3.1. Effect of Aerobic Treatment on Chemical Oxygen Demand (COD)
The effect of the treatment of DSW on COD was found to be highly dependent of the strains used for aerobic treatment (
Table 1). Aerobic fermentation of DSW by
Phanerochaete chrysosporium,
Flavodon flavus,
Fusarium proliferatum and
Gibberella fujikuroi appeared to be less efficient strains for COD reduction, with 23.5%, 28%, 34% and 38%, respectively, whereas
Aspergillus terreus var africanus,
A. parasiticus,
Trametes hirsuta,
T. versicolor and
A. terreus var. terreus showed the highest decrease in COD (76.53%, 74.60%, 74.01%, 73.64% and 73.5%, respectively). Notably, the 9
Aspergillus and anamorph strains used in this study were among the most effective strains for COD reduction (COD reduction was higher than 65% for all 8
Aspergillus strains and 58.65% for
Fennellia flavipes), indicating that these strains are particularly interesting for their reduction of the pollution load of DSW. COD reduction by
Pichia jadinii and
Penicillium sp. could reach 40.91% and circa 62%, respectively.
Some of our results were consistent with other published works. For instance, Gonzalez et al. (2000) reported a high COD reduction (62%) on diluted molasses spent wash treated by
Trametes spp. [
14]. Benito et al. (1997) also found that
T. versicolor was able to reduce COD by more than 70% on supplemented sugar beet molasses [
40]. Similarly, a reduction of 46% and 65% of COD was found for
P. jadinii and
Penicillium sp., respectively [
30,
41]. Aerobic treatment of cane molasses stillage with
A. niger and
A. oryzae led to a COD reduction of up to 78% and 88%, respectively [
31,
32,
33]. On the contrary, Garcia et al. (1997) found that
A. terreus lowered the COD of DSW by only 29% [
34]. Surprisingly, our results using
P. chrysosporium and
F. flavus were found to be well below the observed values from the literature with, respectively, a 73% COD reduction on DSW supplemented with yeast extract and 80% on diluted DSW [
42,
43]. This difference may be explained by the fact that these studies were carried out on diluted and supplemented DSW, while we used crude DSW in our study. Moreover, COD reduction is generally concomitant with the discoloration of the vinasse. In their study, Fahy and collaborators (1997) showed that a sugar addition in the medium could significantly improve the depollution rate of vinasse by
P. chrysosporium [
44]. From these results, the efficiency of the strains to reduce COD is strongly dependent of the origin of the vinasse used (beet or cane for example) and their complementation with other sources of nutriments.
3.2. Effect of Aerobic Treatment on Colour
The effect of aerobic treatment of DSW on color was studied using optical density of DSW supernatant at 475 nm [
19]. Consistent with the COD reduction, we found that the strains showing the highest reduction of color belong to
Aspergillus and
Trametes genus (
Table 1). For instance,
Aspergillus parasiticus,
A. alucateus,
A. terreus var. terreus and
A. itaconicus led to a decrease of OD
475 nm up to 42.46%, 41.88%, 38.84% and 35.36%, respectively (
Table 1). Similarly
Trametes hirsuta and
T. versicolor reached up to 42.46 and 32.46% of decolourisation of DSW. DSW treatment with
A. flavus, A. niger,
A. oryzae,
Fenellia flavipes,
Flavodon flavus and
Phanerochaete chrysosporium also led to decolourization of DSW but to a lesser extent (OD
475 nm reduction was comprised of between 20 and 27%). Surprisingly, we observed that aerobic treatment of DSW by the yeasts (belonging to
Candida,
Clavispora,
Cryptococcus,
Galactomyces,
Issatchenkia,
Komagatella,
Pichia and
Saccharomyces genus) and by
Fusarium sporotrichoides and
Penicillium verrucosum resulted in small to high increase of OD
475 nm. The most important increases of colourization were obtained for
Saccharomyces cerevisiae (90.91%),
P. verrucosum (68.99%),
C. dubliniensis (52.53%),
P. jadinii (41.08%),
F. sporotrichoides (40%),
Issatchenkia orientalis (39.73%),
C. tropicalis (38.72%),
P. guilliermondii (38.38%),
Pseudozyma antarctica (36.03%),
Cryptococcus albidus (35.35%) and
C. glabrata (30.64%). The other yeasts species (
C. albicans,
P. angusta,
Komagatella pastoris,
Galactomyces geotrichum,
Rhizopus microsporus. var oligosporus and
Thanatephorus cucumeris) showed only limited colorization of the broth (less than 18%). A study has already noticed the increase of color after treatment. Kumar and collaborators (1998) reported that the optimum discoloration was closely related to the optimal growth and that the overall discoloration was obtained in the pH range of between five and eight, whereas at extreme pH levels, an increase in color was observed [
42]. We can then hypothesize that the coloration observed in this study is probably due to the high final pH reached at the end of the process (
Table 1). Finally, Kumar and collaborators (1998) reported that optimal discoloration was closely related to optimal growth and that overall discoloration was obtained in the pH range of five to eight, while at extreme pH levels, an increase in color was observed [
42].
With respect to the color of DSW treated with
A. niger,
F. flavus,
T. versicolor and
P. chrysosporium, our results showed a lower impact, as compared to the literature. One of the most studied fungi for potential decolourization of distillery effluent was
Aspergillus sps.
Aspergillus fumigatus G-2-6,
Aspergillus niger,
A. niveus,
A. fumigatus UB260 had an average of 55–79% decolourization [
45,
46,
47,
48,
49,
50]. Miranda et al. (1996) showed that, under optimal nutrient concentrations, aerobic treatment using
A. niger allowed for a decolourization of beet molasses by 69%. Furthermore, they reported that 83% of the total color removed was eliminated biologically and 17% by adsorption on the mycelium [
47]. Under optimal pH, Patil and collaborators (2003) showed that a melanoidin solution was decolourized from 60% to 72% by
A. niger immobilized cells [
51]. Raghukumar et al. (2001) reported that a diluted cane molasses stillage treated with
F. flavus could reach up to 80% decolourization [
43]. Further, aerobic treatment of a diluted molasses spent wash by
T. versicolor had a decolourization yield of 53% [
52]. When beet molasses were used, the decolourization yielded 58–81% OD
475 nm reduction. From 53.5 to 80% of decolourization of supplemented molasses spent wash treated by
P. chrysosporium was reported [
40,
42]. Moreover, Fahy and coworkers (1997) demonstrated that the further addition of a carbon source like glucose in a 6.25% molasses spent wash medium strongly enhanced the decolourization yield from 49 to 80% by
P. chrysosporium [
44].
Some of our results were somewhat contradictory with other published works. For instance, a study showed that
C. tropicalis could reach 75% decolourization level of a supplemented molasses spent wash when incubated at 45 °C [
19]. Likewise, treatment of distillery spent wash with the ascomycetes of
Penicillium genus resulted in about 50% reduction of the color [
46]. With reference to
Thanatephorus cucumeris (Rhizoctonia sp. D-90), Sirianuntapiboon and coworkers (1995) reported the decolourization of a melanoidin medium (molasses) by 87.5% thanks to an absorption mechanism. Indeed, the pigments were accumulated in cytoplasm and around the cell membrane before their degradation by intracellular enzymes [
53]. To the best of our knowledge, no studies have focused on the decolourization of DSW by
Galactomyces geotrichum,
Rhizopus microsporus,
Giberella fujikuroi and
Fusarium sp. Notwithstanding this, considering their use for molasses decolourization,
Galactomyces geotrichum and
Rhizopus microsporus var. oligosporus could achieve a color reduction of diluted molasses of up to 87% and 38%, respectively [
36]. Similarly, Seyis and Subasioglu (2009) showed that molasses decolourization by
Gibberella fujikuroi and Fusarium species were not successful [
54]. The OD
475 nm increase could result from pigments repolymerization, from a higher rate of nutriment consumption and from production by the microorganism of molecules that also absorb at this wavelength [
55,
56,
57].
3.3. Effect of Aerobic Treatment on pH
Compared to the initial pH of the DSW broth (in the range of 4.77–4.95), all microbial treatments of crude DSW led to a significant increase of final pH (
Table 1). Alkalinisation of the medium may be the result of an ammonium release during the assimilation of nitrogen source like proteins for the microorganism growth or a consumption of organic acids or reducing sugar present in DSW [
55]. Among the 37 strains tested in this study, 22 could achieve a pH final value above 8 units. Among the best alkalinising strains, maximum pH (>9 units) was reached for DSW incubated with
A. terreus var. africanus (9.05),
P. verrucosum (9.03) and
A. terreus var. terreus (9.0). More generally, among the
Aspergillus and anamorphs genera, seven strains were found to reach a pH of above 8.3 units.
Several studies have shown that the degradation of melanoidins, which is related to discoloration, tends to increase with alkaline pH. For instance, Hayase and collaborators (1984) reported that the discoloration of melanoidin occurred more rapidly at alkaline pH than at acidic or neutral pH and could reach up to 94% discoloration at pH 10 [
58]. In addition, Mohana and coworkers (2007) reported that melanoidins are less soluble in acidic rather than in alkaline pHs and that pHs less than or greater than 7 units lead to a decrease of discoloration activity [
59]. Similarly, Agarwal and collaborators (2010) claimed that melanoidins were more soluble at alkaline pH [
60].
Contrary to these studies, we found no specific link between pH and (dis)colorisation of DSW was shown (see
Table 1). Indeed, DSW aerobic fermentations using
A. terreus var. terreus and
Penicillium verrucosum led, in both cases, to an alkalinisation of the supernatant pH of DSW up to 9 units, but in the first case, an OD
475 nm decrease of 38.84% could be noticed, whereas an OD
475 nm increase of 68.99% was observed in the second case. Likewise,
A. oryzae and
F. flavus induced a decolourization of DSW by about 22%, but an alkalinisation of pH of 8.86 and 6.17, respectively.
As few sugar remain in residues like sugarcane molasses after sugar fabrication, the ethanol production from these residues conduced the use of harsher processing steps to depolymerize the structural polysaccharides. These processes result in side reaction products and in the acidification of the medium that are potentially inhibitory to microbial growth. Therefore, anaerobic digestion of the vinasse produced from sugarcane molasses may be fraught with problems [
61]. As aerobic fermentation of DSW by yeasts and filamentous fungi bring about alkalinisation of DSW, the anaerobic digestion of the latter could be improved.
3.4. Biomass Production and Mineral Content of DSW after Aerobic Treatment
The biomass production of the 37 yeasts and filamentous fungi strains was measured during growth on crude DSW (
Table 1). Microorganisms that presented the best production of biomass during aerobic treatment of DSW were
Trametes hirsuta (29.40 g·L
−1),
A. terreus var. africanus (29.19 g·L
−1),
Clavispora lusitaniae (28.56 g·L
−1),
T. versicolor (25.96 g·L
−1) and
Issatchenkia orientalis (25.41 g·L
−1). In the same way as COD, OD
475nm and pH, we again found that the
Aspergillus genus was particularly efficient in biomass production on crude DSW. The 9
Aspergillus anamorphs strains showed that biomass productions, after 10 days incubation, were comprised of between 17.98 g·L
−1 (
Fennellia flavipes) and 29.19 g·L
−1 (
Aspergillus terreus var africanus). Smaller amounts of biomass were observed for aerobic fermentation of crude DSW by the yeasts such as
P. jadinii and
S. cerevisiae (14.67 and 11.67 g·L
−1, respectively).
Several studies have concluded that the COD reduction and/or decolourisation of diluted and/or supplemented molasses spent wash from sugarcane or sugar beet feedstocks by strains of
Aspergillus,
Penicillium,
Candida and
Pichia genus was accompanied by a fungal growth on the medium [
62]. Biomass productions in DSW treated by
Aspergillus and anamorphs strains were somewhat higher than those previously reported in literature. For instance, Rosalem and collaborators (1985) showed that biomass production of
Aspergillus niger grown on DSW could vary from 8 to 13 g·L
−1 [
32]. Likewise, cellular concentration of
Aspergillus oryzae grown on DSW were comprised between 12 and 17 g·L
−1 dry weight [
31,
33]. In their study, Rolz and collaborators (1975) also demonstrated that biomass production by
Penicillium sp. grown on DSW can reach up to 16 g·L
−1 [
30].
Data from the literature showed that
Issatchenkia orientalis incubated in DSW supplemented with molasses, MgSO
4, urea and H
3PO
4 could only produce a biomass of up to 8 g·L
−1 [
63]. The growth of
S. cerevisiae on molasses stillage reached a maximum biomass production of about 12.7 g·L
−1 [
64]. Similarly, growth of
P. jadinii on DSW supplemented on molasses produced from 9 to 18 g·L
−1 of dry biomass [
65]. Our results therefore clearly indicate that aerobic treatment of crude DSW by these filamentous fungi and yeast strains could achieve a significant reduction of polluting loads of DSW concomitantly with a high production of dry biomass (
Table 1) that could be further valuated into added value molecules. Unexpectedly, our study did not reveal a clear link between biomass production and COD reduction (
Table 1). This was particularly true for the strains that grow poorly on DSW (biomass production of
P. antarctica,
P. rugulosum,
P. angusta and
G. fujikuroi were comprised between 0.75 and 4.12 g·L
−1), but showed a significant decrease in COD ranging from 38% to 62%. This result indicated that the enzymatic process of the reduction of polluting loads could work independently of the process of using nutriments from DSW for growth.
We also noticed that aerobic treatment by the 37 strains used in this study always resulted in a significant reduction of mineral content of DSW (
Table 1). This decrease was considerable after treatment of DSW by
F. flavus (61.5%),
A. terreus var. africanus (66%),
A. oryzae (66.6%),
P. chysosporium (70.5%),
A. terreus var. terreus (72.4%),
A. alutaceus (73.5%) and
A. niger (77.6%). In agreement with our results for COD, OD
475 nm, pH and biomass production, we found that seven out of the nine
Aspergillus and anamorphs strains showed a mineral reduction in the broth by at least 50%. This result confirmed the high potential of
Aspergillus genus to efficiently reduce the polluting load of DSW concomitantly with a high valuable biomass production. Aerobic treatment conducted with
C. tropicalis (20.8%),
P. angusta (20.8%),
A. flavus (20.9%),
P. antarctica (22.1%),
C. glabrata (26.6%),
P. guilliermondii (28.2%),
C. lusitanea (28.4%) and
C. albicans (29.3%) led to a lesser, but significant decrease in mineral content. The growth of microorganisms is strongly dependent on micronutrients (such as iron, copper, manganese, zinc, and nickel) and macronutrients (like potassium, phosphorus, magnesium, nitrogen, sulphur, and calcium). These nutrients are involved in carbohydrate metabolism, amino-acids and vitamins production, Krebs cycle, nucleic acid production, pigments production and enzyme activities [
66,
67]. However, the absence of clear relationship between mineral content and biomass production may suggest that other phenomena are involved in the reduction of minerals in the media. For example, mineral content may decrease from precipitation as a consequence of DSW alkalinisation during aerobic treatment.
3.5. Statistical Relationships between Physico-Chemical Parameters
A Principal Component Analysis (PCA) was carried out to group the strains according to their performances on the physico-chemical parameters of DSW (biomass production and variations of pH, minerals content, COD and OD
475nm) and we investigated possible correlations between some of them. The Pearson correlation matrix showed that variables were moderately correlated between them (
Table 2).
For instance, we detected some correlations for pH and COD reduction (with a Pearson correlation coefficient
r of 0.508), reduction of minerals content and effect on OD
475 nm (
r = 0.503), biomass production with COD reduction on the one hand (
r = 0.466) and the effect on OD
475 nm on the other hand (
r = 0.447). Applied to the five original variables, the Cattell’s scree diagram [
68] highlighted three significant Principal Components (PC) explaining 84.89% of the total variance, 45.62% for PC
1, 26.23% for PC
2 and 13.81% for PC
3 (
Appendix A—
Table A1). The active coordinates retained by PCA were used to create
Figure 1A,B.
Principal Component Analysis was performed using XLSTAT (Addinsoft). Predicted groups were correlated to CAH clusters. Cluster 1, consisted in strains S2, S3, S4, S5, S6, S7, S8, S9, S36 and S37 which had the most significant COD decreases and biomass production. Cluster 2 includes strains S1, S10, S14, S15, S17, S21, S23, S24, S25, S26, S33 and S35 that had a COD decrease and biomass production yields less higher than ones of the strains of cluster 1. Cluster 3, consisted in strains S11, S12, S13, S16, S20, S27, S30, S31, S32 and S34, which significantly increase OD475 nm. The remaining strains (S18, S19, S22, S28 and S29) constituted the last group (Cluster 4) and had a less important effect on the final pH.
The eigenvectors of the covariance calculated enabled the defining of three PCs (
Table 3). Only the original variables, whose correlation values with the principal components were greater in absolute value than 0.5, were taken into account. The first axis PC
1 was representative of a global average level of the variables and strongly correlated with four of the five parameters (
Appendix A—
Table A2). These four parameters (COD reduction, biomass production, minerals content reduction and effect on OD
475 nm) contributed for 93.77% to PC
1 construction. Additionally, the variables (final pH and the effect on the OD value at 475 nm) are in absolute value the original variables best correlated with the PC2 axis. PC
2 axis (
Appendix A—
Table A2). It can be noted that PC
2 was mainly built by the pH and the effect on OD
475 nm variables, i.e., 83.53% of contribution to PC
2 construction. Surprisingly, we found that some fungal species, such as
P. antarctica (S32), had very little growth on DSW (0.75 g·L
−1) despite a high COD consumption and a significant increase of pH, while species like
P. chrysosporium (S28) showed significant biomass production (17 g·L
−1), concomitant with small pH increase (7.01) and moderate COD consumption (23%). Then the third axis PC
3 was built mainly on biomass production and minerals content reduction (86.56% of the PC
3 construction). The variable reduction of mineral content also greatly contributed to the construction of the PC
3 axis (
Appendix A—
Table A2). By opposition to
A. flavus (S3), which turned out to produce a high amount of biomass (19.77 g·L
−1), but a weak minerals consumption (20.88%),
P. rugulosum (S26) could consume a large amount of mineral content (56.26%) with very little growth on DSW (2.36 g·L
−1)
(Figure 1A). These results suggested that a part of the minerals was indeed used for fungal growth, while another part was precipitated due to the alkalinisation of the DSW.
PCA indicated that the strains could be classified into three to four groups. According to hierarchical cluster analysis (HCA), four groups of strains with close characteristics had been defined, explaining 64.62% of the total inter-variance and 35.38% of the total intra-variance (
Figure 2). The distribution of the clusters according PC
1 and PC
2 (
Figure 1A,B) allowed us to define the common characteristics of strains belonging to the same cluster (
Appendix A—
Table A3). Cluster 1 including the 8
Aspergillus anamorphs strains and the 2
Trametes spp. was characterized by aerobic treatment resulting in both high biomass production, high COD and mineral content reductions and a strong impact on OD
475 nm, resulting in significant decolourization. Cluster 3 included strains that led to a significant increase of OD
475 nm that could reach 190.9% in comparison to the OD
475 nm of crude DSW and conduced to the lesser mineral consumption. Cluster 4 consisted of strains whose effect on final pH was less important and that brought to a lesser biomass production and COD reduction. The final pH of DSW treated by strains defined in Clusters 1 and 3 were generally above pH = 8 whereas the pH of DSW treated by strains of Cluster 4 had a pH lower than seven. Cluster 2, which gathered all the other strains, was formed by strains that influenced COD and mineral contents and produced biomass on DSW, but less significantly than the strains of Cluster 1.
Automatic truncation based on entropy (dotted line) allowed identifying four consistent groups of fungi explaining 64.62% of the total inter-variance and 35.38% of the total intra-variance. Order of appearance of clusters (from top to down) was Cluster 2, Cluster 3, Cluster 4 and Cluster 1.