Introduction

Fuel infrastructure is prone to being contaminated by microorganisms1. It is known that the booming use of biodiesel consisting of fatty acid methyl esters (FAME) entails fungi and bacteria to cause biofouling, fuel degradation and corrosion of metallic fuel infrastructure2. For example, a report on underground storage tank corrosion revealed an increase in corrosion cases attributed to microbial activity following the introduction of biodiesel fuels3. This trend is likely multifactorial: biodiesel has a higher saturation moisture content4 and a higher oxygen content5 than conventional diesel, and FAME are more readily metabolised than the hydrocarbons from conventional fuels. As a result, biodiesel blends tend to exhibit high levels of microbial contamination6,7. Floyd et al.8 however observed that the fungal preference for FAME over alkanes depends on the chain length (e.g. nonane being more readily degraded than most relevant FAME). In general, current strategies to mitigate microbial contamination and its consequences focus on the application of biocides, diagnostic test kits and/or preventive maintenance practices like regular cleaning and dewatering2.

These contaminations of fuel tanks are colloquially referred to as “fuel algae” even though they consist mostly of bacteria and fungi6,9,10. One of the most common isolates is Amorphotheca resinae also known as Hormodendrum resinae, Cladosporium resinae or Hormoconis resinae11. Its natural habitat remains debated: some proposing coniferous resins12, others soil13. In this context, Fürst et al.14 were able to isolate A. resinae from soil underneath conifers like Taxus baccata and Pinus spp. Beyond these natural environments, A. resinae has a long history of being isolated from anthropogenic environments such as tarred wood15, cosmetic face cream16 or oil- or tar-contaminated soil14,17. These days, A. resinae is however most known as the kerosene or diesel fungus, reflecting its dominance and ubiquity as a coloniser of fuel infrastructure9,18. This ability has been linked to its metabolic versatility: A. resinae can degrade a wide range of aliphatic and aromatic hydrocarbons19. Of interest in this context is as well its production of surface active compounds, likely enabling the uptake of these hydrophobic hydrocarbons20.

In contact with fuel infrastructure metals, fungi might contribute to corrosion. Most fungal corrosion studies focus on aluminium alloys21,22,23 or carbon steel8,24,25. Bento et al.24 studied carbon steel corrosion over 100 days with fungi grown on conventional diesel and found that Aspergillus fumigatus enhanced corrosion, Candida silvicola did not and A. resinae only enhanced corrosion at one timepoint. In contrast, both Wickerhamomyces sp. and Paecilomyces sp. were shown to enhance the corrosion rate of carbon steel when grown on a 20% (B20) biodiesel blend8. Our recent in situ investigation of a diesel tank (Gerrits et al.26, under review) showed that A. resinae - dominated biofilms enhanced the corrosion of galvanised carbon steel. In addition to enabling more microbial activity, biodiesel itself can also enhance corrosion: when contaminated with water, biodiesel undergoes hydrolysis, producing corrosive organic acids27. This process likely explains why biodiesel enhances the corrosion rate of carbon steel more than conventional diesel28,29,30.

As our first experiments with A. resinae isolates show melanin secretion when (and only when) grown with carbon steel, we focussed our study on the role of this pigment in corrosion. A. resinae is well known to produce HCl-precipitable melanin, both localised in the cell wall and secreted to its environment31,32. Melanin has the potential to adsorb metals32,33 but also acts as an electron donor or acceptor for redox-active metals like iron34,35. We hypothesise that fungal melanin with its metal redox and adsorbing capacities, has a role in the deterioration of iron-redox-sensitive materials as shown prior for the iron silicate olivine36. Therefore, a CRISPR/Cas9 genome editing method was applied on a diesel-isolate of A. resinae to modify genes putatively involved in melanin synthesis. The corrosivity of the resulting strains was analysed gravimetrically (uniform corrosion) and via interferometry (localised corrosion) in laboratory experiments. Furthermore, we conducted abiotic, chemical corrosion experiments with melanin harvested from different sources. In this study, we show how the multidisciplinary analysis of a built environment (Gerrits et al.26, under review) can be meaningfully extended by dissecting the deterioration mechanisms using axenic cultures of technically relevant isolates in combination with genome editing.

Results

Comparison of six A. resinae isolates from diesel tank with reference sources

Three strains – BAM 1069, BAM 1070 and BAM 1071 – were isolated from a 30-year-old diesel tank of a campervan (Gerrits et al.26 under review) and identified as A. resinae based on their ITS (internal transcribed spacer) sequences (Supplementary Fig. 1). Despite originating from the same source, the isolates were genetically distinct, differing by at least one nucleotide in the ITS region. ITS sequences of BAM 1069 and BAM 1071, but not BAM 1070, were detected in the metagenomic study of the same diesel tank carried out by Gerrits et al.26 (under review). Their growth characteristics were compared to those of three other A. resinae isolates from diverse origins: BAM 835 (isolated from jet fuel), CBS 406.68 (soil) and DSM 1203 (possibly kaopectate) (Supplementary Table 1).

Growth and conidiation varied among the six isolates but showed no clear correlation with either their kinship or the environment from which they were isolated. On solid Bushnell Haas (BH) medium supplemented with glucose (BH+glucose), DSM 1203 exhibited the fastest mycelial extension, significantly (p < 0.05) exceeding that of BAM 1070 and BAM 835 (Supplementary Fig. 2b). Conidiation on solid BH+glucose was significantly (p < 0.05) more pronounced for BAM 1069, BAM 1070, CBS 406.68 and BAM 835 than for BAM 1071 and DSM 1203, the latter forming whitish mycelium without conidia (Supplementary Fig. 2a and 2d). The effect of light (12 h light-dark cycle) was assessed using solid malt extract agar (MEA) medium. Under light exposure, mycelial extension decreased significantly (p < 0.05) for BAM 1069, BAM 1070 and DSM 1203 (Supplementary Fig. 2c), while conidiospore production increased significantly (p < 0.05) for BAM 1069 and BAM 835 (Supplementary Fig. 2e).

In liquid BH+glucose medium, DSM 1203 again demonstrated the highest biomass yield, considerably exceeding BAM 1071, both forming significantly (p < 0.05) more biomass than the other isolates (Fig. 1a). Regarding BH supplemented with biodiesel (BH+biodiesel, Supplementary Fig. 3a), all isolates formed significantly (p < 0.05) more biomass than the abiotic control, DSM 1203 again producing most biomass. However, direct comparisons between growth in glucose or biodiesel-supplemented liquid medium was problematic due to the incomplete evaporation of biodiesel during biomass drying, resulting in an overestimation of growth on biodiesel.

Fig. 1: Growth, acidification, the formation of acid precipitate as a measure of melanin and uniform corrosion rates of all six isolates of A. resinae.
Fig. 1: Growth, acidification, the formation of acid precipitate as a measure of melanin and uniform corrosion rates of all six isolates of A. resinae.The alternative text for this image may have been generated using AI.
Full size image

The isolates were grown in liquid BH+glucose cultures with and without carbon steel. The experiments ran for 35 days at 25°C, in the dark and at 100 rpm. a, b The biomass (g l-1) represents the dry weight at the end of the experiment. Without carbon steel, the differences between the isolates were generally larger than with carbon steel. Note that in the latter case also mineral precipitates could have been incorporated into the biomass fraction. c, d The bulk pH at the end of the experiment, the initial pH being 6.0. Differences between the isolates were again larger without carbon steel. e, f Melanin secretion measured as acid precipitate (g l-1); at the end of the experiment, the pH of the filtered ( < 0.2 µm) supernatant was decreased to ca. 2 using 1 M HCl and the dry weight of the precipitate was measured. No precipitate formed for the set-ups without carbon steel whereas large differences were recorded between the isolates with carbon steel. g To show that this precipitate was indeed melanin, Raman analyses of the filtered, acidified supernatant from the experiments with carbon steel were conducted. The spectra were similar for all isolates, showing peaks at 1361 ± 6 cm-1, 1450 ± 111 cm-1 and 1593 ± 5 cm-1, typical for melanin. h The corrosion rate (mm y-1) was calculated based on the weight loss of horizontally placed, small steel coupons. All isolates increased the corrosion rate, differences among them being minimal. For all experiments, shown in colour are the boxplots of three independent replicates. In grey the values of all three replicates are shown. The results of similar experiments with BH+biodiesel are shown in Supplementary Fig. 3. Significance at the level of 0.05 is indicated via a different letter (a, b, c, …).

The fungal reduction of the pH of the bulk nutrient solution agreed with the growth data: DSM 1203, BAM 1071 and BAM 835 decreased the pH of BH+glucose to 4.03, 4.15 and 4.52 pH units, respectively, significantly (p < 0.05) more than the other isolates (Fig. 1c). Also on biodiesel, DSM 1203, BAM 1071 and BAM 835 induced the most pronounced acidification, to 5.33, 5.10 and 5.57 pH units, respectively, (Supplementary Fig. 3b). Overall, pH reduction was however less than with glucose as the carbon source, indicating lower growth on biodiesel.

No precipitate formed upon acidification of the supernatant (hereafter referred to as “acid precipitate”), indicative of melanin formation as described by Oh et al.31, for any isolate, whenever grown in BH+glucose or BH+biodiesel (Fig. 1e and Supplementary Fig. 3c).

Response of A. resinae isolates to carbon steel exposure: melanin secretion and corrosion rates

Addition of a carbon steel coupon to the liquid BH+glucose set-up changed these parameters to some extent. Most visibly, the colour of the supernatant of all isolates turned brown (Supplementary Fig. 4). Apart for BAM 1069 and BAM 835, the biomass (including mineral precipitates) of all isolates was higher than for the abiotic control and – apart from significantly (p < 0.05) higher for DSM 1203 compared to BAM 1069 and BAM 835 – similar for all isolates (Fig. 1b). Isolates BAM 1069, CBS 406.68 and BAM 835 could significantly (p < 0.05) increase the pH compared to the abiotic control (Fig. 1d). For all isolates, a brown-coloured acid precipitate was observed in the supernatant; DSM 1203 (0.42 g l-1) and BAM 1071 (0.24 g l-1) produced significantly (p < 0.05) more of this compounds than the other isolates (Fig. 1f).

To confirm that the precipitate produced upon addition of steel was indeed melanin, Raman analyses were conducted. These showed the same two large peaks for all isolates: one at 1361 ± 6 cm-1 and the other at 1593 ± 5 cm-1, and one small peak at 1450 ± 11 cm-1 (Fig. 1g). These features are typical for melanin but could also imply the secretion of melanin precursors37. Whether or not these are actual melanin polymerised nanoparticles or only precursors will be discussed below, until so we will refer to these products as melanin.

We attempted to induce the production of melanin through the addition of single or multiple metals present in the steel. The addition of (i) 18 mM FeSO4*7H2O (at once in the beginning or as ten doses of 1.8 mM added over the course of 35 days), (ii) 5 to 50 µM CuSO4*5H2O, or (iii) a combination of 4.4 mM FeSO4*7H2O, 4 µM CuSO4*5H2O, 23 µM MnCl2 and 2 µM NiCl2*6H2O to liquid BH+glucose did not result in the production of an acid precipitate by BAM 1071 (data not shown).

All isolates, except BAM 1070, significantly (p < 0.05) enhanced the uniform corrosion rate of the carbon steel coupon relative to the abiotic control (0.009 mm y-1) (Fig. 1h). Minor differences were observed between the corrosion rates of all isolates: e.g., the rate of BAM 1070 (0.020 mm y-1) was significantly (p < 0.05) lower than that of BAM 835 (0.029 mm y-1).

A. resinae strains with altered melanin synthesis

Since a biotic enhancement of corrosion was observed only under conditions where melanin secretion occurred (i.e. in BH+glucose cultures), we sought to investigate the corroding properties of melanin in greater detail. Genomic analysis of A. resinae revealed orthologs for all genes known to be required for 1,8-dihydroxynaphthalene (DHN) melanin biosynthesis (Supplementary Table 2). These include pks1, encoding a polyketide synthase catalysing the first step in the de novo synthesis of DHN and cmr1, encoding a transcription factor that regulates DHN melanogenesis in several filamentous fungi including Alternaria alternata38 and Botrytis cinerea39. Pks1 and cmr1 are physically linked in the genome with further genes involved in DHN synthesis (Figs. 2a and 2b). We established a protocol for the genetic manipulation of A. resinae. Protoplasts of strain BAM 1071 were transformed with in vitro-assembled, target-specific sgRNA-Cas9 complexes in combination with target-specific donor DNA. Pks1 was deleted by replacing the coding sequence (Δpks1) and its expression was altered by replacing its promoter by the constitutive promoter oliC from Aspergillus nidulans (CE::pks1, Supplementary Fig. 5). As genetic control, pks1 was reintroduced into the ∆pks1 mutant by ectopic integration of the pks1 expression construct (Δpks1COM). To validate these procedures, cmr1 was deleted by a similar approach (Δcmr1), but the gene was reintroduced into the Δcmr1 mutant by a targeted integration approach (Δcmr1COM) (Supplementary Fig. 6). For each approach, several transformants were obtained that exhibited identical phenotypes.

Fig. 2: Melanogenesis in A. resinae.
Fig. 2: Melanogenesis in A. resinae.The alternative text for this image may have been generated using AI.
Full size image

a Biosynthetic pathway for DHN melanin in fungi. A polyketide synthase (PKS) catalyses the first step. From T4HN, multiple sequential enzyme-catalysed steps produce DHN, which is then polymerized to form the dark DHN melanin by poorly characterized laccase reactions outside the cell. Whether a yellowish-green hydrolase (YGH) is involved depends on the nature of the PKS. T4HN 1,3,6,8-tetrahydroxynaphthalene, T3HN 1,3,8-trihydroxynaphthalene, DHN 1,8-dihydroxynaphthalene, T4R T4HN reductase, SDH scytalone dehydratase, T3R T3HN reductase, MCO multicopper oxidase. b Most of the putative DHN-melanogenic genes are clustered in the genome of A. resinae. Genes putatively encoding the enzymes for DHN melanin synthesis shown in (a) were identified in the A. resinae ATCC 22771 genome based on sequence similarity (Supplementary Table 2). Pks1, thr1 and sdh1 are physically linked with cmr1 encoding a C2H2/Zn2Cys6 transcription factor, known to regulate the transcription of melanogenic genes in different fungi. c Genetic modification of pks1 and cmr1 loci in A. resinae BAM 1071 (WT) affects the production of melanin and thereby strain phenotype. Strains were cultivated for 13 days on two different solid media. Δpks1, Δcmr1 – deletion mutants (genes replaced by resistance cassettes); CE::pks1 – strain constitutively expressing pks1; Δpks1COM, Δcmr1COM – deletion mutants with re-introduced genes (genetic complementation).

Growth on solid MEA and SDNG confirmed the expected roles of these genes in melanogenesis. Deletion of pks1 prevented the formation of melanin on both tested solid media, while the constitutive expression of pks1 increased pigmentation, indicating the overexpression of pks1 under these conditions (Fig. 2c). The Δcmr1 mutant also formed whitish mycelium on MEA, but still produced some orange-brownish pigments on SDNG medium which might be pathway intermediates. In both cases, the reintroduction of the genes into the deletion mutants restored the wild type-like melanin formation demonstrating the reliability of the genome editing approaches.

Corrosion by melanin

To directly investigate the role of melanin in corrosion, we performed abiotic corrosion experiments using purified, secreted melanin. Apart from the steel-induced melanin obtained from BH+glucose cultures with carbon steel, we sought another melanin source. As the cultivation of the six isolates in malt extract broth (MEB) did not result in melanin secretion (data not shown), we turned to CE::pks1. When cultivated in MEB, this strain – but not the WT:BAM 1071 – produced a brown-coloured acid precipitate, even though both strains formed the same amount of biomass (Supplementary Fig. 7). This precipitate had a similar Raman spectrum compared to the steel-induced melanin (Fig. 3a), indicating that it represents melanin or a precursor. Scanning electron microscopy (SEM) analyses of non-acidified filtrates of steel-induced melanin contained spheres with a diameter of 20 to 40 nm in diameter, whereas no such structures were observed in the filtrates of CE::pks1 MEB cultures (Fig. 3b).

Fig. 3: The chemical and visual characterisation of melanin (precursors) from different sources together with their effect on the uniform corrosion rate.
Fig. 3: The chemical and visual characterisation of melanin (precursors) from different sources together with their effect on the uniform corrosion rate.The alternative text for this image may have been generated using AI.
Full size image

a Steel-induced melanin was harvested from the supernatant of BAM 1071 cultures incubated with a carbon steel coupon whereas melanin (precursor) from CE::pks1 was harvested from MEB cultures. Liquid samples from both were filtered and precipitated by addition of 1 M HCl. Raman spectra of acid-induced precipitates show the same peaks (1358 cm-1 and 1588 cm-1), albeit more pronounced for the CE::pks1-derived melanin (precursor). b Heaps of spherical structures with a diameter of 20 to 40 nm were observed via SEM imaging of the filtered, non-acidified supernatant of BAM 1071 cultures incubated with steel. These were not observed for the supernatant of the MEB cultures of CE::pks1. c Filtered, acidified, washed and neutralised melanin (precursor) suspensions were added to BH without a carbon source but with a horizontally placed steel coupon to obtain following concentration series: 0, 0.00316, 0.01, 0.0316, 0.1, 0.316 and 1 g melanin l-1. At low concentrations, melanin from both sources show a similar effect on the uniform corrosion rate, whereas at higher concentrations, the steel-induced melanin had a more pronounced corrosion accelerating effect. Shown in colour are the data of one independent replicate.

Purified (filtered, acidified, washed and neutralised) melanin (precursors) from culturing WT:BAM 1071 in BH+glucose with carbon steel (steel-induced) and from culturing CE::pks1 in MEB was added to carbon steel at increasing concentrations (0 to 1 g l-1). In both cases, the uniform corrosion rates increased with the melanin concentration but plateaued above 0.1 g l-1 (Fig. 3c). Interestingly, steel-induced melanin exhibited a stronger effect at higher concentrations (0.033 mm y-1 at 1 g melanin l-1, increase of 102% compared to 0 g melanin l-1) compared to the CE::pks1-derived melanin (0.020 mm y-1 at 1 g melanin l-1, increase of 48%).

Corrosion by A. resinae strains with altered melanin production

Although all strains could grow in liquid BH+glucose, the WT and Δcmr1COM decreased the pH significantly (p < 0.05) less than the other strains (Fig. 4e). On BH+biodiesel, all strains, except Δpks1, did produce significantly (p < 0.05) more biomass than the abiotic control. Δpks1 did moreover not significantly (p > 0.05) decrease the pH (Fig. 4f) of BH+biodiesel. The WT and Δcmr1COM grew significantly (p < 0.05) better than the other strains (Fig. 4b) while decreasing the pH to the highest extent (Fig. 4f). Note that both decreased the pH to the same extent as for BH+glucose. None of the strains were however able to form an acid precipitate in the absence of carbon steel (Fig. 4i, j).

Fig. 4: Growth, acidification, the formation of acid precipitate as a measure of melanin and uniform corrosion rates of the WT:BAM 1071 and its mutant strains.
Fig. 4: Growth, acidification, the formation of acid precipitate as a measure of melanin and uniform corrosion rates of the WT:BAM 1071 and its mutant strains.The alternative text for this image may have been generated using AI.
Full size image

The strains were grown with and without carbon steel and liquid BH medium with biodiesel or glucose. a–d Biomass (g l-1) formed after 35 days was generally higher with glucose than with biodiesel as a carbon source. When with carbon steel, also mineral precipitates could have been incorporated into the biomass fraction. e–h The bulk pH decreased for all strains from the initial pH of 6, when grown with glucose, without carbon steel. With carbon steel, the fungal strains were not able to change the pH strongly. The WT and Δcmr1COM were able to acidify the medium to a similar extent with biodiesel as with glucose. i–l Melanin secretion measured as acid-induced precipitate (g l-1); precipitate only formed with carbon steel and glucose (and for some samples of CE::pks1 with BH+biodiesel and carbon steel). m As also the putative melanin-deficient strains secreted a precipitate (which was lighter coloured), Raman analyses of the acid-induced precipitates were conducted. The spectra of CE::pks1 and Δcmr1COM were similar to the spectra of the WT, showing peaks at 1358 ± 3 cm-1, 1442 ± 6 cm-1 and 1600 ± 2 cm-1, typical for melanin (precursors). The spectra of Δpks1 and Δcmr1 however did not show these typical peaks, indicating an absence of melanin (precursors). The spectra of Δpks1 showed peaks at 560, 1098 and 2953 cm-1, whereas the spectra of Δcmr1 showed a broad peak at 1121 cm-1 and one at 2904 cm-1. Spectra of the glass substrate (control) onto which the samples were dropped showed two distinctive peaks: one at 560 cm-1 and the other at ca. 1090 cm-1. n, o The uniform corrosion rates were obtained from the weight loss of horizontally placed coupons and was similar for all strains: higher than the abiotic control with BH+glucose, similar to the abiotic control with BH+biodiesel. Shown in colour are the boxplots of four independent replicates. In grey the values of all four replicates are shown. Significance at the level of 0.05 is indicated via a different letter (a, b, c, …).

Addition of carbon steel turned the colour of the supernatant of the WT, CE::pks1, and Δcmr1COM to brown but the supernatant of Δpks1 and Δcmr1 to white (Supplementary Fig. 8). Due to the presence of iron oxide precipitates, not significantly (p > 0.05) more biomass could be detected for the biotic set-ups compared to the abiotic control when grown on biodiesel (Fig. 4d). The pH was similar for all abiotic and biotic set-ups, regardless of the carbon source, except a significantly (p < 0.05) lower pH for CE::pks1 and Δcmr1 compared to the abiotic set-up in BH+biodiesel (Fig. 4h).

With glucose, significantly (p < 0.05) more acid precipitate was observed for all strains, except Δcmr1COM, than for the abiotic control (Fig. 4k). However, the WT and CE::pks1 did not yield more precipitate than the other strains. Raman analysis of these precipitates showed the same three peaks for the WT, CE::pks1, and Δcmr1COM: 1358 ± 3 cm-1, 1442 ± 6 cm-1 and 1600 ± 2 cm-1 (Fig. 4m), features typical for melanin (precursors). For Δpks1 and Δcmr1, however, these peaks were absent. Instead, their spectra showed peaks at 560, 1098, and 2953 cm-1 (Δpks1) and at 1121 cm-1, and 2904 cm-1 (Δcmr1). Note that the peaks at 560 cm-1, 1098 cm-1 and 1121 cm-1 are likely from the analysis of the glass substrate (see control spectrum in Fig. 4m). A Bradford assay of the secreted product of Δpks1 and Δcmr1 was negative (i.e., below a detection limit of 125 µg ml-1) (data not shown), indicating that it was non-proteinaceous. For BH+biodiesel, differences in acid precipitate production between the biotic set-ups and the abiotic control were insignificant (p > 0.05) (Fig. 4l).

For BH+glucose, the uniform corrosion rates of all strains except CE::pks1 were significantly (p < 0.05) higher than the abiotic rate (0.010 mm y-1) but similar among each other, ranging from 0.025 mm y-1 to 0.027 mm y-1 (Fig. 4n). With biodiesel, biotic rates were not significantly (p > 0.05) different from the abiotic control, ranging from 0.010 mm y-1 to 0.012 mm y-1 (Fig. 4o).

The localised corrosion rate (mm y-1) was based on interferometric recordings of the deepest pits at the nutrient solution-biodiesel-air interface of vertically placed coupons (the interferometry maps of all coupon sides are shown in Supplementary Fig. 9). Although the differences between the set-ups were not significant (p > 0.05) due to the low repeatability, the localised corrosion rate for the abiotic control (0.29 mm y-1) was considerably higher than those of the biotic set-ups (0.04 to 0.15 mm y-1) (Fig. 5a). The lowest rate was observed for Δpks1 (0.08 mm y-1) and Δcmr1 (0.04 mm y-1). Note that these rates represent the medians of four independent replicates, each of which consisted of the median of the four deepest pits on a coupon. The deepest recorded pit overall (on an abiotically incubated coupon) had a depth of 72 µm at day 35, corresponding to a localised corrosion rate of 0.75 mm y-1.

Fig. 5: Localised corrosion rates and ESEM analysis of vertically placed carbon steel coupons incubated with WT:BAM 1071 and its mutant strains.
Fig. 5: Localised corrosion rates and ESEM analysis of vertically placed carbon steel coupons incubated with WT:BAM 1071 and its mutant strains.The alternative text for this image may have been generated using AI.
Full size image

The experiment ran for 35 days with liquid BH+biodiesel. a Interferometry was conducted on cleaned coupons at the nutrient solution-biodiesel-air interface. The four deepest pits (two on each side) of each coupon were recorded and their median was taken and converted to the localised corrosion rate (mm y-1). The rates of the abiotic control are considerably higher than those of the biotic set-ups. Note as well the low rates for Δpks1 and Δcmr1. Shown in colour are the boxplots of four independent replicates. In grey dots the values of all four independent replicates are shown. b–g ESEM images of the uncleaned coupons at the interface show that the abiotic surface (b) was covered with mineral precipitates and deep pits. Observations for all biotic set-ups were similar, images are shown as an example. The steel surfaces from the biotic set-ups were covered by a thick layer of mineral precipitates within a hyphal network (c for an overview, Δcmr1 as an example; e and g for details, Δpks1 and CE::pks1 as examples, respectively). Hyphae but also conidia were often covered by spherical precipitates (e). Underneath such layers, larger mineral precipitates were observed (d, WT:BAM 1071 as an example). These are likely oxidised vivianite (Supplementary Fig. 11 for chemical analyses). Single hyphae were also observed directly on the carbon steel (f, WT:BAM 1071 as an example). Significance at the level of 0.05 is indicated via a different letter (a, b, c, …).

The measured biomass and pH for these experiments with vertically placed coupons was the same for all set-ups (Supplementary Fig. 10). Acid precipitate was detected in all set-ups, but not significantly (p > 0.05) more than in the abiotic control (Supplementary Fig. 10c). We also measured the surface retreat of the bulk coupon surface below the nutrient solution-diesel-air interface and translated this into a uniform corrosion rate (Supplementary Fig. 10d): no differences were observed between the strains. Environmental scanning electron microscopy (ESEM) images of uncleaned coupons show that the abiotic nutrient solution-biodiesel-air interface was covered with smaller mineral precipitates (mostly prismatic, ca. 1 µm in diameter) and deep pits (Fig. 5b). The corresponding biotically incubated surfaces were covered by a fungal biofilm composed out of hyphal networks embedded in spherical mineral precipitates (ca. 1 µm in diameter; Fig. 5c, for an overview; Fig. 5e and 5g for details). Underneath such biofilms, and thus only for the biotic set-ups, larger mineral precipitates were observed (rosettes of prisms, up to 20 µm in length, Fig. 5d). Energy-dispersive X-ray spectroscopy (EDX) shows that these precipitates contained oxygen (54.17 atom. %), iron (20.03 atom. %) and phosphorus (12.62 atom. %). Raman spectroscopy of the same precipitates revealed one clear peak at 1007 cm-1 (Supplementary Fig. 11), likely reflecting the symmetric P-O stretching vibrations of santabarbaraite (Fe(III)3(PO4)2(OH)3*5H2O)40. This mineral forms via the oxidation of vivianite (Fe(II)3(PO4)2*8H2O), which was probably the original mineral underneath the fungal biofilms. Individual hyphae were also observed directly attached to the carbon steel surface (Fig. 5d, f).

Discussion

The ability of the fungus A. resinae to degrade diesel (a mixture of aromatic and aliphatic hydrocarbons) has been well documented9. A. resinae is able to degrade single alkanes, with optimal growth on undecane (C11H24)19. Another fungus, Paecilomyces variotii, formed more biomass when grown on FAME-based biodiesel compared to conventional diesel41. This is consistent with our observation that all tested A. resinae isolates grew on biodiesel (Supplementary Fig. 3). The three isolates isolated from biodiesel (BAM 1069, BAM 1070 and BAM 1071) did not, however, exhibit more growth on biodiesel compared to the other isolates, indicating that biodiesel degradation ability is independent on the environment of isolation. And even though our isolates tended to grow better on glucose than on biodiesel, it is noteworthy that WT:BAM 1071 was able to acidify the bulk nutrient solution to the same extent on biodiesel as on glucose (Fig. 4e, f).

Our primary focus was however the corrosion capacity of A. resinae. Before discussing the biotic effects, we will discuss the abiotic case, particularly the high localised corrosion rates observed via interferometry (Fig. 5a). The corrosive actions of biodiesel are well known by now: corrosion rates of carbon steel incubated with biodiesel are consistently higher compared to conventional diesel28,29,30. Several mechanisms contribute to these higher rates. (1) The oxygen solubility is higher for FAME mixtures than for water (ca. 42 ppm versus 8 ppm)42. Elevated dissolved oxygen levels accelerate the anodic reaction and thus enhance the corrosion rate of carbon steel43. (2) The saturation moisture content of biodiesel, regardless of its origin, is 15 to 25 times higher than that of conventional petroleum diesel, with values of around 1400 ppm at 21 °C4, enabling corrosion. (3) Water exposure promotes FAME hydrolysis, releasing free fatty acids27 which further enhance corrosion.

The observed high abiotic localised corrosion rates are probably due to the presence of the water-biodiesel interface. Wang et al.44, using a wire beam electrode approach, studied the corrosion of carbon steel submerged partially in tap water and partially in biodiesel. They reported, in agreement with other studies28,29, iron oxide precipitation on steel surfaces exposed to water but not on those in contact with biodiesel. Wang et al.44 also demonstrated that the anodic reaction (i.e. the oxidation of metallic iron) mainly occurs in the aqueous phase, intensifying further away from the biodiesel-water interface. The cathodic reaction (i.e. the reduction of oxygen) occurs at the water-biodiesel interface, where the oxygen concentration is highest. Taken together, these mechanisms explain the high localised corrosion rates we observed for the abiotic control (i.e., up to 0.75 mm y-1). These values are in line with the localised abiotic corrosion rates reported in another study for carbon steel incubated in B20 biodiesel8.

The impact of A. resinae on carbon steel corrosion strongly depended on the available carbon source. With biodiesel, fungal growth inhibited localised – but not uniform – corrosion (Fig. 4, Fig. 5 and Supplementary Fig. 10d). In contrast, with glucose, A. resinae enhanced uniform corrosion (Fig. 1 and Fig. 4).

The inhibition of localised corrosion by A. resinae is not entirely unexpected. The inhibition of steel corrosion by microbial colonisation, i.e. bioprotection, has been reported for other microorganisms grown on both conventional diesel and biodiesel45,46. Moreover, an in-situ study of carbon steel corrosion in biodiesel tanks showed maximum localised corrosion rates of 0.094 mm y-147, similar to our observed biotic localised corrosion rates. The bioprotective effect of A. resinae, observed in our study, is probably linked to the formation of fungal biofilms consisting of hyphal networks which (1) immobilise corrosion-inhibiting, (bio)mineral precipitates, (2) generate localised anoxic microenvironments and thus (3) mediate the biomineralization of other corrosion-inhibiting mineral precipitates like vivianite. Three morphologically-distinct types of mineral precipitates were observed: submicron ( < 1 µm) spheres seen mostly on hyphae or within hyphal networks, ca. 1 µm in diameter prisms observed on the steel surface of all set-ups and larger prisms grouped in rosettes ( < 20 µm in diameter) which were observed underneath the hyphal networks (Fig. 5). We were only able to identify the larger rosettes (Supplementary Fig. 11): the ferrous phosphate vivianite. The smaller spheres are likely amorphous iron oxides. We observed more of these precipitates within the hyphal network than on the abiotic steel surface (Fig. 5), indicating that A. resinae could immobilise these precipitates. Similar observations were made for fungi colonising an iron nail from a whale skeleton: an iron-rich, so-called shell, covering the substrate, embedding fungal cells48. These authors also suggested that the surface of the fungal cells (incl. the extracellular polymeric substances) might have acted as a nucleation site for the mineral precipitates, a mechanism also compatible with our observations of hyphae-associated precipitates (Fig. 5e). The ferrous iron present in such precipitates is thought to inhibit corrosion by reducing oxygen, preventing its diffusion to and reduction at the steel surface49. Vivianite by itself could act as a corrosion inhibitor50,51. However, as the precipitation of vivianite is mediated by the accumulation of ferrous iron under anoxic conditions52, its presence underneath A. resinae hyphal networks indicates that fungal biofilm created local anoxic niches. Such environments can in principle establish oxygen concentration cells, enabling localised corrosion53. This has been the proposed mechanism to explain the higher localised corrosion observed underneath patchy A. resinae-dominated biofilms in a diesel tank (Gerrits et al.26, under review). We, however, consider it unlikely that such a mechanism could have acted here as the observed biofilms were continuous, not allowing a cathode to form.

The observed inhibition of localised corrosion by A. resinae contradicts the findings of Floyd et al.8 who observed the acceleration of localised corrosion by fungi grown on biodiesel. This discrepancy between our results could have been caused by the higher extent of fungal acidification, the use of B20 biodiesel, the use of different species, or — less likely — the static character of their set-up. First of all, Floyd et al.8 observed fungal cultures to reduce the pH by ca. two to three pH units in the presence of carbon steel, explaining the higher fungal corrosion rates. Considering that, without carbon steel, A. resinae was able to lower the pH as well by two pH units (Fig. 4f), corrosion of carbon steel likely buffered the pH to a higher extent in our experiments, counteracting fungal acidification. The most likely explanation is a higher steel-to-solution ratio in our set-up. Secondly, Floyd et al. 8 used a higher biodiesel content (i.e., B20), likely causing more fungal growth and possibly explaining the accelerating effect on localised corrosion. Thirdly, the fungal species used by these authors might produce strong iron chelators, which A. resinae might not be able to secrete. And lastly, the absence of rotation could, by causing oxygen depletion, have hindered fungal growth.

With glucose as the carbon source, A. resinae did enhance the uniform corrosion rate of carbon steel (Figs. 1h and 4n). The increased biomass production and/or the slightly lower pH of glucose-grown cultures may have shifted the balance from net protection to net corrosion. Confirming this hypothesis is the correlation between the absence of an acceleration of uniform corrosion and the much lower biomass production and acidification with biodiesel. However, comparisons of different A. resinae isolates shows that neither biomass yield nor bulk pH correlated with the corrosion rates (Fig. 1). Moreover, acidification is probably not the primary biocorrosion mechanism as some A. resinae strains decreased the pH of the nutrient solution without carbon steel to a similar extent with both carbon sources (Figs. 4e, f), yet only caused uniform corrosion only in the presence of glucose.

In the absence of a clear corrosion-enhancing mechanism it struck us as most interesting that all isolates had a brown supernatant when – and only when – grown with carbon steel (Supplementary Fig. 4). Acidification of the filtrate ( < 0.2 µm) of this supernatant yielded a precipitate (Fig. 1f) for which Raman analyses showed two clear peaks at 1361 ± 6 cm-1 and 1593 ± 5 cm-1 and a smaller one at 1450 ± 11 cm-1 (Fig. 1g). Such peaks were also observed for DHN and L-3,4-dihydroxyphenylalanine (DOPA)-melanised biomass of the fungi Botrytis cinerea37, Aspergillus fumigatus54, Knufia petricola55, and Cryptococcus neoformans54, eumelanin and pheomelanin in the human skin56, and the DHN monomer37. Moreover, deletion of genes putatively involved in DHN melanogenesis (pks1 and cmr1)38,57,58 abolished both pigment formation (Fig. 2c and Supplementary Fig. 8) and the associated Raman peaks. SEM analyses further revealed spherical nanoparticles in the filtered supernatant from the carbon steel set-ups of BAM 1071 (Fig. 3b) – but not of Δpks1 (data not shown) –, resembling those claimed by Oh et al. 59 to be melanin. Together, these data indicate that the acid-precipitated compounds secreted by A. resinae indeed are de novo synthesized melanin. However, due to the possibility of a melanin composed out of different monomers60 or the incorporation of other compounds61, the exact chemistry of the melanin of A. resinae remains unknown. The extracellular polymer should therefore be more correctly referred to as melanin-like. Constitutive expression of pks1 (CE::pks1) was confirmed by the production of reddish-brown mycelium (Fig. 2c), as observed before for B. cinerea39, and the secretion of a dark coloured acid precipitate when grown on MEB (Supplementary Fig. 7). Raman peaks of this precipitate (Fig. 3a) were similar to previously reported melanin (precursors) samples37,54,55 and the steel-induced melanin-like polymers (Fig. 1g). However, unlike for the latter, no nanoparticles could be detected in the filtered supernatant of the MEB culture of CE::pks1 (Fig. 3b). All combined, this suggests that the constitutive expression of pks1 resulted in the secretion of a melanin precursor, whose exact molecular formula and structure are unknown. The chemistry of the whitish acid precipitate observed in the Δpks1 and Δcmr1 carbon steel BH+glucose cultures (Fig. 4k) is as well unknown.

The induction of melanin secretion by the presence of steel has to our knowledge not yet been described but could not be replicated through the addition of single or multiple metals like iron, copper or manganese. This opposes previous reports of copper-enhanced melanin production by A. resinae32,62. We hypothesize that such a steel-induced secretion is a protection mechanism against metal toxicity or free radicals as proposed previously63.

Melanin has the capacity to accelerate both uniform (Fig. 3c) and localised (Fig. 5a) corrosion. The corrosive effects of melanin have been suggested before based on the correlation between the secretion of pyomelanin and the enhancement of carbon steel corrosion by the bacterium Pseudoalteromonas lipolytica upon incubation under light64. Several mechanisms could explain such effects. (1) The pigment might accept electrons from the steel65,66, acting as the cathode, surpassing the bioprotective effects discussed above. (2) It is able to reduce ferric iron to ferrous iron34,67, making it able to prevent the formation of or to dissolve the passivating iron oxides in a process similar to the one proposed for the corroding iron-reducer Shewanella oneidensis68. And lastly, (3) melanin can adsorb ferrous iron34,67,69. Żądło and Sarna34, hypothesised based on extensive spectroscopic analyses that reduced melanin subunits at the exterior of the pigment can reduce ferric iron. Ferrous iron is subsequently adsorbed to the pigment and can then be oxidised by the oxidised, central melanin subunits. This melanin-adsorbed iron could form a galvanic cell70 but also lower the iron concentration at the steel-water interface, preventing the precipitation of iron oxides. We propose that the effect of melanin depends on the local redox environment. In oxygen-rich conditions (e.g., without a biofilm), melanin reduces ferric iron and adsorbs ferrous iron (Fig. 6). In oxygen-poor conditions (e.g., underneath the biofilm), ferric iron is absent, and melanin acts as an electron acceptor while continuing to adsorb ferrous iron. The higher corrosion rate of the steel-induced melanin-like polymer compared to the putative melanin precursor of CE::pks1 (Fig. 3c) can be explained by the former’s higher capacity to accept electrons and reduce or adsorb iron.

Fig. 6: Proposed, hypothetical mechanism for the localised corrosion of carbon steel by A. resinae grown on biodiesel based on the literature and our observations.
Fig. 6: Proposed, hypothetical mechanism for the localised corrosion of carbon steel by A. resinae grown on biodiesel based on the literature and our observations.The alternative text for this image may have been generated using AI.
Full size image

a Abiotic conditions: the high oxygen and water content of biodiesel, along with the presence of fatty acids likely contributed to elevated localised corrosion rates of up to 0.75 mm y-1. b Biotic conditions: melanin-free A. resinae reduces localised corrosion rates by forming biofilms consisting of a hyphal network and corrosion-inhibiting iron precipitates on top of the steel. These biofilms inhibited corrosion by preventing oxygen to reach the steel surface. c Role of melanin: despite the protective biofilm, melanin in the fungal cell wall appears to accelerate localised corrosion. We hypothesise, based on previous findings regarding the redox chemistry of melanin, that underneath the fungal biofilm where oxygen is depleted, melanin might act as an electron acceptor and adsorbs ferrous iron. In contrast, in oxygen-rich environments without biofilm coverage, melanin may reduce ferric iron and continue to adsorb ferrous iron.

The discrepancies between the results from our different experimental approaches are both puzzling and illuminating as they underscore the complexity and limits of the proposed mechanisms. With biodiesel, our strains had either no effect on uniform corrosion (Fig. 4o and Supplementary Fig. 10) or exerted a protective effect on localised corrosion (Fig. 5a). And while uniform corrosion was accelerated by secreted melanin without a biofilm (Fig. 3c), there was no effect when biofilms were present (Fig. 1 and 4n). Localised corrosion, however, was enhanced by either secreted or cell-wall bound melanin (Figs. 4o and 5a). Dissecting these paradoxes, consider that uniform and localised corrosion are driven by distinct processes71. Bioprotection mechanisms may be more effective suppressing localised corrosion, whereas secreted melanin enhances uniform corrosion only when it has direct access to the steel surface. To conclude, in conditions where corrosion of steel buffers the pH (unlike the observations of Floyd et al.8) and where biofilms can grow laterally (unlike the patchy biofilms observed in the diesel tank of Gerrits et al.26 (under review)), A. resinae biofilms appear prone to protect carbon steel from corroding even though their cell wall-bound melanin would impel them to do otherwise (Fig. 6).

Methods

Isolation and cultivation of A. resinae

For isolating fungi from the fouled diesel tank of a Fiat Ducato I 280 campervan in 2022 (described in Gerrits et al.26 under review), a floc sample from the supernatant of the tank was taken and spread onto dichloran rose bengal chloramphenicol (DRBC) agar (Roth). After five weeks, mycelia of three individually growing, filamentous, brown colonies were transferred to individual malt extract agar (MEA) (20.0 g l-1 glucose·H2O, 1.0 g l-1 peptone from casein, 20.0 g l-1 malt extract, 20.0 g l-1 Kobe I agar) and cultivated for twelve days at 25 °C. Ten conidia, harvested from these plates, were incubated for three days on MEA with 100 µg ml-1 ampicillin and 100 µg ml-1 chloramphenicol. Mycelia of the three isolates were finally incubated on MEA with cellophane for DNA isolation. Genomic DNA was isolated with the FastDNA SPIN Kit for Soil (MP Biomedicals) from 7-day-old mycelia from the cellophane-covered MEA. To visualise DNA, it was mixed with Midori Green Direct (Biozym Scientific) and separated in 1% agarose gels with the 1 kb Plus DNA Ladder (New England Biolabs, NEB) as a size reference. PCR reactions for the amplification of regions of interest, e.g., of the internal transcribed spacer 1 and 2 (ITS1 and ITS2) and the 5.8S ribosomal RNA (rRNA) gene localized in between, were carried out with primers from Eurofins Genomics (Supplementary Table 3) and the Q5 High-Fidelity DNA polymerase (NEB). PCR products were purified with the Monarch PCR and DNA Cleanup Kit (NEB) and sequenced with the Mix2Seq Kit at Eurofins Genomics. Evaluation of the sequence reads, the generation of consensus sequences, and sequence comparisons were performed with the embedded tools in Geneious Prime 2024. The isolated strains BAM 1069, 1070, and 1071 were deposited at the German Collection of Microorganismens and Cell Cultures (DSMZ) with the accessions DSM 115.529, DSM 115.530, DSM 115.531. Details on strains and sequences used are summarized in Supplementary Table 1. The exact origin of strain DSM 1203 (BAM 123, NRRL 2778, NBRC 100535) is not entirely certain. Fonken and Murray72, likely deposited the strain at the Agricultural Research Service (ARS) Culture Collection as NRRL 2778. From here it was obtained by Klaus Kieslich, who deposited it at the DSMZ-German Collection of Microorganisms and Cell Cultures as DSM 120373. Gunther S. Fonken also submitted an A. resinae strain to the Westerdijk Fungal Biodiversity Institute in 1959 (CBS number 161.59), stating it was isolated from kaopectate with a reduced amount of methyl p-hydroxy-benzoate (see as well74). Moreover, genotypic identification showed that CBS 161.59 is akin to DSM 1203 (Supplementary Fig. 1). DSM 1203 was therefore likely isolated from kaopectate. Strain BAM 835 was isolated by Pedro M. Martin-Sanchez from aviation fuel from an Airbus A380-841 airplane. Strain CBS 406.68 (BAM 801, JCM 11449, ATCC 200942) was isolated by Douglas G. Parberry75, from “the heart of a grass tussock under firs and pines near Capel Curig”, Wales, the United Kingdom.

Strains were typically kept on MEA at 4 °C. Before the start of an experiment, an agar plug from these plates was placed on fresh MEA and grown for seven days in the dark at 25°C. The inoculant was obtained either in the form of a conidial suspension or as an agar plug. Conidial suspensions were prepared by scraping off conidia from an agar culture using an inoculation loop, followed by two washing cycles with sterile, ultrapure water. The conidia concentration of the final suspension was quantified using a Thoma cell counting chamber (Marienfeld).

Light microscopy of strains grown on solid medium was conducted using a Zeiss Axio Imager M2m microscope with a 63x objective and immersion oil. A sterile glass cover slip was positioned next to a mycelial plug for 10 days and placed on a microscope slide using ultrapure water as mounting solution.

Genetic modification of A. resinae BAM 1071

Genes putatively involved in melanogenesis were identified in the genome sequence of A. resinae strain ATCC 22711 (Supplementary Table 2). Subsequently, chosen genes were amplified from genomic DNA of the A. resinae BAM 1069, 1070, and 1071 and sequenced as described above. For editing loci of interest, protoplasts of A. resinae were transformed with preassembled target-specific ribonucleoproteins (RNPs) and donor DNA. Protospacers for Cas9 were chosen with the CRISPR site finding tool of Geneious Prime 2024 using the genome sequences of the A. resinae strains ATCC 22711, ZN1, and KUC3009 as off-target databases. sgRNAs were synthesised in vitro via the EnGen sgRNA Synthesis Kit (NEB), purified with the Monarch RNA Cleanup Kit (NEB), and assembled in a molar ratio of 1:1 with EnGen Spy Cas9 NLS (NEB) by incubation for 10 min at 22 °C. Donor DNA i.e., resistance/expression cassettes flanked by sequences homologous to the integration sites was generated in different ways. i) Donor DNA with 75-bp-long homologous sequences was generated by PCR using primers containing 5’ overhangs (Supplementary Table 3), plasmids of the pNDR-OGG series76 as template and the Q5 High-Fidelity DNA polymerase (NEB). ii) Donor DNA with longer homologous sequences was isolated by digestion with suitable restriction enzymes from cloned plasmids (Supplementary Tables 4 and 5). Protoplasts of A. resinae were generated according to Erdmann et al.77 using the same buffers and media, but with minor adjustments in the implementation. 20 ml of MEB was inoculated with 105 conidia and incubated for two days at 25 °C and 100 rpm. Mycelia were harvested by centrifugation, washed twice with protoplast buffer, resuspended in 20 ml protoplast buffer containing 0.8 g VinoTaste Pro (Novozymes) and 0.02 g Yatalase (Takara Bio Inc.) and incubated at 27 °C and 80 rpm for 2 h. The suspensions were filtered through a 40-µm cell strainer (BD Falcon). Protoplasts were harvested by centrifugation (5 min, 1000 x g, 4 °C), washed twice with transformation buffer and re-suspended in the same buffer to obtain a titer of 2 ×107 protoplasts ml-1. The RNP, i.e., 1 µg of the sgRNA complexed with 5 µg of Cas9, and 10 to 15 µl of the donor DNA were mixed on ice. 106 protoplasts (50 µl) were added and incubated for 20 min on ice. After the addition of 100 μl of 24% (w/v) PEG 6000 (heated to 37 °C), the samples were incubated for 30 min at room temperature. 750 µl of liquid transformation medium was mixed in gently. Aliquots of 50 µl were transferred to empty Petri dishes. 10 ml of moderately warm TM with 1% agar was added per Petri dish and carefully spread. The next day, 10 ml of TM agar containing either 100 µg ml-1 hygromycin B (HYG) or 90 µg ml-1 nourseothricin (NTC) as selective agent were added per Petri dish. Cultures were incubated in the dark at 25 °C until putative transformants pierced the agar after approx. 5 to 8 days for HYG and NTC, respectively. Mycelia from the upper agar layer were transferred onto MEA supplemented with 50 µg ml-1 HYG or 45 µg ml-1 NTC for secondary selection. The genomic DNA from resistant transformants was isolated for genotyping. The correct integration of deletion and expression constructs as consequence of homologous recombination (HR) events was detected by diagnostic PCR with the Taq DNA polymerase (NEB) by combining primers binding upstream or downstream of the integration site with those binding within the integrated sequences i.e., in resistance or expression cassettes (Supplementary Figs. 5 and 6). For studying the pigmentation, WT:BAM 1071 and the generated strains were cultivated on MEA and solidified synthetic-defined nitrate glucose (SDNG) medium (20.0 g l-1 glucose, 3.0 g l-1 NaNO3, 1.7 g l-1 BD Difco Yeast Nitrogen Base without Amino acids and Ammonium Sulfate, pH 5.0).

Quantification of growth and conidiation on solid medium

Growth and conidiation on solid medium were assessed by growing the strains on solid MEA (pH set at 6) or Bushnell Haas medium (BH78, 1.662 mM MgSO4, 0.136 mM CaCl2·2H2O, 7.3 mM KH2PO4, 5.7 mM K2HPO4, 13 mM NH4NO3, 0.185 mM FeCl3·6H2O, pH set at 6) amended with 15.0 g l-1 glucose·H2O and 15.0 g l-1 bacteriological grade agar (i.e., solid BH+glucose). Petri dishes with 25.0 ml agar medium were inoculated by placing agar plugs in their centre and incubated for ten days at 25 °C, either under continuous darkness or 12-hours alternations of light (90 µmol photons m-2 s-1) and darkness. Growth is shown as mycelial extension which is based on the diameter (mm) of single colonies grown on solid medium from agar plugs, for ten days at 25 °C. Conidia production (conidia / mm2) was quantified by removal of all conidia from the mycelium grown for ten days at 25 °C, quantification via a Thoma chamber and dividing this quorum by the area occupied by the mycelium. Both experiments were independently reproduced three times.

Quantification of growth and acidification in liquid medium

To obtain the biomass production and acidification potential of the strains, 1000 conidia of each strain were grown in 50 ml falcon tubes filled with 10.0 ml liquid BH medium (initial pH of 6). The BH medium was either amended with 15.0 g l-1 filter sterilised ( < 0.2 µm) glucose·H2O (i.e. BH+glucose) or with 50 ml l-1 B7 biodiesel79, which is a blend, expected to contain up to 7% biodiesel (BH+biodiesel). The tubes were covered with the unscrewed lid and two layers of parafilm and incubated for 35 days at 25°C, darkness, and 100 rpm. After 35 days, biomass was harvested via centrifugation (5 min, 5000 g) and dried at 65°C to obtain a dry weight. Not all biodiesel evaporated during the incubation at 65°C, causing an overestimation of the biomass formation in the BH+biodiesel set-ups. The pH was measured of a filtered ( < 0.2 µm) medium sample. This experiment was independently conducted three times.

Running carbon steel corrosion experiments

Carbon steel sheets (1.0338 DC0480, similar to ASTM A620, ca. 1 mm thick) were either cut in 10 mm to 10 mm coupons for the so-called horizontal experiments or as 98 mm to 16 mm coupons for the so-called vertical experiments. These coupons were sterilised by submerging them for 20 s in acetone and placed in 50 ml falcon tubes filled with 10.0 ml liquid BH medium (without FeCl3·6H2O, with an initial pH of 6), which was inoculated with 1000 conidia of the respective strain and covered with two layers of parafilm. The tubes were placed for 35 days on a shaker (100 rpm) at 25°C and darkness. The carbon source was either glucose (i.e. BH+glucose) or biodiesel (i.e. BH+biodiesel), at the same concentrations as described above. Note that the small, horizontal, coupons were completely submerged in nutrient solution whereas the large, vertical, coupons were in partial contact with the nutrient solution, biodiesel and air.

At the end of all experiments, liquid samples were taken and centrifuged (5 min, 5000 g). The pH and melanin secretion were measured as described above. The horizontally and vertically placed coupons were taken out of the Falcon tubes and, after removal of most of the biofilm, incubated whilst shaking in 5 g N,N’-dibutyl-thiourea l-1 37% HCl, diluted 1:1 with ultrapure water for 4 min, 2 min in 1.2 M NaHCO3, rinsed subsequently with ultrapure water and acetone and finally air dried before measurement of the uniform and localised corrosion rates. The biomass remaining in the Falcon tube, removed from the coupon and the liquid sample, was collected and dried overnight at 65°C to obtain an estimate of fungal growth. However, it must be noted that mineral precipitates likely constituted a significant part of this biomass fraction. Additionally, ESEM imaging was performed on the vertical coupons which were not cleaned prior. Corrosion experiments for the mutant strains were independently repeated four times; the experiments for the isolates were independently conducted three times.

Quantification of uniform and localised corrosion

Uniform corrosion was assessed of the small, horizontal coupons by obtaining their dry weight and calculation of their weight loss, allowing the calculation of the uniform corrosion rate (mm y-1) via following formula:

$${corrosion}\,{rate}\,(\frac{{mm}}{y})=87600\times \frac{{weight}\,{loss}\,{coupon}(g)}{{density}\,{steel}\left(\frac{g}{{{cm}}^{3}}\right)\times {surface}\,{area}\,{coupon}\left({{cm}}^{2}\right)\times {duration}\,{experiment}(h)}.$$

The density of 1.0338 DC04 steel is 7.86 g cm-3. The surface area of the coupon was estimated based on the initial weight and the density, assuming an equal width and length (the relative error of this estimated area is smaller than 1.0%, based on the coupon with the largest discrepancy in width and length).

The surface topography of the vertical coupons at the nutrient solution-biodiesel-air interface was determined to assess localised and uniform corrosion. This was acquired using a Nexview (Zygo/AMETEK) scanning white light interference microscope, also called interferometry, allowing to acquire 3D images with a vertical resolution in the range of approximately 1 to 2 nm. A 5.5x Michelson interferometer objective was used with a lateral optical resolution of 1.9 μm. To enlarge the field-of-view, the internal optical zoom with 0.5x magnification and the stitching mode were used. In this mode, adjacent fields-of-view are matched at their edges and joined together, with a defined overlap of 40%, into a large image of 5000 µm by 10,000 µm. Different software-based methods of levelling were chosen, the uncorroded surface was always used as a reference level. A ridge, separating the corroded, bottom part of each coupon with the uncorroded, top part, was observed for all samples at similar locations. To quantify the depth of corrosion pits, single scanlines were drawn within the region of interest (i.e. an area of 5000 µm by 5000 µm below the ridge). By using OEM software Mx (Zygo), the depth of the pits could be quantified. The calibration and traceability of the white light interference microscope is ensured by certified standards within a DIN EN ISO/IEC 17025:201881 accredited lab. Of each side of each coupon, the depth of the two deepest pits was recorded (i.e. four pits per coupon) and a localised corrosion rate (mm y-1) was calculated. The shown data point for each of the four coupons consists of the median of these four recorded pits as we noticed the presence of outliers. We also used the drawn scanlines to measure the bulk surface retreat compared to the uncorroded surface as another measure of uniform corrosion.

Quantification of melanin secretion including its analysis by Raman and SEM

Melanin secretion by the A. resinae strains was quantified according to Oh et al.31. The filtrate ( < 0.2 µm) of liquid cultures grown for 35 days as described above was used for the quantification of the acid-induced precipitate as a measure of melanin secretion. Briefly, the filtrate was acidified to ca. pH 2 using 40 µl 1 M HCl ml-1, incubated overnight at room temperature, centrifugated (5 min, 5000 x g), dried overnight at 65 °C and weighted. The biomass was collected via centrifugation (5000 x g, 5 min), dried for three days at 65 °C and finally weighed.

For the abiotic corrosion experiments with melanin, melanin was harvested from BH+glucose cultures of BAM 1071 incubated with a horizontal coupon or MEB (i.e. MEA without addition of agar) cultures of CE::pks1, both 35 dpi. The supernatant was filtered ( < 0.2 µm), acidified to pH 2 using sterile 40 µl 1 M HCl ml-1 (to precipitate the melanin), and incubated overnight at room temperature. The next day, the supernatant was removed after centrifugation (5 min, 5000 g) and the pellet was washed twice with sterile ultrapure water. After homogenisation, the melanin concentration was determined by measuring the dry weight of an aliquot (overnight incubation at 65 °C). The remaining melanin was resuspended by bringing the pH back to 6 using sterile 1 M KOH. This melanin suspension was added to 10 ml BH without a carbon source to obtain the following concentration series: 0 - 0.00316 - 0.01 - 0.0316 – 0.1 - 0.316 - 1 g melanin l-1. The experiment ran with small, horizontal, coupons for 35 days as described above.

Raman spectra of this filtered, acidified, but not yet dry supernatant were obtained using a Labram HR800 (Horiba/Jobin Yvon) Raman microscope with 532 nm continuous-wave laser excitation (diode-pumped solid-state laser, 16 mW maximum power at the sample surface, reduced to 1.6 mW by a neutral density filter). The laser light was focused onto the sample surface and the reflected and/or scattered light was collected in upright configuration by using a 50x objective, with a focus diameter of approximately 1.2 µm. Dispersion of the Stokes-Raman-scattered light in a 800 mm spectrometer was accomplished with a 300 mm−1 grating and spectra were detected by a Peltier cooled (-60 °C), charge-coupled device (CCD) Syncerity camera (Horiba/Jobin Yvon) having 1024 pixels along the wavenumber axis, resulting in spectra ranging from approx. 190.5 cm−1 to 3327.5 cm−1 with a spectral resolution of 3.7 cm−1 to 2.5 cm−1 per CCD pixel. Raman maps were gathered by LabSpec 6 software (Horiba/Jobin Yvon) allowing stepwise movement of the sample stage through the laser focus with a step size of 17.5 to 21.7 µm. The acquisition time per spectrum, respectively, was 1 s with 5 accumulations. Single spectra, acquired independently of mappings, were typically measured with a 1 s acquisition time and 3 accumulations. The acidified and filtered supernatant of BH+glucose corrosion experiments and MEB cultures was dropped and air-dried on a cleaned glass microscopy slide. Measurement areas on these dried drops were randomly selected. For each sample, one Raman map, comprising 20 spectra, was acquired and the average was taken for all spectra, excluding those which only showed the signal for the glass substrate (control, Fig. 4m). The spectra shown for each strain or isolate, comprise the average of the spectra of three independent replicates (i.e., up to 60 individual spectra).

The melanin in the supernatant of selected cultures was furthermore imaged using a scanning electron microscope (SEM, SEM 460 and Supra 40, field emission, Zeiss). Generally, filtered ( < 20 µm), non-acidified supernatant of cultures (BH with carbon steel or MEB) was diluted twenty times and dried on a silicon wafer. Only the more conductive border regions of each drop were analysed as the centre was too sensitive to beam damage.

Study of the biofilm-steel interface by ESEM, EDX and Raman

High resolution images of the strains grown on the vertical coupons were obtained using an XL30 environmental scanning electron microscope (ESEM) equipped with a tungsten cathode and a secondary electron detector (FEI/Thermo Fisher Scientific). Energy-dispersive X-ray spectroscopy (EDX) for the quantitative analysis of the elemental composition of mineral precipitates was undertaken with the modular EDX system Quantax 200 with an XFlash 6-60 silicon drift detector (Bruker Nano Analytics). For this purpose, coupons were taken out of the reactor and the areas above and below the nutrient solution-biodiesel-air interface were cut off. The remaining section of steel was prepared according to Spurr,82; the sample was fixated for at least 2 h using a 2.5% solution of glutaraldehyde in phosphate buffered saline (137 mM NaCl, 2.7 mM KCl, 10.1 mM Na2HPO4, 1.8 mM KH2PO4, pH set at 7.4), dehydrated via an ethanol dilution series (i.e., 30%, 50%, 70%, 80%, 90%, and ethanol abs.) and finally dried using critical point drying (Leica EM CPD300). All samples were coated with 30 nm gold. Both, ESEM and EDX investigations, were performed in the “high-vacuum mode” of the microscope. The corresponding EDX spectra were collected at an accelerating voltage of 20 keV and the quantification was performed standardless. Due to the uneven sample surface morphology the elemental composition obtained must be considered as a rough estimate. Quantitative comparisons of the results for different samples obtained under the same experimental conditions are more realistic than the absolute values.

Additional Raman spectra were obtained from the larger mineral precipitates on the carbon steel surface incubated with BAM 1071. These were analysed as describe above but only via a single spectrum (i.e. not via a map).

Statistical analysis

As the sample size was too low to test for normality, we showed all data as the median of three to four independent replicates with the first and third quartile (as a box plot). Differences between these medians were analysed via the non-parametric Kruskal Wallis test. The Conover-Iman post-hoc test with Benjamini-Hochberg p-value adjustment was applied (alpha level = 0.05) to analyse the differences between specific medians. For both the conover.test package (v1.1.6) in R software (v4.5.1; R core Team83) was used.