Co-cultivation of Mortierellaceae with Pseudomonas helmanticensis affects both their growth and volatilome

The growth behaviour of Linnemannia is strain-specific

Most strains showed comparable morphological characteristics on both media as well as in pure and co-culture. However, Linnemannia solitaria and Entomortierella galaxiae produced more aerial mycelium on PDA compared to LcA. There was more/less aerial mycelium in co-cultures with P. helmanticensis compared to pure cultures depending on the strain (Fig. 1, SI Fig. S3).

The comparison of Linnemannia and E. galaxiae daily radial growth rates did not support a difference between these genera (p ≥ 0.3). The overall linear model indicated that the fungal daily growth rates mainly differed among species (Table 1). In addition, the effect of strains highlighted the heterogeneity among strains within species (Fig. 2, SI Figs. S4, S5). Although there was no relevant main effect of medium on the daily radial growth rate of the fungi, the medium did affect the fungi in a strain-specific manner (Table 1, Fig. 2, SI Figs. S4, S5). On nutrient poor LcA, the fungal daily radial growth rates were reduced for all species, except for L. solitaria, which grew better on LcA (SI Figs. S3, S4).

The main effect of co-plating P. helmanticensis on radial growth rate was small, yet significant (0.7%, p << 0.001), and highly strain specific (Fig. 2, SI Table S2). Given that the growth rate refers to only the fungus, those strains with differences in growth rates between pure cultures and co-cultures can be considered affected by co-cultivation. On PDA, six strains grew faster (mean ± standard deviation = 1.1 ± 0.69 mm d⁻¹) and six strains grew slower (0.5 ± 0.21 mm d⁻¹) when co-cultivated with the bacterium. On LDA, four strains grew faster (0.7 ± 0.42 mm d⁻¹) and ten strains grew slower (0.9 ± 0.73 mm d⁻¹) when co-cultivated with the bacterium.

Although the geographical origin location of a strain may affect the growth of a strain, here there was no clear pattern. Nevertheless, not all species were isolated from all locations, thus strain heterogeneity might have masked a potential location effect. With regards to historic or co-evolutionary effects of bacterial association, those strains isolated with a bacterium associated were not affected more frequently by co-cultivation (SI Tables S2, S3). Their growth was also not more likely to be in- or decreased by co-cultivation. Similarly, the re-cultivation with P. helmanticensis did not affect the daily radial growth rate of those fungal strains originally isolated in association with this bacterium in a similar manner, i.e. the same direction (de-/increase).

The volatilome

We used two approaches to investigate the volatilome: On the one hand, we explored the effect of species, strains, cultivation mode (pure and co-culture), and time (Fig. 3) on the VOCs concentrations. As expected, the concentrations varied depending on all the experimental factors, mainly cultivation mode, but also species, strain, and partially on time. On the other hand, we asked which VOCs were produced/consumed by the fungus, P. helmanticensis and their co-cultures, respectively (Figs. 3, 4). All cultures both produced and consumed VOCs compared to the sterile PDA medium.

Stability over time

Over a period of four days, the volatilome of every sample (pure and co-culture) was measured four times using PTR-ToF-MS with similar time intervals between measurements and among samples. The rationale for this design was that we expected the volatilomes to be highly variable over time. Generally, timely changes were insignificant (50/113 VOCs) (Fig. 3). For those (63) VOCs that differed over time, with few exceptions, the time explained only a low percentage of variance (median = 5.4% variance, SI Fig. S6).

Fungal consumers and bacterial producers

Both the fungal and bacterial pure cultures as well as the PDA medium emitted VOCs (Fig. 3). PTR-ToF-MS is a very sensitive technique, thus it is not surprising that in all sample groups, several VOCs were detected in low concentrations. While all sample groups, including the PDA, produced a number of VOCs in high concentrations, the concentrations of some VOCs were lower in the cultures compared to PDA, thereby indicating VOC consumption. The consumption/production of VOCs strongly depended on Mortierellaceae strains (Fig. 4). There were several VOCs that were produced by some Mortierellaceae strains and consumed by others. The fungal strains frequently consumed and produced VOCs. P. helmanticensis, in contrast, rarely consumed, but produced a variety of VOCs. Using generalized linear modelling, we estimated the chance for VOC consumption over VOC production for the different strains and P. helmanticensis in pure culture. As the high number of low concentration VOCs might inflate the model, we applied different detection limits and specified several models. All models had a good fit (SI Table S4). Independent of threshold, the binomial model supports that all fungal species and strains were more likely to consume VOCs compared to P. helmanticensis, and that the bacterium was a better producer than any fungal species/strain (SI Figs. S7, S8). Compared to the mean concentration across all fungal strains, VOC concentrations were higher in P. helmanticensis cultures (p_pairedTTest << 0.001). The co-plating of P. helmanticensis and Mortierellaceae strains did not result in additional VOCs detected compared to all pure cultures (Fig. 4).

Without setting any threshold, only three VOCs (ms105.0299, ms121.0684, ms123.9460; SI Table S5) were consumed by P. helmanticensis. Two of them were produced by at least one Mortierellaceae strain in pure culture (Fig. 4), although only in low amounts. For a number of 50 VOCs produced by P. helmanticensis there was at least one Mortierellaceae strain able to consume it (SI Table S6). Those VOCs consumed by the fungi could be potential candidates for fungal-bacterial communication. We will address this aspect in subchapter “The volatilome and growth”.

The volatilome accurately predicts the taxonomy and cultivation mode

We used linear discriminant analysis (LDA) to predict the taxonomic affiliation (both strain and species) and the cultivation mode from the samples’ volatilomes. For each of the factors, including their combination, the prediction was very accurate (Fig. 5, SI Fig. S9a). The species, including P. helmanticensis, could be discriminated with an accuracy of 96% (185/193 samples in the test dataset). The strain was predicted with an accuracy of 84% (156/185 samples in the test dataset). For 94% (144/154) and 61% (103/170) of the test dataset samples, the LDA accurately predicted whether or not the species and strain were co-cultivated with P. helmanticensis, respectively.

We expected that in the LDA ordination, those strains with similar changes in growth behaviour following co-plating with P. helmanticensis would form a cluster or, along the LDs, shift (or not shift) in the same direction relative to their pure cultures (SI Fig. S9b–f). While such pattern was partially observed and pure cultures could be separated from co-cultures, especially across several LDs, changes in LD scores following co-cultivation were strain-specific.

The volatilome and growth

Models predicting the volatilome contribution of P. helmanticensis and Mortierellaceae strains in co-cultures

We applied general linear modelling of the volatilome data in order to (i) link the growth of both bacteria and fungi to the volatilome and (ii) identify individual VOCs that are potentially regulated in the interaction of fungi and P. helmanticensis. Using general linear modelling, we implicitly assumed that at least the concentrations of the majority of those VOCs—not every single VOC—were (i) linearly dependent on the cell density and were (ii) not regulated by the bacterial-fungal-interaction. We think this assumption can be hold, because it is unlikely that all VOCs produced/consumed are (i) regulated and (ii) regulated in this particular situation. Considering strain heterogeneity, models were calculated separately for each strain.

Such approach has advantages: (i) the correlation coefficients of the equations are an estimator for growth; and (ii) those VOCs not predicted well by these models violate the assumption of linearity and, therefore, might be regulated.

From our perspective, the most straightforward model would predict the co-culture volatilome from both pure culture volatilomes. However, this model has two disadvantages: First, it might be inaccurate for the bacterial coefficient as the fungal consumption might underestimate the bacterial production. Second, for some fungal strains their coefficients returned insignificant (Table 2), meaning that according to the model, the fungus did not contribute to the co-culture volatilome. As the fungus was visibly growing in all cultures (see also subchapter 3.1), this implies a violation of model assumptions for the models of those strains: if the model is correct, for those strains clearly the entire fungal volatilome was affected by the co-cultivation. Those strains with insignificant coefficients frequently consumed VOCs, but hardly produced VOCs. In other words, numerically their volatilome was characterized by negative concentrations of certain VOCs. P. helmanticensis, on the other hand, produced a high number of VOCs in high concentrations (Figs. 3, 4). Due to the presence of P. helmanticensis in co-culture, the co-culture VOC concentrations were positive, despite the fungal consumption. Consequently, the fungal contribution to the volatilome was negligible (insignificant coefficient) and the entire volatilome was predicted as a bacterial VOC production.

In order to overcome the disadvantages and weaknesses of this first model (model 1), we calculated two additional models. We compared these additional models to the first model and together all three models provide a conclusive representation of the co-plating experiment (Table 2). Model 2: For each combination of Mortierellaceae strain and P. helmanticensis, there was a number of VOCs that were neither produced nor consumed by the fungal strain in pure culture. Therefore, for each unique combination of Mortierellaceae strain and P. helmanticensis, we used those VOCs produced only by the bacterium to estimate the bacterial contribution in the co-culture. Model 3: We used model 2 for predicting the bacterial contribution to the co-culture volatilome. This predicted bacterial part of the co-culture volatilome was subtracted from the measured co-culture volatilome. Then, we used those VOCs produced by the respective fungal strain for predicting the theoretical fungal contribution to the co-culture volatilome from the fungal pure culture volatilome.

In the following subsections, we will present the results of the three models in more detail. As a proof of concept, we expected the bacterial coefficients predicted by both approaches to be correlated, and in line with visual observations and growth behaviour (3.1). A conserved mechanism involving VOCs in a potential interaction between Linnemannia/Entomortierella and P. helmanticensis, can be detected by showing a taxonomically conserved pattern of coefficients and VOCs repeatedly diverging from linear predictions across strains. In any case, the inductive effect of any VOC, conserved pattern, or speculated mechanisms needs to be tested in the lab.

Mortierellaceae strains limited the growth of P. helmanticensis

The bacterial coefficients of models 1 and 2 were numerically comparable and correlated (r = 0.92, p << 0.001; SI Fig. S10b). Thus, the intuitive model 1 predicting the co-culture volatilome from both pure culture volatilomes (P. helmanticensis and fungal strain), and the model 2 calculated only based on those VOCs produced by the bacterium and neither consumed nor produced by the fungal strain, supported the same conclusion. In both models 1 and 2, the bacterial coefficient was lower than 1, indicating that compared to the pure P. helmanticensis culture, the bacterial VOC production was reduced by the fungus (Table 2).

The bacterial coefficients of models 1 and 2 were higher for E. galaxiae than for all Linnemannia strains (model 2: mean_{Entomortierella} = 0.58, mean_Linnemannia = 0.20, p_TTest = 0.00027; comparable for model 1). Some (bacterial) coefficients from both models 1 and 2 were insignificant. In these co-cultivation experiments, the bacterial pure culture volatilome was an (ecologically and/or statistically) insignificant estimate for the volatilome differences observed between the pure and co-culture (Table 2). In other words, the bacterial contribution to the co-culture volatilome was close or indistinguishable from zero. These results were visually confirmed in culture: In those fungal strains with low coefficients, bacterial growth could neither be observed on the co-culture plate, nor under the microscope while observing a sample of fungal hyphae collected from the co-culture. For strains with very high coefficients, however, the bacterial growth was visible on the co-culture plate as well as under the microscope (data not shown).

The predicted coefficients for the bacterial volatilome contribution in co-culture were highly correlated with the centroid distances (r_{Spearman|model 1} = 0.77, p = 1.6e−5; SI Fig. S10d) observed between the pure fungal strains and their respective co-cultures.

Fungal VOC consumption exceeds their production

The fungal coefficients of models 1 and 3 were numerically comparable and correlated (r = 0.77, p = 1.2e−5, SI Fig. S10c) indicating that the intuitive model 1 predicting the co-culture volatilome from both pure culture volatilomes (P. helmanticensis and fungal strain) and the model 3 calculated only based on those VOCs produced by fungi supported the same conclusion. It should be noted that the concentrations of the VOCs produced by the fungal strains was often very low (across all strains median concentrations ranged between 0.064 and 0.338 ppbV). The centroid distances and fungal coefficients of both models 1 and 3 were not correlated (p_{Spearman|model 1} = 0.21, p_{Spearman|model 3} = 0.53, SI Fig. S10e). There was no correlation between the daily fungal growth rates and the predicted fungal coefficients (p_{Spearman|model 1} = 0.16, p_{Spearman|model 3} = 0.75).

Comparing the coefficients of models 1 and 3, respectively, the fungal coefficients were lower for E. galaxiae than for the Linnemannia strains (model 2: mean_{Entomortierella} = − 0.11, mean_Linnemannia = 0.49, p_TTest = 0.039; comparable for model 1). There were no differences among Linnemannia species.

Potentially regulated VOCs

For all three models 1–3 we plotted the predicted co-culture concentrations of VOCs against the measured ones (for an example, see SI Fig. S11). While VOC concentrations usually followed linearity and confirmed model fit, some VOCs were frequently identified as outliers. This is in line with our approach: The residual diagnostics confirmed our ecological model assumptions that the majority of VOCs would be linearly dependent, probably on cell density, and would not be regulated, and that the production of some VOCs in pure culture would differ from their production in co-culture, i.e. they might be regulated.

Across all models, 14 VOCs did not follow the models’ predictions (SI Table S7). Among these, six VOCs were frequently detected as outliers (7–21 out of 25); their concentrations were often species specific (Fig. 6). Across all species, the VOC tentatively annotated as formaldehyde (ms31.0182, [M + H]⁺) was predicted lower in concentration than it had been measured (Fig. 6a). The VOC (ms33.0340) tentatively annotated as methanol ([M + H]⁺) was predicted lower in concentration than it had been measured in L. exigua and E. galaxiae (Fig. 6b). For L. hyalina, the prediction was accurate. For the other species, the prediction was either too high or too low, depending on the strain. For L. sclerotiella and L. solitaria, both ethanol (ms48.0528) and acetic acid (ms61.0283, [M + H]⁺) were lower in concentration than predicted by the model, while the measured concentration exceeded the prediction for E. galaxiae (Fig. 6c, d, SI Fig. S12 a, b). The VOCs annotated as acetoin (ms90.0633) and phenylethanol (ms105.0717, [M−H₂O + H]⁺) were predicted higher in concentration than measured for all Linnemannia species. In contrast, for E. galaxiae the predicted concentration was usually lower compared to the measured concentration (Fig. 6e, f, SI Fig. S12c–f).

Source: Ecology - nature.com