Rational engineering of Saccharomycescerevisiae towards improved tolerance to multiple inhibitors in lignocellulose fermentations

Background The fermentation of lignocellulose hydrolysates to ethanol requires robust xylose-capable Saccharomyces cerevisiae strains able to operate in the presence of microbial inhibitory stresses. This study aimed at developing industrial S. cerevisiae strains with enhanced tolerance towards pretreatment-derived microbial inhibitors, by identifying novel gene combinations that confer resistance to multiple inhibitors (thus cumulative inhibitor resistance phenotype) with minimum impact on the xylose fermentation ability. The strategy consisted of multiple sequential delta-integrations of double-gene cassettes containing one gene conferring broad inhibitor tolerance (ARI1, PAD1 or TAL1) coupled with an inhibitor-specific gene (ADH6, FDH1 or ICT1). The performances of the transformants were compared with the parental strain in terms of biomass growth, ethanol yields and productivity, as well as detoxification capacities in a synthetic inhibitor cocktail, sugarcane bagasse hydrolysate as well as hardwood spent sulphite liquor. Results The first and second round of delta-integrated transformants exhibited a trade-off between biomass and ethanol yield. Transformants showed increased inhibitor resistance phenotypes relative to parental controls specifically in fermentations with concentrated spent sulphite liquors at 40% and 80% v/v concentrations in 2% SC media. Unexpectedly, the xylose fermentation capacity of the transformants was reduced compared to the parental control, but certain combinations of genes had a minor impact (e.g. TAL1 + FDH1). The TAL1 + ICT1 combination negatively impacted on both biomass growth and ethanol yield, which could be linked to the ICT1 protein increasing transformant susceptibility to weak acids and temperature due to cell membrane changes. Conclusions The integration of the selected genes was proven to increase tolerance to pretreatment inhibitors in synthetic or industrial hydrolysates, but they were limited to the fermentation of glucose. However, some gene combination sequences had a reduced impact on xylose conversion. Supplementary Information The online version contains supplementary material available at 10.1186/s13068-021-02021-w.

of lignocellulose biomass is the recalcitrant nature of the material to enzymatic conversion [3,4]. Physicochemical pre-treatment is thus required to disrupt the compact crystalline structure and allow enzymatic access to the polysaccharides within, to release fermentable sugars [3,5,6]. The majority of such pre-treatment methods result in significant quantities of degradation products being formed, which have inhibitory effects of subsequent biological conversions [5,[7][8][9][10][11].
Saccharomyces cerevisiae cannot naturally utilize xylose, the most abundant pentose sugar within lignocellulosic hydrolysates [12][13][14]. Although co-fermentation of glucose and xylose remains a challenge, advances in strain development have resulted in the development of industrial S. cerevisiae strains with xylose-utilizing capacity. Through metabolic engineering, heterologous xylose catabolic pathways such as the fungal oxidoreductive pathway (XR-XDH) [13,15,16] or a bacterial xylose isomerase (XI) [17,18] have been introduced into S. cerevisiae strains, as well combining both pathways into the yeast, simultaneously [14]. Recently, the industrial strain CelluX TM 1 [19], a xylose engineered strain with a XI pathway, has been developed.
Interestingly, an unanticipated phenotype that has emerged from xylose strain development is hyper-sensitivity of the introduced heterologous metabolic pathways to stressful conditions [2]. Due to the inter-connectivity between metabolism and stress response, strain development for lignocellulose bioconversion technologies have to simultaneously address both xylose utilization and microbial stresses [10]. Xylose engineered industrial strains are thus the ideal genetic background in which to study the impact of microbial stresses, as well as introducing stress resistance genes. Compared to the XR-XDH route, there is limited knowledge on XI pathway-based xylose utilization for genetically engineered industrial S. cerevisiae [20]. Furthermore, there is a lack of studies on the interaction of genes involved in inhibitor tolerance and xylose fermentation in these yeasts.
Fermentation strains are subjected to various microbial stresses during lignocellulose bioconversion that include microbial inhibitory compounds generated during physicochemical pretreatment of lignocelluloses. The concentrations of inhibitory compounds fluctuate depending on both biomass composition and pretreatment method used [9,11,21]. Furans are degradation products of sugars, phenolics are derived from solubilizing lignin and weak acids such as acetic and formic acids are formed during furan degradation and/or de-acetylation of hemicellulose [5,8,22,23]. Such microbial inhibitors negatively impact growth, fermentation and xylose utilization ability of yeast, which results in sub-optimal ethanol productivity and yields [9,10,13,16]. Given the toxicity of these compounds, studies have been undertaken to develop yeast strains capable of not only withstanding the harsh conditions associated with lignocellulosic biomass fermentations, but to also generate ethanol yields expected in industrial processes [5,22]. Thus, microbial inhibitor toxicity represents a bottle-neck in lignocellulosic bioethanol production and negating these detrimental inhibitory effects remains to be a fundamental challenge [22,24,25].
Inhibitor resistance is characterized as a complex function of multiple genes [10,[26][27][28][29], however, relatively few over-expression studies have undertaken to genetically engineer yeast towards multiple inhibitor tolerance phenotypes. Typical rational design strategies are often limited to only a few genes involved in highly specific in situ detoxification mechanisms, resulting in strains with inhibitor-specific detoxification phenotypes [30][31][32][33][34]. Examples include overexpression of either TAL1, FDH1, or HAA1 for weak acids resistance [30,35] and ADH6, ADH7 or ARI1 [36][37][38] for furfural detoxification mechanisms, as well as PAD1 and/or FDC1 that are linked to phenolic detoxification [33,34,39]. These studies have several shortcomings in that (i) the majority of overexpression studies are conducted in laboratory strains limiting applicability to industrial strains; (ii) strategies are limited by the available knowledge of the molecular genetic basis of resistance phenotypes gained from laboratory strains; (iii) studies make use of synthetic cocktails to simulate industrial stresses, and (iv) very few host strains are proficient xylose utilizers. Thus, more research is required to elucidate gene interaction or synergism within the dynamics of inhibitor resistance phenotypes in industrial strains towards industrial hydrolysates. A cumulative strain development strategy that combines multiple positive gene interactions from various stress response pathways could significantly enhance the yeast stress response towards multi-inhibitor resistance phenotypes. As a result, this study aims to combine inhibitor resistance pathways within a metabolic engineering approach towards the development of efficient multiinhibitor resistant xylose-utilizing strains.
This study explored the simultaneous overexpression of multiple native gene targets that confer resistance to weak acids, furan aldehydes and phenolic compounds individually, and thereby identify gene combinations that could generate cumulative multi-inhibitor resistance phenotypes. ARI1 (NADPH-dependent aldehyde reductase), PAD1 (flavin prenyltransferase) and TAL1 (transaldolase) were selected since these genes have been implicated in broader multi-inhibitor resistance phenotypes [32,39,40], whereas FPS1 (aquaglyceroporin), FDH1 (formate dehydrogenase 1), ADH6 (NADP-dependent alcohol dehydrogenase 6) and ICT1 (1-acylglycerol-3-phosphate O-acyltransferase) address specific effects attributed to weak acid [30,41,42], furan aldehyde [37] and organic solvent stresses [43], respectively. The partial deletion of FPS1 and inclusion of ICT1 allowed for the novel regulation of "membrane-modulating" genes into resistance phenotypes [44]. Dual gene combinations were constructed to have a multi-inhibitor resistance gene coupled with a target-specific gene in an FPS1 deletion background. Not only does this approach allow for the introduction of genes in novel sequential combinations, but also the assessment of these genes in industrial strains exposed to industrial-like lignocellulose fermentations. Ultimately, this improves the current understanding of in situ detoxification of lignocellulosederived inhibitors for the development of robust xylosecapable S. cerevisiae industrial strains. Table 1 lists some of the components determined in the liquid fraction of steam-exploded sugarcane bagasse (SCB), as well as SSL from MgO acid sulphite pulping process of mixed hardwood feedstocks. The pretreatment as well as sulphite pulping process resulted in the solubilization of the hemicellulosic fraction, but lignin was solubilized more intensely in the case of the sulphite process. Both liquors were composed of fermentable sugars and inhibitors such as weak acids (15.7 and 6.5 g L −1 for SSL and SCB hydrolysate, respectively), furans (2.3 and 2 g L −1 for SSL and SCB hydrolysate, respectively) and phenolics (2 g L −1 and 0.8 g L −1 for SSL and SCB hydrolysate, respectively). In terms of carbohydrates, the main sugar present was xylose with values of about 93 g L −1 and 8.7 g L −1 for the SSL and SCB hydrolysate, respectively. Literature reports xylose as the main sugar present in SSL liquors generated from hardwood feedstocks (HSSL) [11]. Also, SSL contained considerably more xylose than SCB hydrolysate, as well as significant concentrations of inhibitors, especially acetic acid (15.1 g L −1 , about 3 times more than SCB hydrolysate) and phenolic compounds (2 g L −1 , two times more than in steam explosion hydrolysate). Moreover, the SSL also has other compounds that can act as inhibitors such as SO 2 (used extensively in the wine industry to limit microbial contamination at lower pH values) and MgO (inhibits the growth of S. cerevisiae), making hardwood SSL a particularly challenging lignocellulosic feedstock [11].

Strain development in lignocellulose hydrolysate and inhibitor tolerance assays
The multi-inhibitor-resistant strain construction strategy centred on three rounds of sequential delta integration of double-gene expression cassettes to construct strains overexpressing selected inhibitor resistance genes in different combinations. Overexpression of respective double-gene cassettes was facilitated by homologous recombination of delta-integration cassettes into native delta sequences distributed in the parental CelluX TM 1 yeast genome. Transformants were screened for growth and ethanol yield in lignocellulosic hydrolysate after each round of integration to select the best strains for the next round of transformation, with final transformants assayed for inhibitor tolerance phenotypes.
Before the integration of inhibitor tolerance genes commenced, a partial FPS1 deletion variant of S. cerevisiae CelluX TM 1 was generated. Eight partial FPS1 deletion transformants were selected by screening for higher ethanol yield (g ethanol g −1 total sugar) on 2% YPD supplemented with 65% v/v sugarcane hydrolysate ( Table 1). The best performing transformant, CelluX1∆FPS1-C5, yielded 0.41 g g −1 at 169 h that showed a 5% ethanol yield increment to the parental CelluX TM 1 strain at 0.39 g g −1 (see Additional file 1: Tables S2, S3). Interestingly, strain CelluX1∆FPS1-C5 also exhibited an increase of 19.8% in formic acid detoxification and higher xylose consumption at 53.6% compared to 49.8% of the parental strain. This partial FPS1 deletion CelluX1∆FPS1-C5 strain was used as the host for the first round of delta integrations.
The results from the best transformant per gene combination from the first round of delta integration are listed in Table 2. The first round of transformants often exhibited a trade-off between growth and ethanol yield (see Additional file 2: Figure S1). The pBKD-AA integration cassette with the ARI1 and ADH6 genes that confer furan resistance, in combination with partial FPS1 deletion allowed for furan and weak acid resistance. Final ethanol yields of pBKD-AA transformants showed overall improvement that ranged from 0.34 to 0.38 g g −1 as compared to the parental CelluX TM 1 strain at 0.33 g g −1 , with a maximum increase in the ethanol yield of 15.8% ( Table 2). Partial deletion of the FPS1 gene in combination with pBKD-AA integrations proved beneficial to inhibitor resistance phenotype in terms of cell growth, as 80% of transformants displayed similar or increased growth (measured in absorbance) compared to the parental strain. The pBKD-AF integration cassette overexpressing the ARI1 and FDH1 genes also conferred furan and weak acid resistance. This configuration, however, significantly decreased the ethanol yield by 4.29-17.2%, compared to the parental strain, although an improvement in growth was seen with AF10, showing an increment of 7.92% over the parental strain. The pBKD-AI integration cassette overexpressing the ARI1 and ICT1 genes conferred furan, organic solvent and weak acid resistance however, 62.5% of transformants with this configuration exhibited decreased growth of 12-15%. Likewise, ethanol yields decreased by 1-7% showing improvement for only one strain (AI1) with an increment of 1.25% over the parental reference, respectively (see Additional file 2: Figure S1).
The pBZD-PA integration cassette overexpressing the PAD1 and ADH6 genes and this configuration conferred resistance to furans, phenolics and weak acids. After the first round of delta integrations, the ethanol yields of the transformants were lower by 2-24.6% or equal to that of the parental strain with six transformants showing improved growth by 1-18.8% growth increment (see Additional file 2: Figure S1). The pBZD-PF cassette integration overexpresses the PAD1 and FDH1 genes that confer resistance to phenolics and weak acids. These transformants also exhibited no increases in ethanol yields, but showed an increase in growth by 5-31.7% for most of the transformants. A similar trend was observed in the transformants with the pBZD-PI integration cassette overexpressing the PAD1 and ICT1 genes, which confers resistance to weak acids, phenolics and organic solvents. Ethanol yields were either lower or similar to the parental strain, whereas growth was either similar or higher than that of the reference strain with PI3 transformant showing a maximum growth increment of 20.9%.
The pBHD-TA integration cassette overexpresses the TAL1 and ADH6 genes that confer furan and weak acid resistance. After the first round of delta integrations, the pBHD-TA transformants showed no real differences in ethanol yields relative to the parental strain with only TA6 showing improvement at 3.6% ethanol yield increment and final growth increment of 18.6%. Transformants with the pBHD-TF integration cassette overexpressing the TAL1 and FDH1 genes conferring weak acid resistance exhibited similar or higher ethanol yields relative to parental strain with TF2 showing the highest increment in yield at 16.9%. The pBHD-TI integration cassette overexpressing the TAL1 and ICT1 genes that confers resistance to weak acids and organic solvents proved to be detrimental to both the ethanol yield and growth as transformants exhibited significant decreases in both ethanol yield (8.4-35.2%) and growth (20.4-38.2%).
After the first stage of delta integration, the pBHD-TF and pBKD-AA integration cassettes generated transformants with more than 10% increment on ethanol yield relative to the parental strain. Therefore, these combinations were selected to continue into the second round that involved the integration of the pBHD-TF and pBKD-AA integration cassettes into S. cerevisiae AA6 (resulting in ATF transformants) and TF2 (TFA transformants), respectively. The second round of transformants also resulted in a trade-off between growth and ethanol yield (see Additional file 2: Figure S2). Compared to the parental strain, 70.8% of the ATF transformants favoured growth with ATF13 showing an improved growth increment at 17.3%. Conversely, 58.3% of the TFA transformants exhibited increased ethanol yields. Interestingly, when the concentration of hydrolysate in YPDX reduced from 65% v/v to 50% v/v, no significant differences were observed in resistance phenotypes between parental strain and transformants, indicating possible phenotypic plasticity in transformants.
In the third and final round of delta integration, plasmid pBZD-PI was integrated into the ATF13 (AP transformants) and TFA7 (TP transformants) strains, thus strains overexpress six genes in the FPS1 partial deletion background. Subsequent transformants exhibit resistance to weak acids, furans and phenolic compounds. The strains were evaluated on growth (absorbance), fermentation ability (ethanol yield) and inhibitor detoxification (% conversion). S. cerevisiae CelluX TM 1 was used as an industrial and parental reference strain, whereas ATF13 and TFA7 were used as additional parental reference strains. All transformants showed a significant improvement in growth compared to the parental CelluX TM 1 strain, during fermentations with 2% SC-X media supplemented with 65% v/v sugarcane hydrolysate at pH 5 and spiked with 20 g L −1 of furfural and 20 g L −1 of formic acid (Fig. 1A). The growth profiles of the AP and TP transformants surpassed the performance of the industrial CelluX TM 1 strain. Interestingly, TFA7 with only two gene cassettes did as well as the final transformants. AP1 and AP4 showed the highest growth (OD 600 ), whereas TP1 was the best performer from the TP transformants (Fig. 1A). Only AP1 and TP1 showed a significant difference in growth between 120 h versus 168 h. As expected, ethanol concentrations were very low, ranging from 1.6 to 2.4 g L −1 given the extreme toxicity of fermentation media (data not shown).
The transformants showed a noteworthy difference in inhibitor detoxification phenotypes for formic acid (present as formate in the medium at pH 5) and furfural (Fig. 1B). Although AP4 showed the highest OD 600 at 168 h, the growth could not be linked back to an improved inhibitor resistance phenotype. In contrast, the TP1 strain showed the highest detoxification phenotype with an average of 13.4% and 19% conversion of formic acid (formate) and furfural, respectively, at 168 h (Fig. 1B). CelluX TM 1 as parental and industrial reference strain, however, showed poor growth and poor inhibitor detoxification with no formic acid converted and only 3% of the furfural detoxified. TFA7 and AP1 transformants also surpassed the parental and industrial reference strain with improved detoxification phenotypes. The copy number of the gene inserts were determined via qPCR analysis. The TFA7, TP1 and AP1 transformants each have 1 insert of the TF (TAL1 + FDH1) and AA (ARI1 + ADH6) double-gene constructs, i.e. 1 additional copy of the native TAL1, FDH1, ARI1 and ADH6 genes. Both TP1 and AP1 transformants also have 6 additional copies of native PAD1 and ICT1 genes (see Additional file 1: Table S4). The copies numbers of the inserts between the transformants were similar. However, the observed phenotypes of the TFA7, TP1 and AP1 transformants were different as the AP1 strain performed poorly compared to the more robust TFA7 and TP1 strains.
Two different assay methods were applied to evaluate inhibitor tolerance phenotypes, i.e. inhibitor tolerance plate assays (pH 4.0-4.5, no pH control) and enzymatic assays. Plate assays showed variations between inhibitor phenotypes within the three stages of strain development (Fig. 1C). In particular, the PI3 strain shows increased susceptibility to weak acid stress (6 g L −1 acetic acid and 0.8 g L −1 formic acid) when no pH control was implemented (pH < 5) and this phenotype was confirmed in AP1 and TP1 transformants with pBZD-PI inserts in the third round of delta integrations (Fig. 1C), although this integration improved resistance to the phenolic syringaldehyde. AP1 was the only transformant showing resistance to 1 g L −1 furfural when critical mass was present [45]. The in vitro activities of detoxification enzymes were assayed to determine inhibitor detoxification potential of transformants. Detoxification was measured as the decrease in substrate, i.e. furfural, cinnamic acid or formic acid due to enzymatic degradation (see Additional file 1: Table S5). No significant differences were observed between transformants and parental reference in furfural assays. In the cinnamic acid assays, PI3 and TP1 transformants exhibited enhanced in vivo cinnamic acid detoxification activity. Similarly, formic assays showed AP1 and TP1 transformants to have enhanced formic acid detoxification phenotypes, relative to control.

Detoxification phenotypes in simulated/synthetic inhibitor cocktail fermentations
The TFA7, AP1 and TP1 transformants were subjected to fermentations in 2% SC-X media supplemented with either, 5 g L −1 furfural plus 0.5 g L −1 HMF, 6 g L −1 acetic-plus 0.81 g L −1 formic acid, or 1 g L −1 cinnamic acid to ascertain detoxification phenotypes of the gene combinations to specific microbial inhibitors groups. The S. cerevisiae CelluX TM 1 strain was used as an industrial and parental reference. The transformants from the first round of integration (AA6, TF2 and PI3) were used as secondary controls to determine if second and third-round delta-integrated transformants also exhibit phenotype from first integrations, i.e. cumulative phenotypes.
In fermentations with 1 g L −1 cinnamic acid, there were no differences observed between parental CelluX TM 1 strain and TFA7 or TP1 transformants. However, the PI3 transformant exhibited an enhanced cinnamic acid detoxification phenotypes (Fig. 2A). The AA6, TFA7, AP1 and TP1 transformants showed marked improvement in furfural detoxification phenotype when compared to parental and industrial reference strains. All transformants exhibited a decrease in the lag phase of 24 h compared to the 48 h for the parental strain, with furfural detoxified within the said time period (Fig. 2B). At 48 h, transformants exhibited ethanol yields ranging from 0.25 to 0.29 g g −1 , whereas the parental control fermentation yield was below 0.1 g g −1 (Fig. 2D). Glucose was depleted within 48 h versus 72 h for the parental CelluX TM 1 reference strain with no significant differences in ethanol yield for TFA1 and AA6 strains versus the reference strain at 72 h. No significant differences in CelluX TM 1 and TFA7 fermentation performances were observed with weak acid exposure, however, this fermentation confirmed AP1 and TP1 strains are more susceptible to weak acids due to the pBZD-PI insert (Fig. 2C).
Fermentations with inhibitor cocktail were conducted with 2% SC supplemented with an inhibitor cocktail (IC) based on the composition of SSL (Table 1). Blank media supplemented with the cocktail was used as control to account for the evaporation of volatiles. Two different carbon sources were used to determine the possible effect the carbon source may have on resistance phenotypes, given the sensitivity of the introduced heterologous pathways to fermentation stresses. As anticipated, strains showed a significant difference in observed inhibitor resistance phenotypes in fermentations with glucose and xylose versus xylose only as carbon source (Fig. 3). Transformants in xylose-only fermentations showed poor detoxification phenotypes with < 5% of inhibitor compounds detoxified. In fermentations with both glucose and xylose, differences in detoxification phenotypes between transformants and the CelluX TM 1 reference strain were observed. The AP1 transformant outperformed strain CelluX TM 1 for furfural detoxification, whereas the TP1 transformant outperformed both AP1 and CelluX TM 1 for formic acid detoxification.
Overall, transformants exhibit enhanced detoxification phenotypes compared to CelluX TM 1 parent.

Hardwood SSL fermentations and final growth rate of transformants
Various concentrations of untreated SSL were used to characterize strain performances in lignocellulose fermentations in terms of consumption of glucose and xylose, ethanol concentration, ethanol yield and ethanol productivity (Table 3). In 2% SC media supplemented with 40% v/v concentrated SSL at pH 5.0, glucose was depleted within 72 h, however, xylose consumption was less than 10% for all strains with the AP1 transformant showing no xylose consumption. Ethanol yields at 72 h showed that the TP1 transformant with a yield of 0.255 g g −1 performed better than the CelluX TM 1 strain with yield a of 0.228 g g −1 (Table 3), an 11.8% increment in yield above parental control. In 2% SC supplemented with 80% v/v concentrated SSL at pH 4.5, both parental and transformant strains showed no growth, however, strains appeared to be metabolically active as seen by the consumption of sugars (Table 3). Glucose consumption for all strains exceeded 10%, with CelluX TM 1, TFA7 and TP1 consuming 15.1%, 15.8% and 16.3%, respectively. However, ethanol concentrations remained below 1 g L −1 , with only CelluX TM 1, TFA7 and TP1 strains producing ethanol, at 0.51, 0.66 and 0.48 g L −1 , respectively.
The growth kinetics of the final transformants were characterized in 2% YPDX and YPX at pH 5.0 in terms of consumption of glucose and xylose, ethanol concentration, ethanol yield, ethanol productivity, metabolic ethanol yield and the maximum specific growth rate (µmax) ( Table 3). In 2% YPDX, the TFA7 and TP1 transformants exhibited both increased ethanol yield and productivity at 0.428 g g −1 and 0.713 g L −1 h −1 and 0.432 g g −1 and 0.720 g L −1 h −1 , respectively. The cofermentation of glucose and xylose was only reduced for strain API1 during YPDX fermentations, unlike the low xylose utilization for all strains seen in SSL (0 -6% xylose consumption), highlighting the pronounced effect of inhibitors on % xylose consumption. Furthermore, xylose consumption in 2% YPX decreased from 100% of the reference CelluX TM 1 strain to 42.3 -90.1% for transformants, indicating that the strain modification impacted negatively on xylose consumption. This was confirmed in hardwood-SSL fermentations with xylose as the main carbon source, where transformants exhibited lower % xylose consumption versus reference CelluX TM 1 strain.

Discussion
Efficient carbohydrate bioconversion of lignocellulose hydrolysates remains a challenge given the numerous microbial stresses and inhibitory compounds produced during pre-treatment, despite the progress on the development of industrial yeast strains with the ability of cofermenting glucose and xylose. Overexpression of genes involved in detoxification of inhibitors could be beneficial for improving yeast tolerance. This study used a rational engineering strategy based on sequential delta homologous integrations of dual expression cassettes, allowing the introduction of several copies into the genome of the yeast. Novel target gene combinations were selected to not only modulate the cell plasma membrane against inhibitor influx, but also to direct intracellular detoxification of inhibitors and strengthen carbon/xylose metabolism. One copy of the FPS1 gene in a xylose-capable (XI) industrial strain S. cerevisiae CelluX TM 1 was deleted, whereas the TAL1, PAD1, FDH1, ICT1, ARI1 and ADH6 genes were overexpressed to develop a range of multiinhibitor-resistant strains. The impact of gene combinations on the development of cumulative inhibitor resistance phenotypes was evaluated. The first step of our study was the disruption of the FPS1 aquaglyceroporin gene, which encodes a channel protein responsible for glycerol efflux and intake of acetic acid. The deletion of FPS1 in industrial S. cerevisiae strains has proven to substantially increase both growth and ethanol yield (10-45% improvement) on glucose media under acetic acid stresses [41] and improve xylose fermentation as seen by 3-10% improvement on ethanol yield [42]. The partial deletion of the FPS1 gene in our study, however, resulted in a moderate increase on the ethanol yield (~ 5% increment), that could be ascribed to the partial deletion and/or differences in the genetic background of the strain and media (carbon source: glucose and xylose; concentration of acetic acid 2.9 g/L, pH 5).
A CelluX1∆FPS1-C5 transformant was used as recipient strain for the sequential delta integration of nine different gene combinations. These combinations were assessed for ethanol yield and biomass formation (growth measured at OD 600 ) during fermentations supplemented with 65% v/v sugarcane bagasse hydrolysate (Table 1). Compared to parental strains, there was generally a trade-off between the growth and the ethanol yield of the transformants from the first round of delta integrations (see Additional file 2: Figure S1). Nevertheless, some combinations of genes resulted in significant improvement on ethanol yield (ARI1 + ADH6 and TAL1 + FDH1)  Table 1). The benefits of ARI1 and ADH6 overexpression on biomass and cell viability have been documented [37,46]. Similarly, our results are in line with those of Sanda et al. [47] where the overexpression of both TAL1 and FDH1 resulted in improved ethanol production in xylose-fermenting laboratory strains. None of the transformants screened could ferment the sugars when the SCB hydrolysate concentration in the media was increased from 65 to 75% v/v (data not shown), that suggests selecting 2-3 genes to improve inhibitor resistance fell short of multi-inhibitor resistance phenotypes. Although, it is interesting to note that the inhibitor-specific combinations AA (ARI1 + ADH6) for furans and TF (TAL1 + FDH1) for weak acids were the only combinations to improve both ethanol yields and biomass in 65% v/v hydrolysate fermentations. Inhibitorspecific combinations with different specificities may have a cumulative effect in constructing multi-resistance phenotypes. The AA and TF combinations were used in subsequent strain development, which resulted in strains with these specific gene combinations, but in alternate integration sequences. The PI (PAD1 + ICT1) combination was included only for the final round of integrations as it did not improve ethanol yields but did improve growth ( Table 2).
The second round of delta integrations also resulted in a trade-off between ethanol yield and biomass growth, with the ATF transformants more prone towards biomass (2-17% increment in 70.8% of the transformants) while the TFA transformants were more inclined to ethanol yield improvement (5-24% increment in 58.3% of the transformants) (see Additional file 2: Figure S2). It was also found that transformants presented phenotypic plasticity, with no differences in growth or ethanol yield compared to parental strain when the toxicity of the media was reduced. The selected transformant for the next round of integration, TFA7, was also able to grow in the presence of syringaldehyde at low cells concentration (Fig. 2C).
The third round of delta integration resulted in transformants with different combinations of the selected six genes in a partial FPS1 deletion background. Of particular interest, was whether subsequent additions of inhibitor-specific gene combinations had a cumulative effect or build-up towards multi-inhibitor resistance phenotypes. The performance of the final transformants was evaluated during fermentations supplemented with inhibitors (single inhibitor or in a cocktail) and different source of sugars. The transformants were subjected to fermentations with synthetic inhibitor cocktails, as this allowed for a more controlled assessment to determine if phenotypes for specific inhibitor resistance could be linked to specific gene combinations. Compared to the parental strain, these transformants showed improved growth during fermentations supplemented with 65% v/v hydrolysate (Fig. 1A) and proved to be able to detoxify furfural and formic acid despite their high concentration in the media (20 g/L) (Fig. 1B). Final transformants presented strong furan resistance phenotypes with a 24-h reduction of the lag phase in synthetic media containing only furfural (Fig. 2B), and improved conversion of furfural when the fermentations were carried out in mixed synthetic inhibitor cocktail with glucose and xylose (AP1, Fig. 3A). The inclusion of the PAD1-ICT1 combination in both the first and the third round of integration, however, seemed to increase the sensitivity of the yeast towards weak acids at concentrations found in the sugarcane bagasse pretreatment liquor (Table 1), as inferred from the inhibitor tolerance assays (Fig. 1C) and drop in growth in synthetic inhibitor media containing only weak acids (Fig. 2C). This could explain that the TFA7 transformant, that only contains two gene cassettes, was able to outperform CelluX TM 1 and, in some cases, TP1 and AP1 as well. Alternatively, we speculate that this could also be linked to a lower metabolic burden as compared to TP1 and AP1 (Table 3). In terms of resistance towards cinnamic acids, there was no significant difference between the parental strain and the last set of transformants ( Fig. 2A). In contrast, this resistance phenotype was evident in the first round of integration for the PAD1-ICT1 combo (PI3, Fig. 2A) together with increased susceptibility to higher temperatures (Fig. 1C). The reduced thermo-tolerance could be due to an excess fluidity of the membrane caused by a higher proportion of unsaturated fatty acids incorporated by the ICT1 protein (1-acylglycerol-3-phosphate O-acyltransferase) [43].
The tolerance of the yeast towards lignocellulosicderived inhibitors is also dependent on the carbon source in the fermentation media, and xylose metabolism is much more susceptible than glucose's [2]. However, there is limited information on the possible interaction between genes conferring inhibitor tolerance, and the genes associated with the XI pathway, especially in industrial strains of S. cerevisiae. Our results demonstrate the inter-linkage between carbon metabolism and microbial inhibitor resistance. During fermentations with inhibitor cocktail media containing xylose as only carbon source, the conversion of furfural was decreased drastically in all the strains (Fig. 3B). However, this reduction was not as severe in the case of the transformants (2.4-4 times less) compared to the parental strain (7.4 times less) (Fig. 3B). Invariably, the true test of inhibitor resistance is fermentation ability exhibited with lignocellulose hydrolysates. Fermentations in SSL presented a unique challenge to the strain development as it is both xylose rich and contain microbial inhibitors unique to the paper and pulp production process such as lignosulfonates and high concentrations of Ca +2 or Mg +2 ions besides the typical compliment of weak acids, furans and phenolics (Table 1).
In 2% YPD supplemented with 40% v/v SSL, transformants TP1 and AP1 outperformed strain CelluX TM 1 (4.8% increment on ethanol yield for API1, 11.84% increment on ethanol yield for TP1), confirming enhanced inhibitor resistance phenotypes (Table 3). Given the poor performances of strains in xylose-only fermentations supplemented with inhibitor cocktail, it was expected that strains would not be able to tolerate SSL well. However, this improvement was observed despite a reduction in the xylose consumption compared to the parental strain, especially for the API1 transformant (Table 3). However, the 80% v/v SSL YPD media proved too toxic for all the strains, but it should be noted that the strains were specifically developed with furans, weak acids and phenolic compounds in mind. Hardwood-SSL contains atypical microbial inhibitors such as MgO, lignosulfonates and SO 2 which the industrial yeast strains are as yet unable to tolerate at such high concentrations (Table 1).
Unexpectedly, the ethanol production from xylose in the selected transformants was also reduced when fermentations were carried out with no inhibitors present ( Table 3, values for YPX). Nonetheless, the transformants containing the TAL1 + FDH1 combination from the first integration (TFA7, TP1) were less influenced with about 10% reduction on xylose consumption compared to an almost 58% reduction in the AP1 transformant. The positive synergism of the overexpression of these two genes has been documented for a recombinant xylosefermenting S. cerevisiae laboratory strain, i.e. the ethanol production from xylose was improved, despite the media containing both acetic acid (1.8 g L −1 ) and formic acid (0.96 g L −1 ) [47].
Different gene interactions between gene combinations with no differences in gene copy numbers were observed, suggesting there might also have been a possible "position effect" influencing the multi-inhibitor resistance phenotypes. It is tempting to speculate that the initial or first integration events exhibited a more dominant phenotype because the first cassettes integrated into highly active and assessable sites, whereas subsequent integrations were relegated to less active areas, a so-called "positional effect". As such, the transformation efficiency was reduced with each integration cycle. This positional effect could explain the higher furfural conversion of API1 (Fig. 3A), and the interaction of TAL1 + FDH1 on xylose metabolism when the combination was the first integration cassette (Table 3: TFA7, TP1), as well as the increment of conversion of cinnamic acids only when the PAD1 gene was on the first integration cassette (Fig. 2A).
Additional study into advanced strain development strategies for the manipulation of complex phenotypes, such as microbial inhibitor resistance with a minimal detrimental impact on xylose fermentation on XI engineered industrial yeasts is required. Furthermore, elucidating how gene location, gene dosage and copy number influence exhibited phenotypes, would improve the tuning of the transformation process by adjusting the DNA concentration and/or sequence of integration according to genes (function, size) [48]. Combining targeted rational engineering with techniques such as evolutionary engineering or genome shuffling may pave the way forward in the manipulation of complex phenotypes [49].

Conclusions
The efficient conversion of sugars (glucose and xylose) in the presence of microbial inhibitors for lignocellulosederived biofuels production still remains a challenge. This study aimed to evaluate novel gene combinations that confer resistance to multiple inhibitors (cumulative resistance phenotypes) in recombinant xylose-capable industrial yeast strains. The sequential delta-integration of these genes resulted in strains with improved tolerance towards furans and formic acid, but these acquired abilities somehow negatively influenced the xylose consumption capacity of the yeast. Nevertheless, there were combinations of genes where this impact was minimal, specifically when the sequence of integrations was TAL1 + FDH1 followed by ARI1 + ADH6 (and PAD1 + ICT1). Despite the reduced xylose fermentation, selected strains could outperform the parental strain when grown on synthetic media supplemented with 40% (v/v) xylose-rich SSL hydrolysate. This study highlighted inhibitor resistance as a complex phenotype and contributes towards developing advanced strain development techniques based on positive gene interactions/mechanisms to develop "hardened" multi-inhibitor resistance xylose-capable S. cerevisiae strains.

Construction of plasmids
Standard protocols for DNA manipulation were followed [50]. Genomic DNA was extracted from S. cerevisiae BY4742∆FPS1 [51] and used as template DNA for amplification of target genes open reading frame. Target genes ARI1, ADH6, FDH1, ICT1, PAD1, and TAL1 were amplified via PCR using the Phusion ® high-fidelity DNA polymerase (New England Biolabs, Ipswich, USA) and appropriate primers (Table 4) on an Applied Biosystems 2720 thermocycler (Life Technologies, CA, USA) according to the manufacturers' recommendations. The primers introduced PacI and AscI restriction sites required for directional cloning into the delta-integration plasmids, pBZD [52], pBKD [52] and pBHD [53]. PCR products were initially ligated into the pCLoneJET 1.2 commercial vector (Thermo Scientific, Waltham, USA) according to the manufacturers' guidelines. Gene sequences were verified using the dideoxy chain termination method and an ABI PRISM ™ 3100 genetic analyser (Applied Biosystems, Waltham, USA) at Central Analytical Facility (CAF) of Stellenbosch University.
The first gene expression cassettes were constructed by directional cloning of ARI1, TAL1 or PAD1 into plasmids pBKD1, pBHD1 or pBZD1 (D1), respectively, containing the constitutive PGK1 gene promoter and terminator sequences. Secondary gene expression cassettes were constructed by directional cloning of ADH6, FDH1 or ICT1 into plasmid pBKD2 (D2) containing the constitutive ENO1 gene promoter and terminator sequences.
Double gene expression cassettes were generated by subcloning the pBKD2 ENO1 pt gene cassettes as a SpeI/NotI fragment into corresponding pB(K/H/Z)D1 plasmids to yield single delta plasmids with both PGK pt and ENO pt expression cassettes (see Additional file 1: Table S1). All plasmids used and constructed in the study are listed in Table 5.

Yeast transformation and screening FPS deletion stains
The first step of the rational engineering strategy was to disrupt FPS1 to generate FPS1 deletion strains using plasmid pYFCUP1 [41]. Plasmid DNA was propagated and extracted using cetyltrimethylammonium bromide (CTAB) plasmid extraction protocol [50], and used as a template to amplify the FPS1L-CUP1-FPS1R insert by PCR using Phusion ® high-fidelity DNA polymerase (New England Biolabs, Ipswich, USA) and appropriate primers. The 3030 bp linear PCR product was separated on 1% agarose gel to confirm the insert. The PCR product was then purified using GeneJet PCR purification kit (Thermo Scientific, Waltham, USA) and transformed into S. cerevisiae CelluX TM 1 by electroporation using a Bio-Rad Gene-Pulser Apparatus (1.4 kV, 200 OHMS, and 25 µF). Transformants were incubated in 2% YPDS (20 g L −1 glucose, 10 g L −1 yeast extract, 20 g L −1 peptone, 1 M sorbitol) media at 30 °C for 4-5 h, plated on YPDS agar plates supplemented with 7 mM and 8 mM CuSO 4 and incubated for 72 h at 30 °C. Successful transformants were confirmed with PCR using FPS1-L forward and CUP1-L reverse primers. Partial FPS1 deletion was confirmed and attributed to the aneuploidy nature of parental strain. The partial FPS1 deletion strains were screened in 70-mL fermentations using 2% YPDX supplemented with sugarcane pre-treatment liquor/hydrolysate to a concentration of 50% v/v. Fermentations were sampled at 24-h intervals for 7 days and samples were analysed for fermentations products as described in HPLC analysis sector.

Strain construction
The first round of delta integration involved the transformation of nine distinct double-gene expression cassettes into the partial FPS1 deletion CelluX TM 1 strain. All integration plasmids were digested with either Bst11071 or XhoI (Thermo Scientific-Waltham, USA) according to the manufacturer recommendations and transformed into partial FPS1 deletion CelluX TM 1 strain by electroporation (1.4 kV, 200 OHMS, and 25 µF) using a Bio-Rad Gene-Pulser Apparatus. Transformants were recovered on 2% YPD supplemented with appropriate antibiotics and confirmed via PCR using PGKseq-L (D1) and BKDENOpt-L (D2) as forward primers in conjunction with gene-specific reverse primers to confirm the complete double-gene insert ( Table 4). The transformation frequency of the delta plasmids allowed for a range of copies to be integrated, thus preliminary plate screenings were conducted to identify transformants with higher copy numbers. The plate assays were based on antibiotic resistance, i.e. higher copy numbers transformants exhibit increased This work antibiotic resistance. The screening was done on 2% YPD plates supplemented with increasing concentrations of appropriate antibiotics. Ten transformants were selected and underwent high-throughput screening to determine growth and ethanol yields. Transformants were pre-cultured in 2% YPD supplemented with 20% v/v sugarcane hydrolysate for 24 h and inoculated to an optical density (OD 600 ) of 1 into 5 mL media composed of 2% YPD supplemented with 65% v/v sugarcane hydrolysate. Fermentations were conducted in capped 15-mL conical tubes incubated at 30 °C, on a rotary wheel at 100 rpm, for 120 h with endpoint sampling for fermentation products and growth (OD 600 ). S. cerevisiae CelluX TM 1 was used as a parental reference strain in all screening fermentations. The second round of delta integration involved the integration of pBKD-AA and pBHD-TF plasmids into TF2 (TAL1 + FDH1) and AA6 (ARI1 + ADH6) strains, respectively, thus generating strains with AA and TF double-gene expression cassettes in different combinations. Plasmids were linearized with XhoI (Thermo Scientific, Waltham, USA) and transformed into selected strains via electroporation. Subsequently, transformants were recovered on 2% YPD plates supplemented with appropriate antibiotics and confirmed via PCR using appropriate primer combinations. Twenty-four confirmed transformants were selected for each combination (TF2 + pBKD-AA or AA6 + pBHD-TF) and screened in 2% YPD supplemented with 65% v/v hydrolysate.
The final (third round) delta integration involved the addition of the pBZD-PI expression cassette to the second round of gene combinations to give TF2-pBKD-AA + pBZD-PI and AA6-pBHD-TF + pBZD-PI overexpression strains. Transformation efficiency decreased after each round of sequential integration, with only six strains per combination recovered and confirmed via PCR after the third round of integration. Strains were screened in 2% SC-X supplemented with 65% v/v hydrolysate and spiked with 20 g L −1 furfural and formic acid and final strains selected overexpressed six inhibitor resistance genes in different combinations. Strain names were derived from the initials of the genes inserted (e.g. TF is TAL1 + FDH1). For the second-round transformants, the initial of the second integration (D1 gene) was added to initials of the first integration (e.g. TF + ARI1 resulting in TFA), whereas, for the thirdround transformants, the initial of the first integration D1 gene and initial of the third D1 gene were combined (e.g. TAL1 (D1 gene first integration) + PAD1 (D1 gene third integration) = TP). The copy number of the introduced genes was determined using qPCR (service provided by Inqaba Biotechnical Industries (Pty) Ltd, South Africa). The ALG9 gene was used as housekeeping gene.

Enzymatic activity assays
Different enzyme assays were conducted to determine the enzymatic detoxification capacity of transformants as compared to the parental CelluX TM 1 strain. Transformants overexpressing PAD1, ARI1, ADH6 and FDH1 with direct enzymatic detoxification mechanisms were analysed in vivo for detoxification activity using crude cell extract. Strains were grown on appropriate antibiotic selective plates for 48 h. Single colonies were suspended in 1 mL SC media to an OD 600 of 1 (~ 0.6 g L −1 DW). Cells were harvested via centrifugation and washed first with PBS buffer (8.01 g L −1 NaCl, 0.2 g L −1 KCl, 1.78 g L −1 Na 2 HPO 4 .H 2 O, 0.27 g L −1 KH 2 PO 4 ) followed by two washing steps in lysis buffer (10 mM phosphate buffer, 2 mM EDTA, 1 mM PMSF, pH 7). After the second wash, 100 µL glass beads (0.4 mm diameter, Sigma Aldrich, St. Louis USA) were added to cell pellets with 100 µL lysis buffer in 2-mL Eppendorf tubes and chilled on ice for 2 min. Cells were disrupted in ten cycles with each cycle consisting of 1 min of vigorous vortexing followed by 1-min cooling steps on ice. The cell extract was centrifuged at 13,000 rpm for 5 min at 4 °C and supernatant was aspirated and analysed for enzymatic activity. Aldehyde reductase activity was assayed according to Petersson et al. [37] with modifications. The reaction mixture consisted of 10 mM furfural substrate and 100 µM NADPH cofactor in 100 mM potassium phosphate buffer pH 7.0. The reaction was started with the addition of 20 µL crude cell extract and assays were incubated at 30 °C for 45 min. Formate dehydrogenase (FDH) activity was monitored according to Hasunuma et al. [30] with modifications. Reaction mixture consisted of 50 mM sodium formate substrate and 0.4 mM NAD + cofactor in 50 mM potassium phosphate buffer pH 7.0. FDH reaction was started with the addition of 20 µL crude cell extract and assays were incubated at 30 °C for 45 min. Phenolic detoxification was determined according to Richard et al. [33] with adjustments. The reaction mixture consisted of 0.4 mM cinnamic acid substrate suspended in 20 mM sodium phosphate buffer pH 7.0. The reaction was initiated with the addition of 20 µL crude cell extract and incubated at 30 °C for 45 min. All assays were deactivated via acidification with the addition of 10% v/v H 2 SO 4 and stored at − 20 °C until analysis.

Inhibitor tolerance assays
The inhibitor tolerance phenotypes of transformants were assessed using tolerance plate assays to determine synergistic/antagonistic dynamics in strains containing the different gene expression cassettes. Transformants were streaked out on 2% YPD plates supplemented with appropriate antibiotics and incubated at 30 °C for 48 h. Single colonies of strains were suspended into 200 µL 2% SC and suspensions were made to an OD 600 of 5, 1, 0.5, 0.1 and 0.01. Inhibitor tolerance plates consisted of 2% SC agar supplemented with either 0.6 g L −1 syringaldehyde, 1 g L −1 furfural, or 6 g L −1 acetic acid and 0.8 g L −1 formic acid. S. cerevisiae CelluX TM 1 was used as parental reference strain. The AA6, TF2 and PI3 strains are inhibitor specific for furfural, weak acids and phenolics, respectively, and were used as references for the phenotype of the first double-gene expression cassettes. Transformants were spotted onto inhibitor tolerance plates (5 µL) and incubated at 30 °C for 48-72 h. Strains were also cultivated on 2% SC plates and subjected to a higher temperature of 37 °C as this temperature is more industrially relevant.

Fermentations with single inhibitor group and inhibitor cocktail
Fermentations were conducted in 2% SC-X minimal media supplemented with either; 1 g L −1 cinnamic acid at pH 5.0, 5 g L −1 furfural and 0.5 g L −1 HMF at pH 5.0, or 5 g L −1 acetic acid and 0.81 g L −1 formic acid at pH 5.0. Fermentations in 2% SC-X supplemented with inhibitor cocktail (IC) contained 6 g L −1 acetic acid, 0.8 g L −1 formic acid, 5 g L −1 furfural, 0.5 g L −1 HMF and 0.5 g L −1 cinnamic acid. The concentration of cinnamic acid at 0.5 g L −1 was selected to account for the total phenolic content of spent sulphite liquor (SSL). Strains were precultured in 50 mL 20%-IC SC-X media pH 5.0 until late exponential/early stationary phase and inoculated into 50 mL 2% SC-X or 2% SCX (xylose only) media supplemented with appropriate inhibitors, to an OD 600 of 1. Fermentations were incubated at 30 °C, shaking at 200 rpm for 120 h, with sampling at 24-h intervals. Fermentations were inoculated in triplicate with S. cerevisiae CelluX TM 1 as an industrial and parental reference strain. Media controls were included in all fermentations to account for evaporation of volatile inhibitor compounds. The inhibitor cocktail composition is based in part on inhibitor concentrations found in both SSL and sugarcane steam explosion liquor. Synthetic media supplemented with inhibitors were selected to exercise better control over experimental parameters given the inherent unknown/ unquantifiable mix of inhibitors present in hydrolysate/ SSL.

Lignocellulose hydrolysate and SSL fermentations
Fermentations with concentrated SSL was conducted to ascertain strain performance under industrially relevant fermentation conditions. SSL was kindly provided by Sappi Saiccor (Umkomaas, South Africa) which uses an acid-based sulphite pulping process. Strains were pre-cultured in 2% SC-X media supplemented with 20% v/v SSL to late exponential/early stationary phase (OD 600 > 10) and inoculated into 100-mL serum bottles with 50 mL fermentation media to an OD 600 of 1 (~ 0.6 g L −1 DW). Fermentation media consisted of 2% SC media (20 g L −1 glucose, 5 g L −1 (NH 4 ) 2 SO 4 , 1.67 g L −1 YNB w/o aa, 3 g L −1 KH 2 PO 4 and 100 mM potassium phthalate), supplemented with either 40% (pH 5) or 80% (pH 4.5) v/v concentrated SSL. Samples were taken every 24 h and analysed via HPLC. Sugarcane hydrolysate was generated via steam pretreatment and pressing of pre-treated material. First, sugarcane bagasse was water impregnated for 24 h and dewatered using a spin dryer. The steam explosion pre-treatment experiment was performed at 205 °C at a residence time of 13.5 min. The slurry was then pressed and the hydrolysate was collected, aliquot and frozen for storage at − 20 °C until use.

Chemical composition and analytical analysis
High-performance liquid chromatography (HPLC) was conducted to determine glucose, xylose, ethanol, glycerol, acetic acid and formic acid concentrations. Samples were run on an Aminex HPX-87H Column equipped with a Cation-H Micro-Guard Cartridge (Bio-Rad, Johannesburg, South Africa) at column temperature of 65 °C with a mobile phase of 5 mM sulphuric acid and a flow rate of 0.6 ml min −1 . Peaks were detected with an RI detector (Shodex, RI-101) operated at 45 °C. Furfural and HMF were analysed on a Luna C18 (2) reversed-phase column equipped with a Luna C18 (2) precolumn (Phenomenex). Mobile phases used for elution was 5 mM trifluoroacetic acid in water (phase A) and 5 mM trifluoroacetic acid in acetonitrile (phase B), the column temperature was set to 25 °C and the flow rate at 0.7 ml min −1 . Separation occurred via gradient elution, 5% mobile phase B, increasing to 11% phase B over 14 min followed by an increase to 40% phase B over 3 min and was then kept constant at 40% for 2 min. This was followed by a decrease to 5% phase B over 5 min and ended in a final step of constant composition at 5% B for 4 min to equilibrate. Peaks were detected with a Dionex Ultimate 3000 diode array detector at 215 nm and 285 nm. Phenolic compounds were analysed with Chromeleon 6.8 software on a Dionex 3000 System with UV detector at 285 nm equipped with a Waters XSelect C18 Column (4.6 × 250 mm). Mobile phases used for elution was water (phase A) and acetonitrile (phase B) at a flow rate of 0.7 mL min −1 [55]. The SO 2 content was measured via the Ripper titration [56] at the Department of Wine Biotechnology, whereas magnesium oxide (MgO) content via ICP-MS, at the Department of Geology, Stellenbosch University.

Calculations and statistical analysis
All experiments were conducted in triplicate. Data were analysed using Microsoft Excel data analysis tools,