Many sequenced marine α-Proteobacterial isolates have been observed to contain identifiable GTA-like gene clusters 
, (). Several of these sequenced, GTA containing α-Proteobacterial strains were acquired for this study (). To date, four marine strains have been documented to spontaneously produce GTA-like particles with no precipitous decline in host cell abundance after entering stationary phase growth 
. For this study seven marine α-Proteobacterial strains were screened for the spontaneous production of virus-like particles during stationary phase (). Of these seven strains Oceanicola granulosus
, Roseovarius nubinhibens
ISM, Ruegeria mobilis
strain 45A6 and Nitratireductor
strain 44B9s were observed to produce particles in this way and the particles were considered potential GTAs. The day of maximal production of GTAs was different for each strain, but reproducible within that strain 
. Maximum levels of GTAs produced for the strains examined are listed in . For the determination of GTA production as a function of cell abundance experiments were performed with paired GTA and cell counts. At the day of maximal GTA production for the R. nubinhibens
and the R. mobilis
45A6 strains, GTA particles and cells were enumerated. In these experiments the GTA counts were slightly lower than the observed maxima. For the R. nubinhibens
ISM strain the GTA particles were measured at 3.5×108
and the cell counts were 2.8×109
with GTA particles equating to 10.9% of cell abundance. For the R. mobilis
45A6 strain the GTA abundance was 5.1×108
and the cell counts were 1.4×109
with GTA particles equating to 30% of the cell abundance.
GTA Gene Maps: GTA gene maps from representative, fully sequenced strains used in this study.
Sequenced Alpha-Proteobacterial strains containing GTA gene cassettes and screened for GTA production over time.
The strains Roseovarius nubinhibens
ISM, and Nitratireductor
strain 44B9s were mutagenized with Tn5 
, which has a known sequence and encodes genes for both Kanamycin and Streptomycin resistance. Since GTAs function by packaging random pieces of host DNA the strains mutagenized with Tn5 would produce GTA particles with some proportion containing Tn5 sequences including the genes for antibiotic resistance. These mutagenized strains were subsequently used as donor strains in production of GTAs for environmental gene transfer experiments.
Sampling of natural microbial assemblages was undertaken in several differing natural environments and at varying times of year to determine if GTA-mediated gene transfer could be observed in microbial populations from diverse environments. Samples from a coastal environment in Georgia were obtained in October (). For Site 1 the ambient viral abundance was 2.2×107 viruses ml−1 and the bacterial abundance was 1.2×106 cells ml−1, yielding a VBR of 18.3. The calculated GTA MOI was 332 GTAs cell−1 At Site 2 the measured viral and bacterial abundances were similar at 2.05×107 viruses ml−1 and 1.6×106 cells ml−1, respectively. For this site the VBR was 12.8 and the calculated GTA MOI was 254 GTAs cell−1. Interestingly, both of these sites had a high percentage of cultivable bacteria with a calculated 60% of the bacterial abundance at Site 1 forming colonies and 25% at Site 2, however this calculation was based on plate counts after the experimental overnight incubation and some growth of the microbes in the samples is likely. At both of these sites a statistically significant increase in both single (Kanamycin) and double (Kanamycin and Streptomycin) antibiotic resistance was observed in the GTA treated samples in comparison to controls. Total viable counts were statistically indistinguishable between treatments and controls ().
RnGTA-mediated gene transfer experiments at two sites off Jekyll Island, GA: The background is a satellite map of the island with the arrows indicating the sampling sites.
All of the spontaneous double antibiotic resistant colonies from the controls were screened by PCR for the presence of the Tn5 Streptomycin kinase gene as well as several double antibiotic resistant isolates from the replicate GTA treated samples. The gene was not recovered from any of the spontaneous revertants but the exact match to the Tn5 gene was recovered from 11% of the screened treatment colonies. The high rate of double antibiotic resistance is extremely improbable by chance; nonetheless, only 11% of the viable colonies produced the expected Streptomycin kinase gene. It is possible that even though the active sites of the genes were successfully transferred, some modifications, truncations, or re-arrangements may have occurred. This type of gene alteration has been documented in the past for genes transferred to marine bacterial populations by natural transformation 
Because of the common association between the coral endosymbiont Symbiodinium
and α-Proteobacteria 
, GTA-mediated gene transfer was examined in a reef environment. The samples with their associated microbes from the vicinity of the Looe Key reef were acquired in September. Sampling was performed immediately adjacent to the reef surface. In this sample the ambient bacterial abundance was 1.16×106
and the viral abundance was 2.34×107
measured by flow cytometry, yielding a virus to bacteria ratio of 20.2. The calculated GTA MOI was 150 GTAs cell−1
The water samples were incubated with GTAs produced by both the R. nubinhibens: Tn5 (RnGTA) and the Nitratireductor 44B9s: Tn5 (NrGTA) donor strains. A statistically significant increase in antibiotic resistance was observed in the treated samples for both GTAs and for both single and double antibiotic selection (). In this experiment parallel samples tested for prophage induction also contained inducible lysogens in the ambient microbial population as evidenced by a statistically significant increase in virus abundance in the Mitomycin C treated samples (). The isolates from these experiments were not screened for Tn5 sequences.
GTA-mediated gene transfer in the reef environment: RnGTA and NrGTA mediated gene transfer frequencies at a site near Looe Key Reef.
Environmental parameters measured in conjunction with cruise GTA-mediated gene transfer and prophage induction experiments.
To investigate further, a short research cruise was undertaken with sampling of a series of four stations in the vicinity of the Gulf of Mexico. The stations were selected to represent a gradient of conditions over a short temporal scale including oligotrophic, oligotrophic with upwelling, near shore and estuarine river mouth stations (). In this case, the samples were obtained during late winter (February). The gradient in conditions can be observed in the decreasing salinity and increasing viral and bacterial abundances moving from the oligotrophic areas toward the more eutrophic river mouth inside Tampa Bay ().
Similarly to the reef experiment, samples were incubated with GTA particles produced using both donor strains. In addition, both high and low MOI treatments for each GTA type were included. At the oligotrophic station 4, no spontaneous antibiotic revertants were cultured and the GTA treated samples did produce some colonies demonstrating resistance to Kanamycin (, Station 4). However, statistical significance of the result could not be confirmed due to high variability in the data. At the oligotrophic station with upwelling (, Station 5), GTA-mediated transfer was observed with the RnGTAs at a high MOI and with the NrGTAs at both the low and high MOIs. At the two near-shore stations (Stations 2 and 7), gene transfer was even more prevalent, primarily due to easier detection because of the higher numbers of cultivable bacteria. Note the higher ratios of cultivable bacteria in comparison to total bacterial counts at stations 2 and 7 as compared to the oligotrophic stations (). Higher levels of gene transfer were also observed with the high MOI treatments in comparison to the similar low MOI treatments at all of the stations sampled. Several isolates (a total of 67) from the spontaneous antibiotic resistant controls and the two GTA treatments were screened by PCR for the Tn5 Streptomycin kinase gene. No amplicons were recovered from the spontaneous revertants. However, the correct amplicon was recovered from 12% of the RnGTA treated isolates and 17.6% of the NrGTA treated isolates.
GTA-mediated gene transfer in the Gulf of Mexico: RnGTA and NrGTA low and high multiplicity of infection (MOI) experiments from the Gulf of Mexico/Tampa Bay cruise.
In the all of the February cruise samples parallel prophage induction experiments were performed. In these samples both the bacterial and viral abundances were measured in the control and Mitomycin C treatments. Although a decrease in bacterial abundances was observed at stations 4, 2 and 7 in response to the Mitomycin C treatment, in no case was a corresponding statistically significant increase in viral abundance observed. Thus, prophage induction did not occur at any of the sampled stations ().
Statistical comparison of the GTA-mediated gene transfer frequency and measured environmental parameters for fourteen separate experiments in which GTA-mediated gene transfer was detected, including those discussed in detail above. The measured salinity ranged from 27 to 39, the temperature ranged from 16–30.5°C and the virus to bacteria ratio ranged from 13–51. The average ambient bacterial abundance from all experiments was 1.6×106
), and the average ambient viral abundance was 4.4×107
). Multiple correlation analysis indicated one statistically significant correlation with the GTA-mediated gene transfer frequency, which was an inverse relationship between gene transfer and temperature (r
14). This result suggested that more gene transfer occurred in cooler conditions. Modeling of the GTA-mediated gene transfer frequency by stepwise regression indicated that the environmental parameters most explanatory of increasing gene transfer frequency were the combined parameters of higher salinity, increasing GTA dose (MOI) and lower ambient bacterial abundance with a combined P value
0.019 and an adjusted R2
of 0.992 for this model. Although temperature was correlated with gene transfer frequency, it was not a strong explanatory factor in the best-fit model.