BIOTIC AND ABIOTIC OIL DEGRADATION AFTER THE DEEPWATER HORIZON DISASTER LEADS TO FORMATION OF RECALCITRANT OXYGENATED HYDROCARBONS: NEW INSIGHTS USING GC×GC

Oil that is released into the environment after spills is altered though physical as well as (bio)chemical processes. These processes are often referred to as oil weathering. In a marine oil spill, physical processes consist of dissolution as well as evaporation, while the (bio)chemical processes are microbial degradation (biodegradation) as well as oxidation though sunlight (photooxidation) (Atlas 1981, Garrett et al. 1998, Nicodem et al. 1998, Prince and Walters 2007).

To assess the short- and long-term fate of petroleum hydrocarbons after an oil spill, it is important to identify and assess the relative importance of these weathering processes in the environment. Biodegradation and photooxidation, for example, seem to be very slow for the lingering oil of the 1989 Exxon Valdez oil spill after 16 years (Short et al. 2007). In contrast, for oil washed ashore after the 2010 Deepwater Horizon disaster in the Gulf of Mexico, extensive (bio)chemical weathering of oil was observed within 18 months (Aeppli et al. 2012).

New methods, such as comprehensive two-dimensional gas chromatography (GC×GC) advanced the potential to investigate processes that lead to chemical alternation of oil (Arey et al. 2007a, Arey et al. 2007b). The high chromatographic resolution of GC×GC also has advantages over one-dimensional GC (such as GC-MS) for oil fingerprinting (Frysinger et al. 2002, Eiserbeck et al. 2012).

The aim of this study was to identify physical, chemical, and microbial oil weathering processes in petroleum released after the Deepwater Horizon disaster. To this end, we analyzed a time series of oiled samples, collected along the coast of the northern Gulf of Mexico, using GC×GC and other analytical techniques. Whereas previous papers describe the formation of oxygenated hydrocarbons (OxHC) upon weathering, sequential biodegradation of alkane classes, and chemometric analysis of GC×GC data (Aeppli et al. 2012, Hall et al. 2013, Aeppli et al. 2014, Gros et al. 2014), this paper specifically investigates and summarizes how GC×GC allows for compound-specific degradation assessment of individual petroleum hydrocarbons, in order to gain insights into biotic and abiotic weathering processes.

Over 100 surface slicks, oil-soaked sand samples (“sand patties”) and rock scrapings (oil scraped off rocks) were collected and analyzed (Figure 1). Some of these samples were previously described in Aeppli et al. (2012) and Hall et al. (2013); additional sand patties were collected in 2012. Surface slick samples were directly dissolved in dichloromethane (DCM) and dried with Na₂SO₄. Sand patties and rock scrapings were extracted with DCM / methanol (90/10 v/v) in centrifuge vials (3 extraction, with centrifugation in-between extraction). The sample and solvent amount was adjusted for the final oil concentration to be 10 to 50 mg L⁻¹.

Comprehensive two-dimensional gas chromatography coupled to flame ionization detection (GC×GC-FID) was performed as described in Hall et al. (2013). In brief, 1 μL sample volumes were injected in splitless mode into a GC×GC-FID system (Leco, Saint Joseph, MI), equipped with a Restek Rtx-1 first-dimension column (60 m, 0.25 mm ID, 0.25 μm film thickness) and SGE BPX-50 second-dimension column (1.5 m, 0.10 mm ID, 0.10 μm film). The inlet temperature was held at 300 °C and the carrier gas was hydrogen at a constant flow rate of 1.00 mL min⁻¹. The temperature program of the GC oven was 40 °C for 10 min, then ramped to 340 °C at 1.25 °C min⁻¹ (held 5 min). The second-dimension oven had a constant offset of 5 °C relative to the main oven. A liquid-N₂ cooled two-stage modulator was used (Leco), and the modulation period was either 10 or 15 s.

One-dimensional gas chromatography (GC-FID) was also conducted (Aeppli et al. 2013), using a Hewlett-Packard 5890 Series II GC equipped with an Agilent DB-1MS capillary column (30 m, 0.25 mm I.D., 0.25 μm film) and operated at 5 mL min⁻¹ H₂ carrier gas flow.

Thin-layer chromatography coupled to flame ionization detection (TLC-FID) was performed as described in Aeppli et al. (2012) using an Iatroscan MK-5 TLC–FID analyzer (Iatron Laboratories, Tokyo, Japan). Briefly, 1 to 5 μL of sample extracts were spotted on the base of a silica-gel sintered glass rod (Chromarod S III, Iatron Laboratories) which was then sequentially developed in hexane (26-min development time), toluene (12 min), and DCM/methanol 97/3 (5 min).

To confirm that the investigated samples were MW oil, we used biomarker analysis (petroleum fingerprinting) methods (Wang et al. 2006), with ratios described earlier (Carmichael et al. 2012). GC×GC-FID led to excellent separation of individual petroleum hydrocarbon compounds including biomarkers (Figure 2).

The majority of the oil samples were identified as MW oil. The relative distribution of hopanoids and (dia)steranes of all surface slicks, sand patties, and rock scrapings was the same as for the MW oil (Aeppli et al. 2012). This can be seen by visual comparison (Figure 3), but was also confirmed quantitatively using biomarker ratios. A sub-set of samples, however, did not match the MW signature. These samples were visually different from sand patties or rock scraping, but were sand-free and asphalt-like bricks of 5 to 10 cm diameter, collected mostly along the shores of Grand Isle and Elmer's Island (LA). The sterane-to-hopane ratio and diasteranes-to-sterane ratios were especially different than the MW-type samples (Aeppli et al. 2014).

We found that with increasing weathering, a fraction of non-GC amenable compounds were formed in the weathered oil samples (Figure 4). Elemental, FT-IR, and radiocarbon analysis of this fraction revealed that it consisted of highly oxygenated compounds (average molecular formula was (C₅H₇O)_n with carbonyl and hydroxyl groups), which were formed from petroleum hydrocarbons (Aeppli et al. 2012). We therefore refer to this fraction as oxygenated hydrocarbons (OxHC).

The OxHC fraction was persistent on the time scale of this investigation (up to 28 months post-spill), and increased to up to 80% of the total sample mass. The OxHC fraction also correlated with molecular ratios that are indicative of oil weathering, such as the phytane/17α(H),21β(H)-hopane ratio (Figure 6a). We also linked the formation of OxHC to the disappearance of saturated compounds (Hall et al. 2013). This led us to propose to use the relative amount of the OxHC fraction in a given sample as a proxy for its degree of oil weathering (Aeppli et al. 2012).

Physical oil weathering processes (e.g., evaporation, dissolution) can be easily identified on a GC×GC chromatogram. GC×GC separates compounds according to two different physical properties. The first-dimension column separated according to the vapor pressure of the analytes. Evaporation of compounds in the environment can therefore be identified on the GC×GC chromatogram by removal of compounds with low retention times (i.e, compounds with low n-alkane carbon numbers, Figure 5). Dissolution, on the other hand, can be identified on GC×GC by preferential removal of compounds with higher second-dimension retention times. This is because compounds with higher polarity (i.e., water solubility) are more retained on the second-dimension column.

For the investigated samples we saw a moving evaporation front with time of weathering of the samples in the environment. For slick samples, compounds with carbon numbers < 12 were evaporated relative to crude oil (Figure 5a-b). In contrast to these ~1 to 3 cm thick oil slicks, we observed a more rapid evaporation for thin oil sheen that was collected in 2012 at the Deepwater Horizon site, with compounds with carbon numbers < 16 were evaporated (Aeppli et al. 2013). Evaporation continued once the oil arrived at the beaches (Figure 5c-d).

Beside physical weathering processes, compound-class selective removal of hydrocarbons was observed. This can be seen in a relative disappearance of alkanes and PAHs relative to biomarker compounds (Figure 5).

We investigated to what extent biodegradation, as well as photooxidation, were responsible for this removal of compounds. To this end, we investigated changes in ratios of compounds with the weathering proxy OxHC, and explained this by comparing the physico-chemical properties of the compounds (Figure 6). For example, the decrease of phytane to 17α(H),21β(H)-hopane (Figure 6a) can be driven by the difference in a variety of physical properties as well as in their biodegradability (Table 1). Therefore, this ratio is not suitable to investigate biodegradation; rather it is a general indication of oil weathering.

In contrast, the ratio of phytane to n-octadecane (Figure 6b) is a good indicator of biodegradation, as the major difference in the physico-chemical properties of these two compounds is their susceptibility towards biodegradation (Table 1). For the same reasons, the disappearance of C₃₅-homohopane relative to 17α(H),21β(H)-hopane (Figure 6c) has to be driven by biodegradation (Aeppli et al. 2014).

The concept of comparing two different compounds can also be applied to whole compound classes. By analyzing several alkane compound classes, we found a biodegradation sequence in the order of increasing resistance towards biodegradation of n-alkanes > cycloalkanes ≈ methyl-alkanes > cyclic isoprenoids ≈ isopernoids (Gros et al. 2014).

In contrast to biodegradation of saturated compounds, we observed photooxidation as the main degradation process for aromatic structures. This can be illustrated by the ratio of chrysene and its C3-alkylated congeners, which decreases with increasing weathering of the field samples (Figure 7a, and Aeppli et al. (2012)). Given the physical properties of chrysene and C3-chrysenenes (Table 1), the opposite trend in this ratio would be expected if physical processes were responsible for the shift. Likewise, biodegradation is know to be more effective towards chrysene than towards its alkylated congeners (Prince et al. 2003). In contrast, photooxidation is more efficient with C3-chrysene that chrysene, as alkylation activates the aromatic system. The shift in chysene/C3-chrysene can therefore be explained by photooxidation as the main process leading to depletion of chrysenes.

Note, that lighter PAHs (e.g., naphthalanes, phenanthrenes) are probably also affected by photooxidation. However, these compounds are within the evaporation front (between n-C₁₂ and n-C₁₈; Figures 3a and 5), and are also fairly water-soluble. Consequently, it is challenging to separate the contribution of photooxidation from evaporative and dissolution for these compounds. However, approaches involving numerical modeling would be suitable to do this.

The biomarkers of the triaromatic steroid (TAS) family are also affected by photooxidation. In field samples, we observed a decrease of TAS relative to 17α(H),21β(H)-hopane (Figure 7b-c); interestingly, this degradation affected various TASs, such as C₂₁-TAS and C₂₈-TAS to the same extent (Radović et al. 2013, Aeppli et al. 2014). Given that C₂₁ and C₂₈-TAS have almost three orders of magnitude difference in physical properties (e.g., aqueous solubility, vapor pressure, octanol-water partition coefficient) (Table 1), evaporation and dissolution can be ruled out as the driver of TAS disappearance. Biodegradation can also be excluded as the main driver of this process, as large differences in biodegradability of various TAS have previously been observed in laboratory incubation experiments (Douglas et al. 2012). Given that all TASs have a similarly activated aromatic system, we hypothesize that photooxidation is responsible for the TAS degradation observed for weathered MW oil. To confirm this hypothesis, we performed photooxidation incubations with crude oil, and found a similar TAS degradation across all TAS congeners (Radović et al. 2013).

Our findings suggest that in the investigated samples, straight-chain, branched, and cyclic alkanes were subject to biodegradation, while aromatic compounds were mainly degraded by sunlight. With some exceptions (TAS and homohopanes), biomarker compounds were recalcitrant on the observed time scale of years and are therefore suitable for fingerprinting MW oil.

Knowing which constituents in crude oil are either labile or recalcitrant will not only be useful for the development of new oil fingerprinting techniques, but can also help identify novel recalcitrant and potential toxic compounds of concern that should be included in standard monitoring programs after oil spills. In this study, GC×GC was an excellent tool to determine biomarker ratios. Second, this study illustrates that weathering of oil was accompanied by formation of recalcitrant oxygenated transformation products. Although we were able to characterize this fraction on a bulk level in terms of oxygen contend, functional groups and radiocarbon content, these OxHCs are still poorly characterized on a molecular level. Third, analyzing relative disappearance of compounds or compound classes provided evidence that biodegradation, as well as photooxidation, were acting on MW oil.

This research was made possible in part by grants from the NSF (OCE-0960841, RAPID OCE-1043976, RAPID OCE-1042097, EAR-0950600, OCE-0961725, OCE-1333148), and in part by a grant from BP/the Gulf of Mexico Research Initiative (GoMRI-015) and the DEEP-C consortium.