Enhanced performance of the microalga Chlorella sorokiniana remotely induced by the plant growth-promoting bacteria Azospirillum brasilense and Bacillus pumilus

Remote effects (occurring without physical contact) of two plant growth-promoting bacteria (PGPB) Azospirillum brasilense Cd and Bacilus pumilus ES4 on growth of the green microalga Chlorella sorokiniana UTEX 2714 were studied. The two PGPB remotely enhanced the growth of the microalga, up to six-fold, and its cell volume by about three-fold. In addition to phenotypic changes, both bacteria remotely induced increases in the amounts of total lipids, total carbohydrates, and chlorophyll a in the cells of the microalga, indicating an alteration of the microalga’s physiology. The two bacteria produced large amounts of volatile compounds, including CO2, and the known plant growth-promoting volatile 2,3-butanediol and acetoin. Several other volatiles having biological functions in other organisms, as well as numerous volatile compounds with undefined biological roles, were detected. Together, these bacteria-derived volatiles can positively affect growth and metabolic parameters in green microalgae without physical attachment of the bacteria to the microalgae. This is a new paradigm on how PGPB promote growth of microalgae which may serve to improve performance of Chlorella spp. for biotechnological applications.


Results
Promotion of algal growth by airborne microbial volatiles from PGPB. Both PGPB significantly enhanced growth of C. sorokiniana remotely up to six-fold, relative to the controls, in four days (Fig. 1a). Initial enhancement during the first three days was higher with A. brasilense; after four days, enhancement of population growth with B. pumilus was higher (Fig. 1a). Reducing CO 2 in the atmosphere of the Kitasato flask by incorporating a filter made of lithium hydroxide completely stopped growth of the microalgae over short and long time intervals, as expected (Fig. 1b,c). However, with the two PGPB located remotely and also when adding the control bacterium Escherichia coli to the experimental setup, promotion of growth occurred when the incubation period was extended. Cell volume increases were also observed in algal cells remotely connected to both PGPB, whereas microalgae cultured alone did not increase in cell volume (Compare Fig. 1a and Fig. 1d).
Production of CO 2 and mVOC by PGPB. Both PGPB grew well in Brunner's medium (Fig. 1e), and both bacteria increased the pH of their growth medium in a similar manner (Fig. 1f). After two days in Brunner's medium, both PGPB produced significant amounts of CO 2 , as expected, which accumulated in the headspace at concentrations significantly higher than atmospheric CO 2 ( Table 1). B. pumilus produced a larger amount of CO 2 with fewer cells than A. brasilense (Table 1). E. coli is well known to produce significant amounts of CO 2 under these conditions 44 . SPME coupled to GC-MS analysis identified 47 volatiles produced by PGPB (Table 2). A representative gas chromatogram is shown in Suppl. Fig. 1, showing the differences in volatile emission patterns between strains of A. brasilense, B. pumilus, and E. coli. GC-MS analysis of mVOCs revealed consistent differences in the volatile blends released by specific strains of A. brasilense and B. pumilus, and to E. coli ( Table 2, mass spectra of major labeled peaks are presented as Supplementary Material, Fig. S4). In all cases, reported values represent the relative percentile of each percentage of total volatile to the total amount of volatiles detected with volatile peak area being first normalized to the amount of the spiked internal standard hexneyl acetate. No absolute measurements of individual of volatiles were made considering that no standard calibration curve was made for each target volatile with its standard. Such an approach has been reported from our work reporting plant volatiles analysis using headspace SPME 45 . Growth-promoting volatiles 2,3-butanediol and 3-hydroxy-2-butanone (also referred to as acetoin) were consistently major volatile components (75% and 62% of total volatile blend) produced by A. brasilense and B. pumilus, respectively, whereas these volatiles were released at much lower levels by E. coli (4%). 3-methyl-1-butanoic acid (isovaleric acid), 2-methyl-butyric acid, and 3-methyl-1-butanol (also referred to as fusel alcohol) were found only in A. brasilense, and B. pumilus. Several short-chain fatty acyl esters (octanoic, decanoic, and dodecanoic acid ethyl esters) were identified, though at much lower levels compared to short-chain alcohols in A. brasilense and B. pumilus volatile blends.
Scientific RepoRts | 7:41310 | DOI: 10.1038/srep41310 A remote effect induced by the PGPB and E. coli on accumulation of total lipids, carbohydrates, and chlorophyll a in C. sorokiniana. Total lipids were significantly enhanced by a remote effect from all three species of bacteria (two PGPB and the control E. coli), with B. pumilus inducing the highest effect. Lipids produced by microalgae growing alone were below the detection limit of the method used (Fig. 2a). Similar results occurred for carbohydrate production by the microalgae, but in this case, A. brasilense induced the highest effect (Fig. 2b). All bacterial species enhanced the quantity of chlorophyll a, relative to the level of chlorophyll a in microalgae growing alone (Fig. 2c).

Discussion
Culturing microalgae in the presence of PGPB is an artificially created experimental situation. For biotechnological purposes, any beneficial interaction occurring under these artificial conditions would result in an economic gain and therefore be worth exploring. The main purpose of this proof-of-concept study was to show that the enhancement of performance of microalgae, by volatiles of PGPB, is an effective strategy for common biotechnological applications involving cellular growth and metabolite production. As our results demonstrate, The control bacterium (E. coli) was used only in experiments measuring the potential effect of produced CO 2 on microalgae growth and metabolite production. Values on curves denoted by different capital letters differ significantly using one-way ANOVA combined with LSD post-hoc analysis at P < 0.05. Points at each time interval denoted by different lowercase letters differ significantly at P < 0.05 in (a-d, f) using ANOVA and in (e) using Student's t-test).     Table 2. Relative quantification of volatiles expressed as relative percentile and biological function of the volatile compounds produced by the three bacterial species along with their reported biological functions.
Compounds denoted by an asterisk were confirmed with authentic standard.
the enhancement of algal production by PGPB can be accomplished without attachment of the PGPB to plant surfaces (as normally occurs when PGPB is inoculated on plants). The significant positive effects of both A. brasilense and B. pumilus on general growth and metabolism of Chlorella spp. when the microorganisms are in forced close proximity inside an alginate bead is well documented (see Introduction for references). At least one genus of PGPB, Bacillus, was known to produce a variety of volatiles 27 . Both PGPB are expected to produce CO 2 during normal aerobic respiration, the growth conditions used in this study. Consequently, it is possible that the effects of volatiles will affect performance of microalgae.
The design of the experiments presented here ensures a lack of physical contact between any PGPB and the microalgae. Supplying cultures of microalgae and other plants with CO 2 to increase growth is a commonly used laboratory and even industrial technique 46 . The growth promotion of Arabidopsis thaliana by the PGPB Serratia odorifera (produces quantities of volatiles) was partly attributed to enrichment with CO 2 30 . Consequently, the significant improvement in growth from exposure of C. sorokiniana to volatiles of both PGPB and E. coli could also reasonably be attributed to the effects of CO 2 produced by the bacteria.
The relative importance of CO 2 in growth promotion in the experiments reported here was evaluated by reducing the quantity of CO 2 in the headspaces of the experimental flasks using lithium hydroxide. Removal of CO 2 in the headspace completely inhibited autotrophic growth of C. sorokiniana growing alone, but not when a PGPB was present in the adjacent flask. Under conditions of physical separation, growth of the microalgae was initially delayed but resumed after a few days. We suggest that this happened as the concentrations of CO 2 produced by the bacteria increased and the filter could not absorb all the CO 2 that was continuously produced by the PGPB. Alternatively, this was also assisted by the organic volatiles produced by the PGPB. Finally, CO 2 can affect the pH of the microalgae substrate. CO 2 that accumulated in the headspace of the culture can be incorporated as carbonate in the medium and consequently raise the pH to more neutral values than the initial pH, as shown in this study. Extremely low pH negatively affects growth of Chlorella spp. 19 . Reduction of O 2 partial pressure in the enclosed flasks is less likely. Although theoretically this can inhibit photosynthesis and growth of microalgae, these parameters were enhanced in our system.
We have also found that the enteric bacteria E. coli had similar growth promotion effects on C. sorokiniana as both PGPB. While co-culturing E. coli with C. minutissima yielded faster growth and cell density in the microalgae 47 and a surprising phytostimulation of maize by E. coli was also reported 48 , so far, no solid explanation for these effects have been provided. Consequently, we assumed that the growth effect can be partly attributed to production of large quantities of CO 2 , as well known for E. coli. It seems unlikely that the results of these experiments can be explained as simply due to the effects of CO 2 enhancement, however, as B. pumilus had the highest CO 2 production, but did not produce the highest growth enhancement of the microalgae during the first 72 h.
The literature on organic volatile-mediated bacterial-plant interactions (see Introduction for references) suggests that production of growth-promoting volatiles is a widespread phenomenon among rhizosphere bacteria. The presumptive effect of volatiles of A. brasilense was predicted 34 , but neither determined nor quantified. In contrast, the presence of Bacillus sp. volatiles and their effect on plant growth are well documented 28,29,49,50 . Both PGPB tested in our study produced a large array of volatiles, some of which are known for growth promotion, such as 2,3-butanediol and acetoin. Their role as plant growth-promoting volatile determinants has even been proven through exogenous application to plants 29 . The abundance of these two key volatiles in A. brasilense and B. pumilus volatile blends was far higher than what was previously reported for B. subtilis and B. amyloliquefaciens 28 . This comparison requires that a reservation be added. A direct comparison cannot be made because different volatile collection methods were used in these studies, namely dynamic versus static SPME headspace sampling. Additionally, using SPME fiber coatings in our study may have limited sensitivity by preferentially absorbing or excluding particular volatiles, based on polarity or size. For example, divnylbenzene/carboxen/PDMS fibers favor short-chain polar compounds such as 2,3 butanediol 51 .
In this study, as in others involving plant-volatile interactions, the biological functions for several identified volatiles were not established 26,52 . Some volatiles (nonanoic acid, indole, nonanal, isovaleric acid, and pentanoic acid) affect other organisms, whereas other detected mVOCs have no currently known function. In view of the chemical complexity and diversity of mVOCs, assessment of many of these compounds as relevant growth enhancers remains to be done 27,52 .
Both CO 2 and organic volatiles produced by PGPB were measured in this study. CO 2 serves as a natural substrate for growth of photosynthetic microalgae and organic volatiles have been shown to promote the growth of several plants. Our experimental design, while demonstrating that these volatiles did affect growth and metabolism of C. sorokiniana, could not distinguish between the relative contribution of the compounds. The extremely rapid multiplication of microalgae (seen after removal of CO 2 by the lithium oxide filter followed by replenishment of CO 2 by the remote presence of the PGPB) could be attributed to a starvation effect for carbon in the microalgae. Similar responses in microalgal growth rate are known after being deprived of nitrogen and phosphorus 53 .
Cells of A. brasilense attached to C. vulgaris and C. sorokiniana significantly promote accumulation of starch 15,16 , lipids 18 , fatty acids 17 , and pigments 18 in the microalgae. Volatiles of B. subtilis and E. coli promote accumulation of starch in Arabidopsis 41 . As a result of the minimal mineral culture medium we used, the absolute quantity of these metabolites is lower than quantities produced when the microalgae are growing in rich medium. Still, our study clearly demonstrated that an enhancement of the production of key metabolites can be observed in the absence of any physical attachment of the PGPB to the microalgae cells. This provides insights into a novel mechanism in microalgae-bacteria interactions. This phenomenon may be complementary to the established paradigm of attachment of PGPB to plant surfaces, but can only be investigated under special culturing conditions.
The biotechnological ramifications of this study need to be explored further. Both PGPB used here are produced on a large industrial scale as agricultural inoculants 54 . E. coli is also used for many biotechnological applications 47 . During culturing, volatiles from these bacteria are diffused to the air and lost. The results presented here suggest a potential alternative biotechnological option. If these volatiles can be captured, they can be used to promote the performance of Chlorella, which is the most common economically produced genus of microalgae.

Conclusion
This study demonstrated that diffusion of CO 2 and organic volatile compounds produced by two PGPB promote cell growth and metabolite content in the microalgae C. sorokiniana without physical contact.

Materials and Methods
Microorganisms and initial growth conditions. The 56 . This microalgal strain can grow at pH 5 and higher. In experiments that measured the potential effect of CO 2 , Escherichia coli DH5α (Invitrogen, Carlsbad, CA) served as the negative control for microalgal growth and promoting metabolites because it does not have any plant growth-promoting effects; it also served as a positive control in the CO 2 experiment because it produces CO 2 , as any E. coli. All cultures were tested under sterile conditions.
For initial culturing of the microalga, 10 mL axenic C. sorokiniana culture, cultivated in sterile mineral medium (C30), was added to a sterile flask containing 90 mL sterile C30 medium, composed of (in g·L  (for A. brasilense). Cells were grown for 48 h at 32 ± 2 °C and 120 rpm. After incubation, the bacterial cultures were rinsed, as described above, diluted in saline solution, and subsequently served as the inoculum for 250 mL flask, each containing 2 × 10 9 CFU·mL −1 inoculum concentration. C. sorokiniana was inoculated into the same flasks using 10 mL of a suspension containing 8 × 10 6 cells·mL −1 .
Experimental design and culturing conditions. All experiments were run in batch cultures in pairs of 250 mL sterile Kitasato (Büchner) flasks (Corning, Corning, NY) containing 100 mL medium in each flask, with each pair of flasks serving as a single experimental unit (Fig. 3). Contents of the flasks were inoculated with the respective microorganism, in the above mentioned media, hermetically sealed with a new rubber stopper, and incubated at 28 ± 1 °C for 96 h under illumination of 90 μ mol photon·m −2 ·s −1 on a rotary shaker at 120 rpm. Some experiments were incubated for up to 216 h. A lithium hydroxide filter (described below) was used in experiments designed to remove CO 2 . In all other experiments, the filter was not used.
Each experiment was performed with five replicates per treatment, where a pair of Kitasato flasks served as a single replicate. Each experiment contained the following treatments: A. brasilense and C. sorokiniana; B. pumilus and C. sorokiniana, C. sorokiniana and distilled water (as a control) and, in experiments involving potential production of CO 2 , C. sorokiniana and E. coli (as a positive control). Analytical methods. Determination of total carbohydrate content. Microalgal cells extracted by centrifugation were hydrolyzed with acid for 60 min at 100 °C 16 to release carbohydrates. Quantification of carbohydrates was by the phenol-sulphuric acid method 58 , adapted to a microplate using glucose as standard.
Determination of total lipid content. Extraction of lipids followed the standard method 59 with small modifications to adapt it to microalgae, which involved sonication to disrupt cell walls 17 . Lipids were quantified in the range of 70 μ g to 1.33 mg permitted by this analytical method 60 .
Determination of chlorophyll a content. Extraction of chlorophyll a from cells used the method described by Youngman 61 , with minor modifications, where 5 mL freshly harvested culture were centrifuged for 10 min at 6000 × g. The supernatant was discarded and 5 mL 90% methanol was added to the pellet and heated in a water bath for 10 min at 60 °C. After cooling, the samples were incubated in the dark for 24 h at 4 °C. Then, the samples were centrifuged for 10 min at 4 °C at 6000 × g. Absorbance in the supernatant was recorded at 655 and 750 nm. To quantify chlorophyll a content, we used the equation 61 : Chlorophyll a (mg·L −1 ) = [13.9 (OD 655 -OD 750 ) × U] / V, where, U is the final volume of methanol and V is the volume of the sample.
Determination of production of bacterial CO 2 and reduction of CO 2 in the headspace of the flasks. Bacterial species were cultivated on modified Brunner's medium, as described earlier, for 48 h at 30 °C in sealed serum bottles. Concentrations of headspace CO 2 were quantified by gas chromatography (model 8610 C, SRI Instruments, Torrance, CA) equipped with methanizer. Briefly, the CO 2 in 100 μ L injections of samples was converted to methane via the methanizer (a device for the high temperature reduction of CO 2 to methane in the presence of a catalyst) held at 380 °C, with the methane subsequently passed via a 1 m silica gel column held at 80 °C and detected with a flame ionization detector. The gas chromatograph was calibrated using CO 2 standards (Matheson Tri-Gas, Basking Ridge, NJ). Reduction of quantities of CO 2 in the headspace of the flasks was accomplished by incorporating a UV-sterilized (lithium hydroxide plus water) filter, which strongly adsorbs CO 2 (12) at quantities of 0.3 or 0.5 g per filter into culture flasks.
Analysis of microbial volatile organic compounds (mVOCs). Bacteria cultures were inoculated by pipetting 10 μ L glycerol stock prepared in tryptic soy agar (TSA) containing 20% glycerol in 50 mL MS liquid medium 62 containing 1.5% (w/v) sucrose, 0.4% (w/v) TSA and kept for 24 h at 37 °C. For volatile capture, 5 mL aliquots of the broth were placed in a 20 mL solid-phase microextraction (SPME) vials in a laminar flow hood. Then, 10 μ L of ultra-pure (Resistivity: 18.2 MΩ·cm at 25 °C) water containing 1 μ g cis-3-hexeneyl acetate (Sigma-Aldrich, St. Louis, MO) as the internal standard was added and the vials sealed withTeflon ® -lined magnetic caps using a hand crimper to prevent the escape of volatiles. SPME and gas chromatography-mass spectrometry analysis of the volatiles were performed as detailed below 49 . Earlier studies, in which the volatile composition of media alone was determined using the same analytical approach, showed a very negligible background (i.e., Fig. 2 in Ryu et al. 29 ).
To capture the collected volatiles, a 50/30 μ m DVB-CAR-PDMS SPME fiber (57328-U, Supelco, Bellefonte, PA) was inserted into the headspace above the bacterial culture and the vials were placed in a temperature-controlled oven at 50 °C. Heating is essential and used in all SPME methods to ensure equilibration and saturation of volatiles in the headspace. Adsorption of volatiles was performed for 30 min, and fibers were desorbed at 210 °C for 1 min in the injection port of a gas chromatograph interfaced with a mass spectrometer (GC-17A GC and QP-5000 MS, Shimadzu, Kyoto, Japan).
Volatiles were separated on a DB5-MS capillary column (30 m length, 0.25 mm inner diameter, (J&W Scientific/Agilent Technologies, Santa Clara, CA). Injections were made in the split-less mode for 30 s. The gas chromatograph was operated under the following conditions: injector 220 °C, column oven 36 °C for 3 min, then programmed at a rate of 12 °C min −1 to 180 °C, kept at 180 °C for 5 min, and then increased by 40 °C·min −1 to 220 °C and held for 2 min. Helium carrier gas was injected at 1 mL·min −1 . The transfer line and ion-source temperatures were adjusted at 230 °C and 180 °C, respectively. The quadrupole mass spectrometer (QP-5000 MS, Agilent Technologies) was operated in the electron ionization mode at 70 eV. The scan range was set at 40-500 m/z (mass-to-charge ratio). Volatile components were identified using the procedure described in Farag and Wessjohann 63 and peaks were first deconvoluted using AMDIS software (www.amdis.net) and identified by its Kovat retention indices (RI) relative to n-alkanes (C 6 -C 20 ) in the NIST/EPA/NIH mass spectral library database and with volatile standards, when available. The Kovat index refers to relative retention time measurements comparing hydrocarbon stand mixture C 8 -C 20 to allow comparisons among databases. Statistical analysis. Each experiment was repeated at least twice. Results presented are the average of two or three experiments, in each case. All data was analyzed by ANOVA, employing Fisher's post-hoc analysis at p < 0.05 in Statistica 8.0 software (StatSoft, Tulsa, OK).
Considering the complexity of GC-MS data, multivariate data analyses are commonly used to detect compositional differences between species and help identify potential chemical markers for discrimination in an untargeted manner. Principal component analysis (PCA) is the most commonly used unsupervised multivariate data analysis method. Models allow clustering of samples according to intrinsic variance between them and without being biased by desired outcomes 64 . Multivariate data analysis of mVOCs was done by PCA and orthogonal partial least squares-discriminant analysis (OPLS-DA), performed using the program SIMCA-P 13.0 (MKS Umetrics, Malmö, Sweden). The PCA was run to obtain a general overview of the variance of volatile metabolites, and OPLS-DA was performed to obtain information on differences in the composition of volatiles between A. brasilense and B. pumilus strains. The Distance to the Model (DModX) test was used to verify the presence of outliers and evaluate whether a sample fell within the model applicability domain.