SCIENTIFIC REPORTS | (2020) 10:15091
|
www.nature.com/scientificreports
Horizontal ‘gene drives’
harness indigenous bacteria
for bioremediation
Katherine E. French
1*
, Zhongrui Zhou
2
& Norman Terry
1
Engineering bacteria to clean-up oil spills is rapidly advancing but faces regulatory hurdles and
environmental concerns. Here, we develop a new technology to harness indigenous soil microbial
communities for bioremediation by ooding local populations with catabolic genes for petroleum
hydrocarbon degradation. Overexpressing three enzymes (almA, xylE, p450cam) in Escherichia coli
led to degradation of 60–99% of target hydrocarbon substrates. Mating experiments, uorescence
microscopy and TEM revealed indigenous bacteria could obtain these vectors from E. coli through
several mechanisms of horizontal gene transfer (HGT), including conjugation and cytoplasmic
exchange through nanotubes. Inoculating petroleum-polluted sediments with E. coli carrying the
vector pSF-OXB15-p450camfusion showed that the E. coli cells died after ve days but a variety of
bacteria received and carried the vector for over 60 days after inoculation. Within 60 days, the total
petroleum hydrocarbon content of the polluted soil was reduced by 46%. Pilot experiments show that
vectors only persist in indigenous populations when under selection pressure, disappearing when this
carbon source is removed. This approach to remediation could prime indigenous bacteria for degrading
pollutants while providing minimal ecosystem disturbance.
Oil spills in recent decades have le a long-term mark on the environment, ecosystem functioning, and human
health
1–3
. In the Niger Delta alone, the roughly 12,000 spills since the 1970s have le wells contaminated with ben-
zene levels 1,000× greater than the safe limit established by the World Health organization and have irreparably
damaged native mangrove ecosystems
4,5
. Continued economic reliance on crude oil and legislation supporting
the oil industry mean that the threat of spills is unlikely to go away in the near future
6
.
At present, there are few solutions to cleaning up oil spills. Current approaches to removing crude oil from
the environment include chemical oxidation, soil removal, soil capping, incineration, and oil skimming (in
marine contexts)
7,8
. While potentially a ‘quick x,’ none of these solutions are ideal. Soil removal can be costly
and simply moves toxic waste from one site to another
9
. Chemical oxidants can alter soil microbial community
composition and pollute groundwater
10
. Incineration can increase the level of pollutants and carbon dioxide in
the air and adversely aect human health
11
. Practices such as skimming only remove the surface fraction of the
oil while the water-soluble portion cannot be recovered, negatively aecting marine ecosystems
12,13
.
Synthetic biology has now given us the tools to tackle grand environmental challenges like industrial pollu-
tion and could usher in a new era of ecological engineering based on the coupling of synthetic organisms with
natural ecosystem processes
14–18
. Consequently, using bacteria specially engineered to degrade petroleum could
present a viable solution to cleaning up oil spills in the near future. Previous studies have identied which bacte-
rial enzymes are involved in petroleum hydrocarbon degradation (reviewed in references
9,19
) and have engineered
bacterial enzymes like p450cam for optimal invivo and invitro degradation of single-substrate hydrocarbons
under lab conditions
20,21
. However, there are several critical gaps in or knowledge of engineering bacteria for
oil-spill bioremediation. First, we know little about how the performance of these enzymes compare and which
enzyme would present an ideal target for over-expression in engineered organisms. Second, it is unclear how
well engineered organisms can degrade petroleum hydrocarbons compared to native wild-type bacteria which
naturally degrade alkanes, such as Pseudomonas putida. ird, the environmental eects of engineered bacteria
on native soil populations are unclear. For example, do these bacteria persist over time in contaminated soils?
OPEN
USA.
*
email: katherine.french@
berkeley.edu
Vol:.(1234567890)
SCIENTIFIC REPORTS
| (2020) 10:15091 |
www.nature.com/scientificreports/
Although the use of genetically modied bacteria in bioremediation is attractive, this solution faces signicant
regulatory hurdles which prohibit the release of genetically modied organisms in the environment
22
.
Here, we propose a new bioremediation strategy which combines synthetic biology and microbial ecology and
harnesses natural processes of horizontal gene transfer in soil ecosystems. We screened ve enzymes involved
in petroleum degradation in E. coli DH5α (alkB, almA, xylE, ndo and p450cam) to identify (1) where these
enzymes localize and their eect on crude oil using advanced microscopy and (2) to asses each enzyme’s ability
to degrade three petroleum hydrocarbon substrates (crude oil, dodecane, and benzo(a)pyrene) compared to two
wild type bacteria (Pseudomonas putida and Cupriavidus sp. OPK) using bioassays and SPME GC/MS. Based on
these results, we selected one vector (pSF-OXB15-p450camfusion) to determine whether small, synthetic vectors
carrying catabolic genes could be transferred to indigenous bacteria found in petroleum-polluted sediments and
whether this shi in community metabolism could increase rates of pollutant degradation.
Results and discussion
Overexpression of petroleum hydrocarbon-degrading enzymes in E. coli. To compare the locali-
zation and activity of known petroleum hydrocarbon-degrading enzymes, we inserted ve enzymes (alkB, almA,
xylE, ndo, and p450cam) and required electron donors into the vector backbone pSF-OXB15 using Gibson
Assembly
23
(SI Fig.1). To identify where each enzyme localized within E. coli DH5α, we tagged each enzyme
with a uorophore (gfp or mcherry). Fluorescence microscopy revealed that alkB was localized to bacterial
cell membranes and almA was found throughout the cytoplasm. e camphor-5-monooxygenase camC from
the p450cam operon was expressed throughout the cell cytoplasm while another enzyme in the operon, the
5-exo-hydroxycamphor dehydrogenase camD, was expressed within a small compartment at one end of the cell
(Fig.1A). e dioxygenase ndoC from the ndo operon was also localized to a small compartment at one end of
the cell. e dioxygenase xylE was found in small amounts in the bacterial cell membrane and larger amounts in
a microcompartment at one end of the cell. In all cases, these compartments were 115–130nm wide and could
be seen in young, mature and dividing cells. e presence of microcompartments in E. coli expressing p450cam,
ndo, and xylE could reect the known use of protein-based microcompartments by bacteria to concentrate
highly reactive metabolic processes
24
.
Over-expression of all ve enzymes imbued E. coli DH5α with metabolism-dependent chemotactic behavior,
where cell movement is driven towards substrates aecting cellular energy levels
25
. E. coli DH5α do not have
agella, but rather, moved towards petroleum hydrocarbon substrates via twitching
26–28
. Fluorescence microscopy
showed E. coli DH5α expressing alkB ‘clinging’ to oil droplets (Fig.1B) and those expressing xylE seemed to use
the compartment-bound enzyme as a ‘guide’ towards crude oil (SI Fig.2). Both behaviors mimic the interactions
of wild-type oil-degrading bacteria
7
.
Fluorescence microscopy also revealed for the rst time the key role of extracellular enzymes in degradation of
petroleum hydrocarbons. ree enzymes, alkB, almA, and p450cam were found in extracellular vesicles ranging
in size from 0.68μm to 1.67μm (SI Figs.3 and 4). ese vesicles were only seen when E. coli DH5α was exposed
to petroleum hydrocarbons. ey are larger than minicells (which range from 200–400nm in diameter)
29
and
seem to serve some other function. Confocal images suggest that these vesicles may come into contact with oil
droplets, potentially attaching to (or merging with) their surface (SI Fig.4).
We also found three enzymes, alkB, xylE, and p450cam, within the E. coli exopolysaccharide (EPS) matrix.
AlkB and xylE were concentrated around the 500nm pores within the EPS and found dispersed in smaller
amounts throughout the exopolysaccharide (Fig.1C; SI Fig.5). In contrast, p450cam was distributed in high
amounts throughout the EPS. Cryotome sectioning of the EPS from bacteria expressing p450cam indicates that
the monooxygenase camA from the p450cam operon co-localized with a second enzyme involved in hydro-
carbon degradation, the dehydrogenase camD (SI Fig.6). Protein levels of EPS varied signicantly among the
dierent strains of wild-type and engineered bacteria (ANOVA test, F
7,16
= 11.3, p < 0.001, adjusted R
2
= 0.76).
Figure1. Expression and localization of bacterial monooxygenases and dioxygenases involved in petroleum
degradation in E. coli DH5α. (A) Structured Illumination Microscopy (SIM) image of E. coli DH5α expressing
proteins involved in petroleum degradation (cam A, B, C and D) from the CAM plasmid in E. coli. camC(fused
to mcherry) was found throughout the cell while camD(fused to gfp) localized to a microcompartment at one
end of the cell. e scale bar is 5μm. (B) E. coli DH5α expressing alkB fuse to gfp were found clinging to spheres
containing crude oil, mimicking a behavior seen in wild-type oil-degrading bacteria. (C) EPS from E. coli DH5α
expressing xylE. Gfp-tagged xylE were found around small pores (ca. 500nm) within the EPS matrix. (B) and
(C) were taken using the GFP lter on a Zeiss AxioImager M1.
Vol.:(0123456789)
SCIENTIFIC REPORTS
| (2020) 10:15091 |
www.nature.com/scientificreports/
Bacteria expressing alkB (0.84 ± 0.04mg/ml), xylE (0.85 ± 0.13mg/ml), and p450cam (0.97 ± 0.06mg/ml) had
higher levels of EPS protein than E. coli DH5α expressing the empty vector pSF-OXB15 (0.44 ± 0.02mg/ml) (SI
Fig.77 and were comparable to the EPS protein levels of Cupriavidus sp. OPK (1.2 ± 0.15mg/ml), a bacteria
known to use biolms to degrade crude oil
7
. Although previous studies suggest that EPS may be involved in the
extra-cellular metabolism of environmental pollutants
7,30–34
, this is the rst study to identify several enzymes
which may play a role in this process.
Finally, extracellular enzyme expression also inuenced the size of the oil droplets within the cell culture
media. For example, E. coli DH5α expressing xylE produced very small oil droplets (primarily > 1μm in diam-
eter) of crude oil while those expressing alkB and almA produced droplets ranging in size from 1μm-120μm in
diameter (SI Fig.8). Although xylE was not seen in vesicles, confocal suggests that crude oil droplets can ow
through pores in biolms and may become coated in EPS in the process. Potentially, attachment of enzymes to
oil droplets (through fusion with vesicles or contact with EPS) may inuence how fast a droplet is degraded over
time. Previous studies have shown that vesicles embedded with enzymes can catalyze chemical reactions
35,36
. In
addition, Dmitriev etal. have shown that two bacteria, P. putida BS3701 and Rhodococcus sp. S67, use vesicular
structures containing oxidative enzymes which attach to and play a role in degradation of crude oil droplets
37
.
Our results thus suggest that bacterial monooxygenases and dioxygenases involved in petroleum hydrocarbon
degradation may be involved in multiple, complex inter and intra-cellular processes that lead to the degradation
of crude oil.
Comparison of enzyme activity. To determine which enzymes were most useful for degrading long-
chain hydrocarbons, polyaromatic hydrocarbons (PAHs), and crude oil, we conducted 96-well plate assays
exposing wild-type and genetically engineered bacteria to 1% (v/v) of dodecane, benzo(a)pyrene or crude oil.
We found that bacteria engineered to over-express specic enzymes in petroleum degradation were able to
degrade single-carbon substrates better than the wild-type bacteria P. putida and Cupriavidus sp. OPK. One
way analysis of variance (ANOVA) of the assay data showed that there was signicant variation in bacterial
growth when exposed to dodecane (F
8,31
= 33.4, p = < 0.001), benzo(a) pyrene (F
8,31
= 73.03, p = < 0.001) and
crude oil (F
8,31
= 240.6, p = < 0.001). When exposed to dodecane, E. coli DH5α expressing p450cam increased
in biomass the most (139.4%), followed by E. coli DH5α expressing xylE (136.3%), alkB (120.8%), and almA
(97.6%) (Fig.2A). Expressing p450cam and xylE led to signicantly greater conversion of dodecane to biomass
compared to P. putida (t = 4.71, df = 3.17, p < 0.01 and t = 4.41, df = 3.17, p < 0.01 respectively). Solid-Phase Micro-
Extraction (SPME) GC/MS analysis of these cultures revealed that all three bacteria degraded 99% of dodecane
in 10days. When exposed to benzo(a)pyrene, P. putida had the greatest increase in biomass (119.2%) followed
by E. coli DH5α expressing almA (117.1%), xylE (94.8%), and p450cam (90.8%) (Fig.2B). T-tests showed there
was no signicant dierence in the biomass of P. putida and these three strains (p > 0.10). SPME GC/MS showed
that E. coli expressing P450cam, almA and xylE degraded 90%, 97% and 98% of the benzo(a)pyrene respectively
while P. putida degraded 86%.
In contrast, when engineered and wild-type bacteria were exposed to crude oil, P. putida converted the oil to
biomass more eciently, increasing in biomass by 110.9% (Fig.2C). Only two genetically engineered bacteria,
E. coli DH5α expressing p450cam and almA, had comparable increases in biomass to Cupriavidus sp. OPK
(61.93%, 52%, and 48.7% respectively). e assay was repeated with crude oil stained with Nile Red and rates of
degradation were determined according to French and Terry
7
. P. putida degraded 79% of crude oil while E. coli
Figure2. Growth of wild-type and engineered bacteria on dodecane (A), benzo(a)pyrene (B), and crude oil
(C). Wild type strains are denoted as ‘Cupriavidus’ and ‘P. putida.’ Synthetic strains are denoted according to
what enzyme they are engineered to express (e.g. alkB, almA). e two negative controls are a control with the
carbon substrate but no cells and E. coli DH5α transformed with the vector backbone used in the experiment
(pSFOXB15) but without genes inserted for hydrocarbon degradation. Optical density measurements were
taken at OD 600nm. Scale bars show standard error of the mean.
Vol:.(1234567890)
SCIENTIFIC REPORTS
| (2020) 10:15091 |
www.nature.com/scientificreports/
DH5α expressing p450cam and almA degraded 64% and 60% respectively. E. coli expressing alkB, xylE, and ndo
only grew ~ 25% and degraded 35–40% of crude oil. e high performance of p450cam when exposed to crude
oil likely reects the enzyme’s known substrate promiscuity
38,39
which makes it a better catalyst for degrading
crude oil, a complex substrate made of over 1,000 compounds
40
.
Vector-exchange between E. coli DH5α and indigenous bacteria. To determine whether our
engineered bacteria could transfer non-conjugative, synthetic vectors containing petroleum-degrading genes
to indigenous soil and marine bacteria, we conducted a series of mating experiments. We found that wild-type
bacteria readily received the vector pSF-OXB15-p450camfusion through horizontal gene transfer (HGT) (Fig.3;
SI Fig.9). Frequencies of transformation ranged from 19 to 84% in 48h depending on the recipient species (SI
Table1) and were > 90% aer seven days of incubation for all tested species. Plasmid expression was stable for
over three months in the absence of antibiotic pressure. Although our vectors carried a ColE1 origin of replica-
tion, this did not seem to present a barrier to HGT. is agrees with previous studies which suggest ColE1 plas-
mids can be found in wild bacteria
41
and wild-type bacteria can receive ColE1 plasmids from E. coli
42
.
HGT can occur through transformation, transduction, conjugation, transposable elements, and the fusing
of outer membrane vesicles (OMVs) from one species to another
43,44
. To test whether wild-type bacteria could
take up naked plasmids from cell culture, we adding 1μl of puried plasmid (at a concentration of 1ng/μl and
10ng/μl) to LB cultures containing wild-type bacteria and by spreading diluted vectors onto agar plates. We saw
no transformed cells. In addition, neither uorescence microscopy or TEM showed OMV production or release
by the E. coli DH5α strains created in this study. OMVs are 50–250nm in diameter
45
, much smaller than any of
the vesicles produced by our strains.
To determine whether HGT of the synthetic vectors to wild-type bacteria was achieved through mating,
we conducted TEM of wild-type bacteria aer exposure to E. coli DH5α carrying pSF-OXB15-p450camfusion.
TEM suggests several mechanisms of HGT through direct cell-to-cell contact may explain how vectors were
transferred between transgenic E. coli DH5α and wild-type bacteria
46
. In our study, we found E. coli DH5α and
wild-type bacteria engaging in DNA transfer through conjugation and cell merging and in cytoplasmic transfer
via nanotube networks. TEM showed E. coli DH5α expressing p450cam tethered to wild-type cells by conjuga-
tive pili over long distances (Fig.4A), the formation of mating pair bridges between wild-type cells and E. coli
DH5α (Fig.4B), E. coli with conjugative pili (Fig.4C), and E. coli and wild-type cells connected via nanotubes
(Fig.4D) (see also SI Fig.10). Although the plasmids used in this study were non-conjugative, such plasmids
can be mobilized and transmitted via conjugation in the presence of other conjugative plasmids or by merging
with conjugative plasmids
47–49
. Previous studies have also shown that plasmids (and chromosomal DNA) can be
transferred through cell-contact dependent transfer without the use of conjugative pilii. is mechanism was rst
observed in 1968 in Bacillus subtilis and subsequently in other species (e.g. Vibrio, Pseudomonas, Escherichia)
(reviewed in
49
). For example, Paul etal.
50
found that lab strains of E. coli could transfer plasmid DNA to Vibrio
through this process. Nonconjugal plasmids can also be transferred from between bacteria through nanotubes
51
.
Dubey and Ben-Yehuda
52
show in their classic paper that gfp molecules, calcein, and plasmids could be trans-
ferred between B. subtilis cells. ey also show that a non-integrative vector carrying a resistance marker from
B. subtilis could be transferred to Staphylococcus aureus and E. coli (with recipient cells expressing antibiotic
resistance). is transfer was rapid and could happen in as little as 30min. In our study, it is impossible to say
denitively by which mechanism our vectors were transferred from E. coli DH5α to the wild-type bacteria and
in reality multiple mechanisms of transfer may be possible.
We conducted an additional experiment to determine the survival rate of engineered bacteria in petroleum
polluted sediment from a former Shell Oil renery in Bay Point, CA and whether these genes could be transferred
to native, complex soil microbial communities. is sediment is contaminated with high levels of petroleum
hydrocarbons, arsenic, heavy metals, and carbon black. At D
0
, E. coli DH5α containing the plasmid pSF-OXB15-
p450camfusion were seen in aliquots of contaminated sediment and there were no autouorescent bacteria
Figure3. Expression of pSF-OXB15-p450camfusion in the marine bacteria Planococcus citreus. (A) DIC image
of wild-type P. citreus (no exposure to E. coli). ese bacteria are coccoid-shaped with cells found individually or
in groups of 1–4. (B) Image of P. citreus and E. coli expressing p450cam aer 48h of co-cultivation (described in
methods). (B) was taken using the Texas Red lter on a Zeiss AxioImager M1 (excitation/emission 561/615).
Vol.:(0123456789)
SCIENTIFIC REPORTS
| (2020) 10:15091 |
www.nature.com/scientificreports/
visible in the media. Aer D
5
, the population of E. coli DH5α had declined. Instead, a number of diverse native
soil bacteria now contained the plasmid, a trend which continued over the course of the experiment (Fig.5).
Based on morphological analysis of over 100 microscopy images and pairwise mating between bacteria isolated
and identied (via sequencing) from Shell Pond, these bacteria belonged to the Pseudomonas, Flavobacteria,
and Actinomycete genera among others. Plating out aliquots of soil at regular time points conrmed the data
gathered by microscopy: the number and diversity of bacteria expressing the plasmid increased 50-fold over
the rst 30days of the experiment (from 2.6 × 10
–4
CFU at D
0
to 1.25 × 10
–6
CFU at D
30
) (SI Fig.11). e spider-
silk-like biolms formed by native soil microbiota present in the soil were also uorescent (Fig.5), suggesting
the p450cam enzymes also play a role in extracellular degradation of petroleum hydrocarbons under real-world
conditions. GC/MS analysis showed that the amount of total petroleum hydrocarbons within the contaminated
sediments decreased by 46% within 60days compared to untreated soil samples. We le the experiment running
and aer 120days bacteria carrying the vector were still prolic (SI Fig.12). A pilot experiment using articially
contaminated water samples suggest that genes are transferred from E. coli to indigenous bacteria only when
oil is present. In the control samples without oil, we saw no bacteria carrying the vector over the course of the
30-day experiment (the experiment will be published fully in a later publication).
HGT is thought to play a role in the degradation of environmental toxins
53
and several studies have shown
that wild-type bacteria carrying large plasmids with degradative genes can pass these genes on to a limited num-
ber of bacteria
54
. However, this is the rst study to provide evidence for HGT between E. coli DH5α carrying a
small, non-conjugative vector and wild soil microbiota. Our results show that adding engineered E. coli DH5α
carrying small synthetic plasmids to polluted environmental samples may be even more eective than adding
a wild-type bacteria with a larger catabolic vector. Previous studies show that frequencies of HGT between the
donor and recipient bacteria in soil are low (e.g. 3 × 10
–3
CFU transconjugants per gram of sterile soil), recipient
cells come from only a few genera, and the spread of the catabolic vector through the microbial community
does not always lead to enhanced degradation
55–57
. e high transfer frequency of synthetic plasmids like the
ones used in this study could be due to several reasons, including the small size of the vector (natural plasmids
Figure4. TEM conrmation of horizontal gene transfer among E. coli expressing pSF-OXB15-p450camfusion
and P. putida. (A) E. coli DH5α (E) connected to P. putida (P) via conjugative pili. (B) Mating-pair bridge
between P. putida and E. coli DH5α. (C) E. coli DH5α with conjugative pili. (D) Nanotube connecting P. putida
and E. coli DH5α.