**4. Discussion**

*4.1. Surfaces of Coupons Before Cleaning*

The black corrosion products on the surfaces and sulfur detected in the biofilms (by EDS) of coupons from several treatments (T2, T3, T5 and T6) indicates iron sulfide production, and active sulfate reduction. This matches well with the presence of *D. desulfuricans*, when specifically added to the tests (i.e., T3, T5 and T6) and the relatively high potential for sulfate reduction determined from the METAGENassist analysis of the T2 orange tubercle inoculum. The iron:sulfur ratios calculated for the biofilms indicated that the highest to lowest relative amounts of sulfur detected were for T3 > T2 > T6 > T5, which qualitatively matches the thicknesses of the biofilms observed in SEM images.

One somewhat unexpected observation was that the nominally aerobic test with the *D. desulfuricans* (T3) produced a visually thicker biofilm than that for the corresponding anaerobic test with *D. desulfuricans* (T6). While they are more commonly known as strict anaerobes, many sulfate-reducing bacteria, including the culture collection strain used in this work, can survive/grow in aerobic environments [42,43]. It is also possible that the visually thicker biofilm observed for the *D. desulfuricans* aerobic incubation (T3) is a result of a stress response by the bacteria. Although the aerobic tests (including the T3 treatment) were nominally aerobic at the start of the tests it is likely that the oxygen concentration in the solutions dropped over time due to microbial oxygen consumption, making the environment more suitable for optimal growth of *D. desulfuricans*. A suggestion for future work would be to monitor the oxygen concentration in the solutions as a function of time.

## *4.2. Corrosion Attack*

Coupon weight loss indicates that T1 (uninoculated) and T3 (*D. desulfuricans* aerobic) treatments demonstrated greater general corrosion attack than the other treatments. However, T6 (*D. desulfuricans* anaerobic), T3 (*D. desulfuricans* aerobic) and T5 (four isolates + *D. desulfuricans* aerobic) treatments had significantly greater localised pitting than the other treatments. The pitting results for *D. desulfuricans* anaerobic incubation (T3) (average pit depth 45 μm) were greater than for previous tests for the same bacterial strain using a modified Baar's medium (average pit depth ~27 μm) with the same steel type [32]. However,

the corresponding weight loss data for the previous modified Baar's medium tests with *D. desulfuricans* were much greater (~80 μm/year) than obtained in this work (10 μm/year). It is important to note that in addition to pitting, general corrosion of the steel surface was observed in the previous study. This indicates that test medium used can have an impact on steel corrosion and pitting outcomes. Previous work indicated that the addition of Fe ions/lactate to modified Baar's medium can have a significant effect on the weight loss results, which are indicative of any general corrosion taking place [40]. No addition of Fe ions to the bulk test solutions was performed in the current work.

The main hypothesis for the work was that a consortium of microbes would result in greater corrosion rates than single isolates (T3 and T6) or an uninoculated control (T1). This was because a range of phenotypes were considered important in promoting ALWC. The consortium tests (T2, T4 and T5) actually showed the three lowest weight losses (typically indicative of general corrosion) of all of the tests performed. Microbial corrosion (including ALWC), however, is more often linked to localised rather than general/uniform corrosion. In relation to localised corrosion, the average pit depths were smaller for the consortia treatments (T2, T4 and T5) than for sole *D. desulfuricans* tests (T3 and T6). The volume of pits for the T5 treatment (four isolates + *D. desulfuricans*) were similar to the sole *D. desulfuricans* tests (T3 and T6), which were much greater than for any of the other test treatments.

One possible reason for the lower levels of corrosion obtained for the mixed microbial communities compared to the uninoculated control or single *D. desulfuricans* tests was that the types of microbes present (orange tubercle, T2) or chosen (pure cultures) were inappropriate for ALWC. The orange tubercle inoculum (T2), was taken from a site which had pitting corrosion and diagnosed ALWC beneath the orange tubercle. Thus, this orange tubercle consortium should have been appropriate to generate optimal ALWC conditions in our laboratory tests. Indeed, we previously found increased corrosion when orange tubercles (taken from another site with suspected ALWC) were used as inocula in laboratory tests, compared to an uninoculated control test [25]. The metabarcoding analysis of T2 identified several microbes present in both the biofilm and planktonic phases that were capable of sulfate reduction (more detail below) and sulfur was clearly identified in the biofilm by EDS.

High corrosion rates of ~100–125 μm/year (from weight loss) were previously [25] found for tests with orange tubercle inocula compared to ~40 μm/year in uninoculated controls when using a similar test arrangemen<sup>t</sup> (e.g., steel type, test duration) to that used in the current work. These previous higher corrosion rates were observed for samples with uniform corrosion across sample surfaces and with little localised corrosion. Two obvious key differences between the previous [25] and current studies were the initial microbial community structure of the inocula and the types of nutrients added to the test solutions. While a Deltaproteobacteria SRB (*Desulfarculus baarsii*) was detected in the orange tubercle inoculum (T2) test, this strain has not typically been linked to corrosion. In the previous study [25], an unidentified species from the Desulfobulbaceae family of Deltaproteobacteria was found at reasonably high levels (up to ~40%) in the planktonic phase. This family includes *Desulfopila corrodens* which was directly linked to rapid corrosion in ALWC studies [23].

Glucose (a carbon and electron donor source) and yeas<sup>t</sup> extract were components of the earlier test medium [25], while peptone and yeas<sup>t</sup> extract were present in relatively low amounts in the current work's medium. It has been shown that the types of nutrients available for microbial growth and the physicochemical conditions can have substantial effects on the composition of microbial communities that ensue (e.g., [44,45]). These two features likely explain the microbial differences observed between the earlier [25] and the current research. Additionally, changes in the level of carbon sources present in test media can affect the extent of corrosion caused by a single SRB strain (e.g., [29]).

The majority of microbes chosen for the defined consortium tests survived in reasonable numbers throughout the test period, although there were some clear changes in the microbial species composition over the test duration (discussed in more detail below).

There was also clear evidence of sulfate reduction, as seen by the black solution and presence of sulfur in the biofilm for the defined mixed microbial treatment, which included *D. desulfuricans* (T5), indicating that *D. desulfuricans* were metabolically active. As discussed above, it is possible that relatively low levels of carbon sources/electron donors in the test media may have altered physicochemical processes (i.e., sulfate reduction) that play a role in microbial corrosion and had an effect on corrosion rates.

There have been varying reports on the effect of the presence of different species of *Bacillus* on corrosion with some indicating increased corrosion [46–48] and others showing corrosion inhibition [49–53]. Both defined mixed microbial treatments, with (T5) or without (T4) *D. desulfuricans*, had similar levels of general corrosion but qualitatively different biofilm thicknesses, where T5 was greater than T4 (Figures 1 and 2). The 16S rRNA gene metabarcoding showed very different relative amounts of *B. aquimaris*. T5 had a lower general corrosion level than either of the tests with *D. desulfuricans* alone (T3 and T6), indicating that the additional microbes reduced the corrosion mediated by *D. desulfuricans*. This could indicate corrosion inhibition, perhaps due to competition for substrates among the added pure cultures.

#### *4.3. Analysis of Microbial Populations*

Although the *D. desulfuricans* strain used is a facultative anaerobe, plate counts showed that there were still reasonable numbers (103–104 cfu/mL) of viable *D. desulfuricans* present in solutions after 8 weeks when tested in nominally aerobic test conditions. The presence of sulfur in the biofilms of the coupons from these tests (e.g., T3) indicated that *D. desulfuricans* were metabolically active despite the possible presence of oxygen in the bulk solution. Metabarcoding analysis of the planktonic and biofilm phases of the defined consortium including *D. desulfuricans* (T5) showed a relatively high abundance of *D. desulfuricans* throughout this nominally aerobic test. Given the specific inclusion of aerobic microbes in T5 it is highly likely that they used the available oxygen creating anaerobic conditions, which are optimal for the growth of *D. desulfuricans*. This is important as it shows how testing with microbial consortia can produce an environment conducive to anaerobic microbes such as *D. desulfuricans* in a more natural way rather than by exogenously producing low oxygen levels by means such as nitrogen purging. There have been some reports on the potentially important role of oxygen in MIC of SRB [22,54–57] and of other microbes [58,59]. In any case, results from the current work support further studies on this topic.

The microbial consortium composition in the orange tubercle inoculum test (T2) differed over time in the planktonic phase (between 2 weeks and 8 weeks) and also between the planktonic and biofilm phases at 8 weeks. This is expected as the environmental conditions in the laboratory test was quite different to the tubercle's native location. The nature of the community succession will be different in the planktonic and biofilm phases as there will be local differences in oxygen content and corrosion products from the carbon steel could act as a nutrient source for certain microbes [6,60].

At 8 weeks incubation both planktonic and biofilm phases in T2 contained microbes potentially capable of sulfide/sulfur oxidation (e.g., *Thioalkalivibrio* sp.) and sulfate reduction (e.g., *Desulfosporosinus* spp. and *D. baarsii*). A combination of both sulfur oxidisers and sulfate reducers could form a closed sulfur cycle, which has previously been reported to potentially lead to rapid corrosion [5,8,13–17]. However, extensive corrosion was not seen in T2. It is possible that the system may have needed more time to develop an active corrosion state. Although strains of the Deltaproteobacteria (including *D. desulfuricans*) are the most commonly regarded sulfate reducers, other microbial groups also are capable for sulfate reduction. These include *Desulfosporosinus* spp., which is a member of the Firmicutes phylum. Various microbes relevant to the nitrogen cycle (e.g., nitrogen fixation, nitrite reduction and ammonia oxidation) were identified in the liquid and biofilm of T2 at 8 weeks. Nitrifying bacteria are important as they can potentially produce fixed nitrogen that is used to support/maintain the growth of anaerobic bacteria such as SRB [14]. However, conditions would need to be aerobic to support their nitrifying phenotype.

A number of interesting spatio-temporal changes were observed in the microbial communities of the defined consortia tests (T4 without and T5 with *D. desulfuricans*). Although *H. korlensis* comprised a reasonable proportion of the initial inoculum in both T4 and T5 (7% and 5%, respectively), this bacterium was only detected in relatively small relative abundances in the biofilm phase at week 8 (<10% of total). This might sugges<sup>t</sup> a preference for biofilm growth. Large differences were seen in the numbers of *B. aquimaris* present in the T4 and T5 treatments. It was the most abundant bacterium in both the planktonic phase and biofilms in T4, but it was only found in low abundances in T5. No explanation could be suggested for these differences, but since the presence of *D. desulfuricans* was the only difference, this could point to some interaction. *P. bellariivorans* made up ~1% of the initial inoculum according to viable cell counts, but it was quite abundant in T5 when in combination with the *D. desulfuricans* relative to its abundance without *D. desulfuricans* in T4. In T5, *P. bellariivorans* was more abundant than *B. aquimaris* or *H. korlensis* according to metabarcoding at weeks 2 and 8; despite that both of these latter bacteria comprised more cells in the initial inoculum relative to *P. bellariivorans*. These observed changes show how it is difficult to predict the species development in mixed consortia tests, which is likely to be affected by numerous factors including the test medium composition and physicochemical conditions. An increasing number of reports are starting to be produced investigating the development of multispecies biofilms (e.g., [61] and the references in Table S1). It will be interesting to keep a track of the development of this area of research and see what additional insights it might be able to provide microbial corrosion research.

The overall aim of this work was to investigate whether defined microbial species communities could be used to generate conditions conducive to the rapid corrosion of steel, similar to what occurs in ALWC in the marine environment. Notable differences in the biofilms and the microbial consortia were observed for the treatments tested. Lower uniform corrosion and weight loss were recorded for treatments with multispecies communities, however, the results for localised corrosion (typically more indicative of microbial corrosion) were not as straightforward, with one of the multispecies treatments (the defined consortium with *D. desulfuricans*) producing large sized pitting while another (orange tubercle inoculum) had much lower localised corrosion. Table 3 has been provided to help summarise some of the main observations of the work undertaken.


**Table 3.** Summary of observations from corrosion experiments using a variety of test conditions. Colours of cells have been used to highlight related results observed in different test conditions.
