standards of 10⁷ copies were prepared according to the method of [10] for each gene. For each sample from 1^st instar aphids, the number of copies of ef1α and bioA were compared on a Roche LightCycler using the FastStart DNA Master^plus SYBR Green I kit according to the kit instructions. Copy number was determined using LightCycler 3.0 software in comparison to the standards for each gene. For individual bacteriocytes, only the copy number of bioA was assessed, though the bacteriocytes are polyploid [18]. For each clone, three aphids from each subclone were tested, and each clone was measured twice, starting with the initiation of new subclones.

Buchnera titer was determined by the ratio of Buchnera genomes to A. pisum genomes. The regression of Buchnera genome copies and aphid genome copies was linear but the slopes of the variables were unequal and the intercepts were non-zero, so ANCOVA analysis for unequal slopes was applied in JMP 8 (SAS). Buchnera genome copy number of individual bacteriocytes was log transformed and analyzed by ANOVA. Bacteriocyte counts were normally distributed and analyzed by MANOVA. Summary statistics are included in Table S2.

Buchnera titer (Buchnera genomes per aphid genome) varied significantly among field-collected clones of A. pisum in the first instar (F_9,95 = 3.21, p = 0.0019, ANCOVA), with the lowest titer clone–5A–having an average of 35 Buchnera genomes per aphid genome, while the highest titer clone–File–had an average titer of 73.4 (Figure 1). The copy number of both ef1α and bioA varied significantly between clones (p < 0.0001 ANOVA), but the variation in ef1α copy number only explained 4.4% of the variance in Buchnera titer while bioA copy number explained 36.6% of the variance based on measuring effects of each variable independently.

Buchnera titer also varies across the development of the host. Four clones were tested, and there were significant changes in the titer of Buchnera from 1^st to 4^th instar (F_7,97 = 10.37, p < 0.0001; ANCOVA, Figure 2), with clones 8-10-1 and Tuc7 increasing significantly in titer (p = 0.0005 and p = 0.027, respectively student's t-test), and the titer in clones 5A and 9-2-1 not changing significantly (p = 0.82, p = 0.33, respectively student's t-test). Titer varied significantly between the clones at the 4^th instar (Figure 2), with clones 8-10-1 and Tuc7 exhibiting similar titer and clones 5A and 9-2-1 having significantly lower titer (p < 0.05, Tukey's HSD). The copy number of both Buchnera genomes and aphid genomes varied significantly between the 4^th instars of the clones tested. In 4^th instar aphids, Buchnera genome copy number explained 62.9% of the variation in titer (p < 0.0001, ANOVA) while the aphid genome accounted for 18.5% (p < 0.0001, ANOVA) based on measuring effects of each variable independently.

To determine the pattern of heritability in titer, clones 5A and 8-10-1 were reciprocally mated to produce a panel of full-sib clones with distinct genotypes. The Buchnera titer of 1^st instar aphids were significantly different among F₁ clones (F_6,93 = 2.9, p = 0.0004, ANCOVA, Figure 3). The clones also had significantly higher and lower average titers than those observed in the lab clones–those established from field collected asexual females–or their parental clones (F_1,15 = 8.48, p = 0.0042, ANCOVA). Titers ranged from 23.4 Buchnera genomes per aphid genome in clone 58-3″B and 105.6 in clone 85-1″F. These values were significantly lower and higher than those observed in the parental clones (p < 0.05, Tukey's HSD).

As Buchnera are maternally transmitted, aphids from the two matrilines have different Buchnera genotypes. There was no significant difference observed in the Buchnera titer of F₁ clones from the two matrilines, suggesting that differences between the genomes of Buchnera from 5A and 8-10-1 are not responsible for the differences in symbiont titer between these two clones (F_1,93 = 0.046, p = 0.83, ANCOVA).

Figure 4 shows the average number of bacteriocytes per aphid at the 4^th instar. Bacteriocyte counts varied significantly between clones tested (F_3,172 = 21.3, p < 0.0001, ANOVA). These data are mostly consistent with the Buchnera titer of 4^th instar aphids (Figure 2), with Tuc7 and 8-10-1 having similar numbers of bacteriocytes per aphid while 9-2-1 and 5A have significantly fewer (p < 0.05, Tukey's HSD). Clone 9-2-1 has fewer bacteriocytes than clone 5A (p < 0.05, Tukey's HSD), though these clones do not differ in the titer of Buchnera at the 4^th instar (Figure 2). While variation in Buchnera titer in 1^st instars is not consistently reflected in measures of number of bacteriocytes or the number of Buchnera per bacteriocyte, the symbiont titer of 4^th instar aphids recapitulates the patterns observed in the number of bacteriocytes in the clones tested, suggesting that control of bacteriocyte number is one mechanism by which the symbiont populations within an individual is regulated by the host.

The number of Buchnera genomes per bacteriocyte was significantly different among the clones examined (F_3,94 = 10.24, p < 0.0001, MANOVA, Figure 5). 9-2-1 had the fewest Buchnera per bacteriocyte, significantly fewer than all the clones except 8-10-1 (p < 0.05, Tukey's HSD). Tuc7 and 5A had an equivalent number of Buchnera per bacteriocyte, while 8-10-1 had fewer than Tuc7, but the difference between 5A and 8-10-1 was not significant.

Together, results for bacteriocyte numbers are consistent with the differences in total Buchnera titers in three of the four clones examined. 5A and 8-10-1 exhibited the largest difference in 1^st instar Buchnera titer, and 8-10-1 had significantly more bacteriocytes than 5A. In these clones, there was no significant difference between the numbers of Buchnera genomes per bacteriocyte, suggesting that the number of bacteriocytes may be responsible for the difference in Buchnera titer between the clones. Tuc7 had more bacteriocytes than 5A and more Buchnera per bacteriocyte than 8-10-1, though it had intermediate 1^st instar titer compared to these clones. While there was a strong correlation between bacteriocyte numbers and Buchnera titer for the four clones tested (R² = 0.62), the relationship was not significant (F_1,4 = 3.39, p = 0.2, ANCOVA) though a lack of power likely contributed to this. Significant variation in bacteriocyte numbers was also seen in the F₁ clones (F_5,106 = 8.00, p < 0.0001, ANOVA, Figure S1), though no comparison could be made between the Buchnera titer and bacteriocyte number in these clones, as they were measured at 1^st and 4^th instars, respectively.

The clones used in the present study were previously assayed for their dietary requirements of essential amino acids [9]. In that study, we measured the mass of aphids reared on artificial diets with and without essential amino acids. We used the ratio of aphid mass on the diet without essential amino acids to the mass of aphids reared on diet with all the essential amino acids as a measure of the dietary requirement for essential amino acids for each clone. As different individuals were measured for the previous study and the current study, we compared the mean amino acid requirement of a clone against the mean 1^st instar Buchnera titer as measured in the current study. We found a significant, positive association between amino acid requirements and Buchnera titer for the set of clones tested (R² = 0.38, F_1,14 = 8.71, p = 0.01, ANCOVA, Figure S2).