Supplementary Methods

Phylogenetic Analysis of speG and arcA loci

Because multiple paralogs of arcA and speG exist in S. epidermidis species we began by performing large-scale gene family phylogenetic analysis and isolated only closely related orthologs that clearly grouped with genes from ACME. Preliminary phylogenies were constructed using PAUP 4b10(1), and clades containing genes from USA300 and closely related S. epidermidis genes were selected for further analysis.

For parsimony analysis we used random addition (RA) in PAUP(1) followed by tree bisection and reconnection (TBR), Nreps=1000, with characters and character state transformations given equal weight. Node support was calculated based on 1000 bootstrap pseudoreplicates (2) using the heuristic search strategy above.

For MrBayes 3.1.2(3) we used the following settings: Ngen=100000000 nruns=2, samplefreq=100, nst=1, rate=equal, nucmodel=4by4, ngammacat=4, nbetacat=5, covarion=no, omegavar=equal. This resulted in a best speG LogLikelihood of –8875.37 [final average standard deviation of split frequencies fASDSF=0.004437], and arcA LogLikelihood was –2541.35 [fASDSF=0.003807]. More complex models in MrBayes (e.g. nst=6 with Dirichlet prior of 1.0 for all substitution rates, rate=invgamma, nucmodel=4by4, ngammacat=4, nbetacat=5, covarion=yes, omegavar=equal) yielded identical topologies and similar clade credibility values. Best state speG and arcA LogLikelihoods under these assumptions were -8678.58 and -2508.03, respectively. Parsimony and Bayesian tree topologies were 100% consistent differing only in resolution at several nodes.

To assess congruence we used the incongruence-length difference (ILD) test (4) or HOMPART function in PAUP 4b10 (nreps=1000 of RA followed by TBR). Gene partitions of speG and aliD were congruent (P=1.0) so these partitions were combined for further analysis. The arcA partition showed significant incongruence with the combined speGaliD partition (P=0.0020). Deletion of sequences derived from ATCC12228 eliminated all incongruence in this analysis (P=1.0).

For tree reconciliation we reconciled trees both by hand and using the TreeMap 3.0(5), assigning duplications, losses, and recombination events equal weight, and solved for the solution that minimized the sum of these events. Unresolved nodes were resolved to minimize differences between topologies.

Estimation of divergence times and evolutionary rates

Using the phylogenetic tree tip dates (year of isolation or admission of an isolate) to calibrate our temporal estimation in combination with two among-lineage rate variation models in BEAST 1.7.4 (6):1) strict clock, where one single evolutionary rate is applied to the whole tree, and 2) uncorrelated lognormal relaxed clock, where evolutionary rates are allowed to vary across the tree. See Supplementary table 2 for isolate information. Where a range of dates was the most precise isolation date available we used the median year, e.g., 2008 for November 2007 to March 2009 (7) in the case of CO-23, IS-91, IS-24, and IS-111. The evolutionary rates were drawn from a prior uniform distribution in the range 10-10–10-2 substitutions x site-1 x year-1. A constant-size coalescent prior was placed on the tree adapted for serially sampled sequence data (8). The two clock models were contrasted using Bayes factors (9, 10) in Tracer 1.5 (http://tree.bio.ed.ac.uk/software/tracer).

Blastn-based Screen

For S. epidermidis we used two publicly available whole genome sequences (ATC12228 and RP62A) as well as 29 draft genome sequences. For BLAST-based screening in S. aureus we used 17 publicly available whole genome sequences, 57 publicly available draft sequences, and 51 draft sequences. Draft genome sequences were provided by Dr. B. Kreiswirth and Dr. G. Archer from the NARSA genome project.

Clinical isolate PCR screening

The speG and aliD primer specificity testing was performed in triplicate, using the reference strains FPR3757 (MRSA) as a positive control and strain Newman (MSSA) as a negative control, with subsequent purification and sequencing to confirm the target sequence.

All strains were initially screened by colony PCR. All clinical isolates with negative results for either gene product were confirmed at least in duplicate. In cases where there was discrepancy between presence and absence of ACME genes (i.e., at least one gene positive by PCR and any of the remaining 3 genes PCR negative) we did further confirmation using purified genomic DNA.

Assays were performed in a total volume of 20 μl (1 μl of bacterial colony diluted in H2O, 19 μl PCR mixture containing, 0.125 μM of the forward and reverse primers, and 4 μl Taq 5X Master Mix (New England Biolabs)). The cycling protocol was: 5 min at 95°C, followed by 35 cycles of 15 s at 95°C, 30 s at 56°C and 15 s at 72°C. The PCR products (10 μl) were resolved in a 1% Tris-acetic acid-EDTA (TAE) agarose gel in 1✕ TAE buffer at 112 mV/cm for 1 hour and PCR products were visualized with ethidium bromide.

Genomic preparations

After growth over night, 2 ml of bacterial culture was centrifuged at 14,000 rpm for 2 minutes. The bacterial pellet was resuspended in cell wall lysis mixture (6 g/ml lysostaphin, 2.7 g/ml mutanolysin, 16.7 g/ml lysozyme, in 50mM Tris-HCL-10mM EDTA pH8) for 2 hours at 37°C. DNA was then extracted using column-based genomic DNA isolation (DNeasy kit, Qiagen), and PCR was performed again.

Broth MIC determination

For broth culture MIC experiments of antibiotics and polyamines, strains were grown overnight in and then to OD600 of 1.0. Then, strains were inoculated 1:50 in TSB 0.4% glucose at the appropriate concentrations of spermidine, spermine, and putrescine. Aliquots (200 ml) of each condition were arrayed in flat bottom 96 well plates, grown overnight, and then read in TECAN I-Control plate reader at OD600.

Crystal violet biomass assay

Strains were grown overnight, re-inoculated at 1:150 and grown to an OD600 of 1.0. After washing cultures in PBS each strain was inoculated at 1:50 of culture:TSB at the appropriate concentrations of spermidine, spermine, putrescine, N-acetylspermidine (Sigma Aldrich) and NaOH for pH adjustment. All biofilm media contained 0.4% glucose added after autoclaving. Aliquots of 200 mL were grown overnight, statically, at 37°C in a 96-well flat bottom polystyrene plate. Then, each well was drained and washed twice with normal saline. To fix the remaining biofilm, 100% methanol was added to each well and allowed to incubate at room temperature for 15 minutes. Plates were stained with 100 mL of 2% crystal violet for five minutes. Excess stain was rinsed off by washing the plates vigorously in a water bucket. After washing, dye bound to adherent cells was re-solubilized by adding 200 mL of 33% glacial acetic acid for 15 minutes. 100 mL of re-solubilized mixture was transferred to a new flat bottom 96 well plate and a TECAN I-Control plate reader read the OD at 570nm.

Quantitative Real-Time PCR (qRT-PCR)

After exposure to spermidine, samples were then centrifuged at 14,000 rpm for 5 min, and the bacterial pellets were resuspended in 100 mL of RNAlater (Ambion). Total RNA was extracted using Ribopure bacterial RNA extraction kit (Ambion). RNA was reverse transcribed to cDNA using the high capacity cDNA reverse transcription kit (Applied Biosystems). qRT-PCR was carried out using Power SYBR-Green Master Mix in a StepOne Plus thermal cycler (Applied Biosystems). See the list below for primers. Relative quantification (RQ) values were calculated using a comparative threshold cycle (ΔΔCr) program using the software StepOne 2.0.

Primers used in this study

arcAF / GAGCCAGAAGTACGCGAG
arcAR / CTAACACTGAACCCCAATG
opp3ABR / GCAATCTGTAAATGGTCTGTTC
opp3ABF / GAAGATTGGCAGCACAAAGTG
speGF
speGR
UaliDF
UaliDR
aliDF
aliDR / TCTGTTTTAAATCCTTGTGATCG
GCCTTATGAATCCTTAACGGAAC
GTGTAAAATAACAGATAAGAATAGAC
GACAAACGATCAAATCCCTGA
TTGATAATCCTAGCGAACACGA
GACAAACGATCAAATCCCTGA
16S_F / GCG CTG CAT TAG CTA GTT GGT
16S_R / TGG CCG ATC ACC CTC TCA
agrC_F / CCA GCT ATA ATTAGT GGT ATT AAG TAC AGT AAA CT
agrC_R / AGG ACG CGC TAT CAA ACA TTT T
ahpC_F / GCA TGA CCA TTC AGA TGC AA
ahpC_R / CCA ATT CCG TCA GCG TTA AT
atl F / TTT GGT TTC CAG AGC CAG AC
atl R / TTG GGT TAA AGA AGG CGA TG
cidA_F / AAT TTC GGA AGC AAC ATC CA
cidA_R / CTT CCC TTA GCC GGC AGT AT
clfA_F / TTT CAA CAA CGC AAG ATA
clfA_R / GCT ACT GCC GCT AAA CTA
clfB_F / TTT GGG ATA GGC AAT CAT CA
clfB_R / TCA TTT GTT GAA GCT GGC TC
clpC_F / GAA AAA TTA ACG GGC GGA TT
clpC_R / GCA ATC TCT CCG AGT GG
fnbA_F / CCA GGT GGT GGT CAG GTT AC
fnbA_R / TGT GCT TGA CCA TGC TCT TC
icaA_F / CGC ACT CAA AGG CAT T
icaA_R / CCA GCA AGT GTC TGA CTT CG
icaR_F / CCA AAT TTT TGC GAA AAG GA
icaR_R / TAC GCC TGA GGA ATT TTC TG
rbf_F / ACG CGT TGC CAA GAT GGC ATA GTC TT
rbf_R / AGC CTA ATT CCG CAA ACC AAT CGC TA
rot_F / TCG CTT TCA ATC TCG CTG AA
rot_R / CGA CAC TGT ATT TGG AAT TTT GCA
saeS_F / AAT CCA GAA CCA CCC GTT TT
saeS_R / ACG CCA CTT GAG CGT ATT TT
sarA_F / GCA CAA CAA CGT AAA AAA ATC GAA
sarA_R / TTC GTT GTT TGC TTC AGT GAT TC
sarX_F / GTC CTA CTT AAA TCT AGC TCA TCC
sarX_R / CTG AGA AAT TAG AAA CAT TGC TTG GC
sod_F / CCA ATG TAG TCA GGG CGT TT
sod_R / GTT CAG GTT GGG CTT GGT TA

Supplementary Results

Co-occurrence of the genes of USA300 ACME in S. aureus

To test the hypothesis that the entire ACME locus was transferred en bloc from S. epidermidis along with the speG gene, we surveyed a collection of 192 clinical and environmental samples collected in New York City between January 2009 and May 2010 10. We predicted that the entire ACME locus would be intact in most strains if it had been transferred in a single event. Indeed, PCR-based screening showed that speG, aliD, arcA and opp-3a genes were never found in isolation from one another, strongly suggesting that the entire ACME locus was present in every ACME positive strain.

ACME was also strongly associated with USA300 MRSA strains. In this collection of isolates the USA300 clonal complex (spa-CCt008) was the predominant clone accounting for 70% (135/192) of all isolates. Of the 101 strains that were ACME positive only 2 were non-spa-CCt008 strains (t189 [ST188], t692 [ST254]). One of these strains (t692, ST254) is close relative of the ST8 complex 11. ACME was also strongly associated with MRSA (mec type IV) in this lineage with 99% (98/99) of spa-CCt008 MRSA strains being ACME positive and 95% (19/20) of spa-CCt008 MSSA strains being ACME negative. Only 15% of spa-CCt008 MRSA strains were ACME negative. Of the two non-spa-CCt008 strains that were ACME positive both were also MRSA (mec type IV) (Table 1).

To further search for independent occurrences of USA300-type ACME genes in S. aureus we surveyed a set of genome sequences from 125 diverse isolates using a BLAST-based approach. These genomes included 75 publically available whole and draft genome sequences as well as 50 diverse draft sequences of genomes drawn from the NARSA collection (http://www.narsa.net). Together, these genomes represent the largest database of S. aureus genomic diversity available. For arcA, while multiple paralogs and orthologs existed in many genomes, the top matches to USA300 ACME arcA were always S. epidermidis sequences (e.g., 99% nucleotide identity to ATCC12228). Likewise, for aliD the top hits outside of known USA300 strains were exclusively S. epidermidis (e.g., 98% nucleotide identity to RP62A) and other coagulase negative staphylococci. Only one strain of S. aureus, Wood 46 NRS105, had a speG gene with strong nucleotide identity (98%) to the speG gene from USA300 strains. It is notable that the arcA homolog found in this strain was significantly different from USA300 ACME arcA gene with 79% amino acid identity and 78% nucleotide identity. The best hit for aliD in the Wood 46 strain had a similarity score of 23% at the amino acid level and no significant nucleotide similarity.

Thus, in our survey of 317 S. aureus strains we found only one that had a speG gene independent of other ACME-type genes. In contrast, a survey of S. epidermidis genomes uncovered multiple isolated occurrences of ACME-like genes (Fig. 1b).

Estimation of divergence times and evolutionary rates

Both clock models yielded virtually identical estimates for the age of the S. aureus USA300 clade and the node clustering the former and two S. epidermidis isolates (VCU050 and NIHLM049). The evolutionary rate estimated by the two models was also virtually identical. Although the relaxed clock model had a slightly better marginal likelihood score than the strict clock model, the Bayes factors comparison showed that there was not sufficient evidence against a strict clock model (BF<3), therefore we chose to report the strict clock results (Table S3).

Supplementary Table 1: 192 Clinical Strains.

ACME + / ACME –
USA300 (spa type 008)
MRSA / 98 / 17
MSSA / 1 / 19
Total / 99 / 36
Non USA300 (not spa type 008)
MRSA / 2 / 39
MSSA / 0 / 16
Total / 2 / 55

Supplementary Table 2: Information on the 14 isolates used in the time divergence and rate estimation of ACME elements.

Isolate / Date of isolation / Institution
S. aureus
FPR3757 / 14 Mar 2003 / UCSF
TCH1516 / 13 Sep 2002 / BCM
CO-23 / Nov 2007 to Mar 2009 / JCVI
CGS01 / 10 Jan 2005 / Drexel
IS24 / Nov 2007 to Mar 2009 / JCVI
IS91 / Nov 2007 to Mar 2009 / JCVI
IS111 / Nov 2007 to Mar 2009 / JCVI
S. epidermidis
NIHLM049 / 22 Jul 2008 / NHGRI
NIH05001 / 14 Jan 2005 / NHGRI
NIH05005 / 12 Dec 2005 / NHGRI
VCU013 / Sep 1981 / VCU
VCU014 / 13 Mar 2001 / VCU
VCU050 / 19 Oct 2001 / VCU
VCU120 / 28 Jun 2003 / VCU

Supplementary Table 3: Molecular clock models.