Enhancing Bioconversion of Crude Glycerol into Butanol and 1,3-Propanediol After Pretreatment by Coupling Fermentation and In Situ Recovery: Effect of Initial pH Control

Abstract

The sharp rise in the worldwide production of biodiesel has created an excess in the crude glycerol market, so it is essential to develop new added-value alternatives for crude glycerol. This paper describes a study on fermenting high concentrations of two types of medium-pure crude glycerol to solvents by Clostridium pasteurianum. The effect of media composition (iron, yeast extract, and vitamins) on solvents production was assessed by a full factorial design with pure glycerol. Granular activated carbon (GAC) adsorption was highly effective in removing impurities from crude glycerol. Following GAC pretreatment, fermentation of glycerol at initial concentration as high as 60 g L⁻¹ was possible, resulting in a butanol production of ~9 g L⁻¹. Based on these results, a batch fermentation with in situ gas stripping and pH controlled at ≥6.5 was shown to be the best alternative to enhance biomass growth, glycerol uptake, and solvent production. The combination of controlling pH in the early stages of fermentation with in situ butanol removal stabilised the metabolism of the strain and showed that the fermentation performance with crude glycerol is very similar to that of pure glycerol. With a notable uptake of glycerol (>83%), solvent production was >11 g L⁻¹ butanol (yield > 0.21 g g⁻¹_{glycerol consumed}) and >6 g L⁻¹ 1,3-propanediol (yield > 0.13 g g⁻¹_{glycerol consumed}). Setting the fermentation conditions to achieve a high uptake of high levels of glycerol with a similar product distribution is of great interest for the viability of the industrial processing of crude glycerol into chemicals via biological conversion.

1. Introduction

The global warming emergency is promoting the development of technologies for the circular economy, which in turn is strongly linked to the necessary energy transition. In this context, the European Union has set the share of renewable energy in the transport sector to at least 29% of the final consumption of energy as an objective for 2030 [1]. A special contribution of 5.5% is foreseen for advanced biofuels and biogas from different raw materials. In this context, the biodiesel market has raised its production from 410,000 to 847,000 barrels of oil equivalent per day for the period 2012–2022 [2]. Biodiesel is produced from the transesterification of vegetable oil or animal fats and generates crude glycerol as the main byproduct at a ratio of 10:100 (w/w). Crude glycerol composition includes: glycerol, water, methanol, inorganic salts, soap, and organic matter non glycerol (MONG) [3]. The purification of crude glycerol is highly demanding in terms of energy. Crude glycerol is usually considered as a low-value by-product, if not as industrial waste. Despite the wide commercial applications of glycerol, the current market of both refined and crude glycerol is highly volatile and with falling prices [4], meaning that there is a growing interest in developing alternative uses.

Biological processes are considered greener options with economic advantages for the potential exploitation of crude glycerol [5]. Research on the use of glycerol as a fermentation substrate has been carried out on the biosynthesis of value-added products. The main biosynthesized products include butanol [6], succinic acid [7], lactic acid [8], butyric acid [9], or diols as 1-3 propanediol (1,3-PDO) [10], and 2,3-butanediol [11]. Of these products, butanol is a promising biofuel showing better properties than ethanol related to miscibility on gasoline, heating value, and vapour pressure [12]. Butanol can be obtained by Clostridium pasteurianum using glycerol as the sole carbon source [13,14,15]. The main byproducts are 1,3-PDO and ethanol, and small amounts of acetate, lactate, and butyrate [16]. Under the usual conditions, more glycerol is fermented to butanol than to 1,3-PDO, probably because of the higher energy yield of butanol formation, although the product distribution can be strongly influenced by slight variations in the fermentation conditions [14]. Although the metabolic regulation between butanol and 1,3-PDO still remains unclear, several environmental parameters such as glycerol concentration or iron content can affect the product distribution.

One of the limiting factors for developing commercial processes for butanol production lies on the low tolerance of the Clostridia species to butanol (<1.5% (w/v) for growth, <2% (w/v) for viability, [17]). Two pathways are under study to overcome this bottleneck: one is the selection of strains with more tolerance to butanol by C. pasterianum mutagenesis [15,18,19,20,21] and the other deals with an engineering approach based on fermentation with in situ recovery of products [22,23]. The presence of impurities such as fatty acid methyl esters (FAME), salts, and methanol in biodiesel-derived crude glycerol hinders the reported productivities from pure glycerol [24]. The impurities in crude glycerol restrict the initial concentration to be processed, being a major limiting factor in achieving high productivities [25,26,27]. In order to improve crude glycerol processing, pretreatment strategies are required as a prior step to reduce inhibitory compounds. Alternative methods have been explored, such as acidification, carbonation, electrodialysis, and adsorption [22,28].

The aim of this work was to demonstrate the feasibility of fermenting industrial medium-pure crude glycerol by C. pasteurianum. To assess the influence of the glycerol origin on product yield and glycerol consumption, we used crude glycerol from two biodiesel facilities. One of the industrial sites is based on used cooking oil (UCO) as the raw material for biodiesel production, while the other factory processes mixtures of animal fat and UCO. The main objective is to ferment crude glycerol at a high initial concentration. For this, pretreatment by activated carbon adsorption was selected to remove impurities from industrial products. Media composition (iron, vitamin dosing, and yeast extract) was established for promoting butanol production over 1,3-PDO. To clarify the role of pH on the metabolic pathway regulation to butanol, the effect of pH media was assessed by performing pH-controlled batch fermentations in the range of 4.5–6.5. The effect of in situ butanol recovery by gas stripping was explored as a strategy for enhancing crude glycerol exploitation. In all cases, the results with both types of crude glycerol were compared with those of pure glycerol as the model substrate.

2. Materials and Methods

2.1. Sources of Crude Glycerol and Pretreatment

Crude glycerol was provided by two Spanish biodiesel producers. The first source obtained crude glycerol as a byproduct of UCO transesterification (GLY1), while the second was from a mixture of UCO and animal fat for transesterification (GLY2). This glycerol was collected from the industrial sites, analysed, and stored at 4 °C. The compositions are summarised in

Table 1. Properties of crude glycerol derived from transesterification of used cooking oil (GLY1) and of a mixture of used cooking oil and animal fat (GLY2).

Fat Source	GLY1	GLY2
Glycerol (g L−1)	508.3	546.9
Water (g L−1)	602.7	629.2
Solids (g L−1)	9.27	9.54
Ash (g L−1)	0.69	1.45
MONG (g L−1)	66.8	42.6
Methanol (g L−1)	n.d.	6.66

n.d.: non detected.

. Both contained similar amounts of glycerol and solids, but the MONG concentration was higher in GLY1 than in GLY2. Methanol content was very low in GLY2 (methanol:glycerol mass ratio: 0.012) and undetected in GLY1, so the pretreatment sought to remove organic impurities from the fatty materials rather than the removal of methanol impurities from the biodiesel production. In this case, both industrial sites performed efficient methanol recovery after transesterification.

The crude glycerol was pretreated to remove potential inhibitory compounds by two alternative strategies: (1) filtration or (2) adsorption by granular activated carbon (GAC) followed by filtration. GAC with a specific surface area of 820 ± 10 m² g⁻¹ and an apparent density of 430.5 g dm⁻³ was purchased from Scharlab (Barcelona, Spain). Of the potential pretreatment methods, these two were selected since they are simple to scale up. The effect of both strategies was assessed by the fermentation of the pretreated crude glycerol. Samples were homogenised by shaking, and aqueous solutions of 100 g L⁻¹ of crude glycerol were prepared. These solutions were (1) filtered through a 1.2 µm mesh or (2) treated by GAC adsorption followed by 1.2 µm-filtration. GAC adsorption was carried out for 16 h at 25 °C and 150 rpm at a ratio of 1:2 w/w. Prior to filtration, GAC was separated from the aqueous solution by centrifuging for 8 min at 4000 rpm. Samples were taken before and after pretreatment for characterisation. The final solutions were used as sole substrates of fermentation in different initial glycerol concentrations (30–60 g L⁻¹) in serum bottles at 37 °C and at 150 rpm inside an orbital shaking incubator (SI500, Stuart, London, UK) for 48 h. All experiments were conducted in duplicate.

2.2. Strain Maintenance

C. pasteurianum strain DSM 525 was purchased from CECT (Spanish Type Culture Collection, Paterna, Spain). The culture was stored at −80 °C in a Reinforced Clostridial Medium (RCM) with 20% (v/v) glycerol. The cryocultures were reactivated in a culture medium containing 19 g L⁻¹ of RCM and 30 g L⁻¹ of glycerol, without shaking in sterile and anaerobic conditions for 8 h at 37 °C. Actively growing cells were inoculated in the ratio of 10% (v/v) into the fermentation medium.

2.3. Formulation of the Fermentation Medium

The effect of adding iron (0.01–5 mg FeSO₄ L⁻¹), yeast extract (1–3 g L⁻¹), and vitamins (0–0.01 mg L⁻¹ of biotin, 0–1 mg L⁻¹ of thiamine, and 0–1 mg L⁻¹ of para-amino-benzoic acid) on butanol production and glycerol consumption was evaluated. Supplementation of iron and vitamins was additional to the contributions of these substances from the yeast extract. The rest of the components of the culture medium were those according to [14]: 0.50 g L⁻¹ KH₂PO₄, 0.50 g L⁻¹ K₂HPO₄, 5 g L⁻¹ (NH₄)₂SO₄, 0.2 g L⁻¹ MgSO₄·7H₂O, 0.02 g L⁻¹ CaCl₂·2H₂O, 2 g L⁻¹ CaCO₃, 6 mL L⁻¹ of trace element solution SL 7 [29]. Resazurin sodium salt in 1 mg L⁻¹ concentration was injected for visual identification of anaerobiosis. Antifoam 204 was also added at 0.5% (v/v). The duration of the fermentation was first evaluated, resulting in <48 h. A full factorial design 2³ (FFD) consisting of 8 axial and 3 central points was carried out. For these experiments, 60 g L⁻¹ of glycerol 99.5% (technical grade) was used as the sole carbon source. Initial pH was adjusted to 6.5. Butanol concentration at 48 h was established as the response variable. The experimental designs, regression analysis and analysis of variance (ANOVA) were performed on MINITAB^® v.2020.1.0 (Minitab Inc., State College, PA, USA). All assays were carried out in serum bottles as previously described. All the chemical reagents were analytical grade.

2.4. Reactor Fermentation

A comparative study of the fermentation of both crude glycerol types was performed on a bench scale by using a 1 L reactor (700 mL working volume) with on-line pH control. Solutions of 60 g L⁻¹ of pretreated crude glycerols were used as the carbon source. The fermenter consisted of a glass stirred tank reactor (STR). The temperature was maintained at 37 °C by a heating bath (AL 18, Lauda, Lauda-Königshofen, Germany). Fermentation time was 48 h. The pH was controlled by a Tris-compatible flat pH sensor connected to LoggerPro software 3.16.2 (Vernier, Beaverton, OR, USA). The pH set-point was controlled with the automatic addition of 3M NaOH. In order to select the pH set-point a prior test with pure glycerol was performed using three minimum pH thresholds (4.5, 5.5, and 6.5). For the pH value yielding the highest butanol production, the crude glycerol was processed by in situ butanol recovery. The effect of product removal on glycerol utilisation and product yields was assessed. Fermentation gas was continuously bubbled from the bottom of the reactor by a vacuum pump (VP 86, VWR, Darmstadt, Germany) at a gas flow rate of 4.85 vvm, and butanol was recovered by a flat condenser (250 mm) with a round-bottom glass flask (500 mL). The gas circuit was purged with N₂ prior to the operation. The flat condenser and glass flask were maintained at 5 °C by a thermostatic bath (AD15R-30, VWR, Darmstadt, Germany) with a monoethylene glycol/water mixture. Sterile distilled water was periodically injected into the STR to keep constant volume. All the experiments were performed in duplicate. The results were compared with those obtained from pure glycerol.

2.5. Analytical Methods

The crude glycerol composition was determined by measuring glycerol, methanol, water and MONG concentrations along with suspended solids and ashes. Glycerol and methanol concentrations were measured by High-Pressure Liquid Chromatography (HPLC). The HPLC system (1100 Series, Agilent Technologies, Santa Clara, CA, USA) was equipped with a refractive index detector and a diode array detector. An Aminex^® HPX-87H column (300 mm × 7.8 mm, Bio-Rad Laboratories Inc., Hercules, CA, USA) was used at 35 °C, with a mobile phase of 5 mM H₂SO₄, operated at 0.6 mL min⁻¹. Water content was determined by Karl Fischer (standard method ISO 12937 [30]). MONG was quantified as g C L⁻¹ by using a total organic carbon analyser (TOC-VCHS, Shimadzu Corporation, Kyoto, Japan). MONG concentration was determined by subtracting the contribution of glycerol to TOC. The suspended solids were gravimetrically measured after filtration through 0.22 μm. Ash content was determined after burning at 550 °C for 2 h.

Fermentation samples of 1 mL were periodically withdrawn for analytical determinations. The OD₆₀₀ was measured on a SpectroFlex6600 spectrophotometer (WTW, Weilheim, Germany). Cell density (g-dw L⁻¹) was obtained from correlation with optical density at 600 nm (g-dw L⁻¹ = 0.1873 OD₆₀₀ + 0.3396; R² = 0.9996). Samples were centrifuged at 10,000 rpm for 5 min and filtered through 0.22 µm to quantify glycerol, methanol, and fermentation products (butanol, 1,3-PDO, ethanol, butyric, acetic, and lactic acids) by HPLC.

3. Results and Discussion

3.1. Effect of the Pretreatment of Crude Glycerol on Butanol Production

Table 2. Characterisation of crude glycerol types on the pretreatments by (1) filtration and (2) GAC adsorption + filtration.

	GLY1				GLY2
	Initial	After Filtration	After GAC Adsorption + Filtration		Initial	After Filtration	After GAC Adsorption + Filtration
Glycerol (g L−1)	102.5	100.6	97.7		100.9	98.9	97.7
Solids (g L−1)	1.19	0.93	0.34		1.36	0.70	0.24
Ash (g L−1)	0.05	0.05	n.d.		n.d.	n.d.	n.d.
MONG (g L−1)	11.2	10.2	4.3		9.3	8.5	3.6
Methanol (g L−1)	n.d.	n.d.	n.d.		1.21	1.16	n.d.

n.d.: not detected.

summarises the change in the composition of both crude glycerol types after applying the two alternative pretreatments to aqueous solutions of 100 g L⁻¹: (1) filtration and (2) GAC + filtration. For both glycerol sources, the sole filtration through 1.2 μm resulted in very small glycerol losses (<2%). The suspended solids content was reduced by 22% in the case of diluted GLY1, while the removal of solids for diluted GLY2 was doubled (48%). As ash content did not change with filtration of GLY1, and it was already undetectable in GLY2 samples, the higher solids removal from GLY2 indicates that the unreacted solids on biodiesel synthesis had a bigger particle size for animal fat than those from UCO.

Regarding potential fermentation inhibitors, MONG concentration was only slightly changed by filtration in both cases (<10% removal). For the aqueous dilution of GLY2, which contains little methanol (1.21 g L⁻¹), filtration, as expected, scarcely changed it, although this value is far from 5 g L⁻¹, which has been previously reported as non-inhibitory for C. pasteurianum growth on pure glycerol [25].

In the pretreatment with GAC, glycerol losses were slightly higher than those achieved solely by filtration, but in any case, losses were <5%. Solids removal was substantially higher for both glycerol types (>70%) due to the retention of the micro-solids in GAC. GAC showed a better ability to remove MONG due to the adsorption of soluble organic components, which were not retained by filtration only. The treatment was effective to a similar extent for both crude substrates. Regardless of the fat source used for biodiesel production: for UCO (GLY1) or UCO + animal fat (GLY2), similar MONG removal was observed (61.6% for GLY1 and 61.3% for GLY2). However, the difference in the fat source resulted in different MONG compositions. In fact, HPLC chromatograms of GLY2 indicated a more complex and diverse MONG composition, with a greater number of non-identified peaks than those in GLY1 (Figure S1). However, HPLC chromatograms of both glycerol types showed the reduction in the peak magnitudes, especially on dilution prior to GAC pretreatment, but also after GAC adsorption. This confirmed the additional positive effect of GAC adsorption plus dilution to reduce the levels of fatty acids and their derivatives.

The efficiency of both pretreatment strategies was assessed by fermentation at different initial glycerol concentrations, ranging from 30 to 60 g L⁻¹. The fermentation medium was supplemented with FeSO₄ (5 mg L⁻¹) and yeast extract (1 g L⁻¹). The results are plotted in Figure 1. Regardless of the type of crude glycerol or pretreatment, butanol production was successful for all the samples with an initial concentration of 30 g L⁻¹. For the two highest initial concentrations (45 and 60 g L⁻¹), no biomass growth was observed after the sole filtration of crude glycerol, and butanol production was null. Interestingly, fermentation of all samples pretreated with GAC was possible, resulting in butanol productions of from 6 to 10 g L⁻¹. All the samples with positive fermentation were characterised by MONG concentration < 3 g L⁻¹ while any concentrations over 3.9% were fully inhibitory. These results indicate that GAC adsorption + filtration allowed fermenting up to double the initial substrate concentration of that of filtration only, whatever the kind of fat used in the biodiesel plants. When inhibitory compounds were successfully removed (GAC + filtration), a positive relationship was found between initial glycerol concentration and butanol production. The highest butanol production (9.78 ± 0.39 g L⁻¹ to GLY1 or 8.64 ± 0.37 g L⁻¹ to GLY2) was obtained for the highest initial glycerol concentration (60 g L⁻¹). GAC adsorption + 1.2 μm filtration was thus selected as the pretreatment step for further experiments with crude glycerol carried out in this work. The initial glycerol concentration was selected as 60 g L⁻¹, since it demonstrated adequate fermentation after GAC refining.

3.2. Effect of the Medium Formulation on Butanol Production

Fermentation tests were carried out in a culture medium with technical-grade glycerol to evaluate the impact of iron, yeast extract and vitamin concentrations on glycerol consumption and product formation. As the main objective was to derive the metabolic fluxes to butanol formation instead of 1,3-PDO, butanol production at the end of the fermentation (48 h) was selected as the response variable for the statistical study.

Table 3. FFD matrix and results of glycerol consumption and product concentrations with an initial concentration of 60 g L⁻¹ of technical grade pure glycerol.

FFD				Glycerol Consumption (g L−1)	Biomass Growth (g-dw L−1)	Product Formation (g L−1)
Run	X1	X2	X3			Butanol	Ethanol	1,3-PDO
1	5.00	1.00	0.00	26.49	2.78	8.77	3.21	3.92
2	5.00	1.00	6.00	26.60	2.93	10.00	2.10	2.60
3	0.01	3.00	0.00	12.85	1.34	3.85	0.90	4.10
5	5.00	3.00	0.00	30.05	2.63	9.29	2.37	3.81
6	5.00	3.00	6.00	27.86	2.62	9.05	2.70	3.69
8	0.01	3.00	6.00	16.31	1.45	4.77	1.10	4.42
9	0.01	1.00	6.00	6.41	0.96	2.56	1.05	3.31
10	0.01	1.00	0.00	10.25	0.95	2.50	1.25	3.12
4, 7, 11	2.50	2.00	3.00	29.73 ± 1.66	2.85 ± 0.07	8.68 ± 0.31	2.74 ± 0.04	3.52 ± 0.07

X₁: FeSO₄ (mg L⁻¹); X₂: Yeast extract (g L⁻¹); X₃: Vitamins dosed from stock solution (mL L⁻¹).

summarises the FFD matrix along with the fermentation results. Standard deviation (STD) for butanol production was as low as 3.6% as determined on the central point (runs 4, 7, and 11). Iron limitation interferes in the activity of the alcohol dehydrogenases that catalyse the reduction in acetaldehyde to ethanol and of butyraldehyde to butanol [31,32]. The results in this study confirm the adverse influence of the low level of supplemented iron on alcohol synthesis. Butanol production ranged between 2.50 and 4.77 g L⁻¹ for runs 3, 8, 9, and 10 (0.01 mg L⁻¹ FeSO₄). Of these, the lowest values (<2.56 g L⁻¹) were achieved for runs 9 and 10, which had the combined lowest level of iron (0.01 mg L⁻¹ FeSO₄) and yeast extract (1 g L⁻¹ yeast extract). For runs 3, 8, 9, and 10 butanol yield and productivity at 48 h were of 0.308 ± 0.067 g_butanol g⁻¹_{glycerol consumed} and 0.070 ± 0.024 g L⁻¹ h⁻¹. On the other hand, the highest butanol production was achieved for runs 1, 2, 5, and 6, the assays that were performed with a high iron level (5 mg L⁻¹ FeSO₄). For these assays, butanol concentration varied between 8.77 and 10.00 g L⁻¹, with an average value (AVG) of 9.28 g L⁻¹, resulting in the highest butanol productivity at 48 h (0.193 ± 0.013 g L⁻¹ h⁻¹). The highest iron concentration also increased glycerol consumption, resulting in values ranging from 26.49 to 30.05 g L⁻¹. Nevertheless, butanol yield (0.335 ± 0.031 g_butanol g⁻¹_{glycerol consumed}) barely changed by increasing the iron level. Ethanol, which is synthetized following the pyruvate metabolic route [14,33], as is butanol synthesis, is a minor compound that was produced in a quasi-stable proportion over butanol (for runs 1–11: AVG butanol:ethanol of 3.4:1 with relative STD of 28%; ethanol yield of 0.103 ± 0.031 g_ethanol g⁻¹_{glycerol consumed}). In the case of 1,3-PDO, a shift in the yield was observed between the assays conducted under low iron level (runs 3, 8, 9 and 10: 0.353 ± 0.114 g_1,3-PDO g⁻¹_{glycerol consumed}) and those with high iron level (runs 1, 2, 5 and 6: 0.128 ± 0.021 g_1,3-PDO g⁻¹_{glycerol consumed}). The importance of iron supplementation to trigger butanol formation compared to yeast extract and vitamin dosing is also shown by thebutanol:1,3-PDO ratio. When the metabolic route favoured the formation of butanol and ethanol (with little 1,3-PDO), the AVG butanol:1,3-PDO ratio was 2.6:1 (relative STD of 25%; runs 1, 2, 5 and 6), while for the low iron level (runs 3, 8, 9, and 10) this proportion is totally inverted, with 1,3-PDO being the major product (AVG butanol:1,3-PDO of 0.9:1; relative STD of 16%).

Iron limitation substantially reduced the glycerol consumption due to the biomass growth being strictly limited. Without supplementing additional iron to the medium broth (runs 3, 8, 9, and 10), biomass growth was much lower (1.18 ± 0.25 g-dw L⁻¹) than when a minimum of 2.5 mg L⁻¹ of iron was added (2.74 ± 0.13 g-dw L⁻¹), indicating that the content of iron of the yeast extract was insufficient to trigger the biomass growth. Despite the fact that the metabolism of C. pasterianum is recognised as having a weak regulation, high initial glycerol concentration favoured butanol formation under non-iron limited conditions. Analysis of variance model regression (ANOVA) corroborated the strong influence of iron (X₁ p-value = 0.001), and the slight effect of the interaction of iron with yeast extract (X₁ X₂ p-value = 0.045) on butanol production (Table S1). The coefficient of determination (R²: 0.9977) and the adjusted coefficient of determination (Adj. R²: 0.9887) indicated a good correlation between the observed and predicted data. From the FFD results, the high level of iron (5 mg L⁻¹ FeSO₄) was selected for the rest of the experiments. As yeast extract as a single variable presented a non-significant effect in the butanol production, the low level of yeast extract (1 g L⁻¹) was chosen for further experiments, thus cutting the cost of the medium formulation. As the vitamin dosage was the factor with the lowest influence on butanol production, it was decided not to supplement the fermentation medium with additional vitamins apart from those dosed in the 1 g L⁻¹ of yeast extract.

The factors affecting metabolic regulation of C. pasteurianum are still under research. The results of this work are compared in

Table 4. Comparison of batch fermentation of pure glycerol under different levels of supplemented iron.

Initial Glycerol (g L−1)	Supplemented Iron	Equivalent Fe2+ (g L−1)	Butanol Production (g L−1)	1,3-PDO Production (g L−1)	Butanol Yield (g g−1) a	1,3-PDO Yield (g g−1) b	Reference
25	0.1 g L−1 FeSO4·7H2O	2.0 10−2	5.5 c	7.6 c	0.24 d	0.33 d	[26]
20	0.02 g L−1 FeSO4·7H2O	4.0 10−3	~6.0 e	~1.8 e	n.a.	n.a	[35]
80	0.01 g L−1 FeSO4·7H2O	2.0 10−3	10.0	6.6	0.25	0.16	[18]
30	0.05 g L−1 FeSO4	1.8 10−2	8.7	~1.0 e	0.29	--	[34]
50	0.01 g L−1 FeSO4·7H2O	2.0 10−3	6.5	~3.3 e	0.16	~0.08 c	[6]
	0.1 g L−1 FeSO4·7H2O	2.0 10−2	8.6	~4.5	0.21	~0.11 c
45	0.943 mg L−1 FeCl2	4.2 10−4	7.1	6.8	0.20	0.19	[15]
20	0.05 g L−1 FeSO4	1.8 10−2	~5.7 e	n.a.	~0.28 c	n.a.	[13]
60	5 mg L−1 FeSO4	1.8 10−3	8.8	3.9	0.33	0.15	This study (run 1-Table 3)

^a g_butanol g⁻¹_{glycerol consumed}; ^b g_1,3-PDO g⁻¹_{glycerol consumed}; ^c calculated from reported data; ^d average results reported previously; ^e read from graphical data.

with data from the literature when using pure glycerol (data from the selected iron, yeast, and vitamin levels (run 1, Table 3) are included). As can be seen, a wide range of iron supplementation was tested regardless of the initial glycerol concentration, with values as high as 2.0 × 10⁻² g equivalent Fe²⁺ L⁻¹. When comparing butanol production, the concentration achieved in this study (8.8 g L⁻¹) with 1.8, 10⁻³ g equivalent Fe²⁺ L⁻¹ was at the top of the reported range (5.5–10.0 g L⁻¹). Also, the high butanol concentration in this study was accompanied by the highest butanol yield (0.33 g_butanol g⁻¹_{glycerol consumed}). As can be seen, the use of 10-fold iron concentrations (1.8 10⁻²–2.0 10⁻² g equivalent Fe²⁺ L⁻¹, [6,13,26,34]) did not improve either the butanol production or butanol yield achieved. The data with an initial glycerol concentration ≥ 45 g L⁻¹ confirmed that lowering the level of Fe²⁺ 4-fold from that used in this study (4.2 10⁻⁴ g equivalent Fe²⁺ L⁻¹, [15]) has an adverse impact on butanol production versus 1-3 PDO (butanol:1-3 PDO ratio: 1.05). Increasing the supplementation of Fe²⁺ 10-fold derives in a similar proportion (butanol:1,3-PDO ratio: 2.0, [6]) to that achieved with the iron levels used in this study (butanol:1,3-PDO ratio: 1.9 ± 0.3, [6,18] and the present study). It was therefore shown that the medium formulation used herein contains enough Fe²⁺ to modulate the metabolism of C. Pasterianum to butanol formation over 1,3-PDO production when processing high substrate concentrations.

The fermentation of the two pretreated crude glycerol types was tested with the selected levels (high for iron, low for yeast, and low for vitamins) of the medium formulation. Fermentation profiles along with those corresponding to the pure glycerol counterpart (run 1, FFD experiment) are plotted in Figure 2. For all assays, the glycerol consumption rate declined in 24 h, corresponding approximately to the end of the exponential growth phase. This was accompanied by a solvent production > 88% of the final amount; the rest of the product appeared mainly from 24 to 40 h (>97.5%), with nearly no noticeable change in product concentration at 40 h. A very similar product formation and distribution between butanol, ethanol, and 1,3-PDO was achieved, whatever the glycerol source, confirming the feasibility of fermenting crude glycerol after partially removing MONG content with GAC. The main difference lay in the maximum biomass concentration between crude glycerol (2.64 g-dw L⁻¹ for GLY1; 2.35 g-dw L⁻¹ for GLY2) and pure glycerol (2.98 g-dw L⁻¹). This seems to indicate that the remaining impurities limited biomass growth to a certain extent, and consequently glycerol uptake. The limited biomass growth resulted in a slightly lower total solvent production (butanol + ethanol + 1,3-PDO). A total solvent production of 15.62 g L⁻¹ was achieved with pure glycerol, while 14.97 g L⁻¹ and 14.14 g L⁻¹ were obtained for GLY1 and GLY2. Interestingly, the lower biomass growth did not adversely affect butanol production. In fact, butanol production was independent of the glycerol source (AVG: 8.91 g L⁻¹; STD: 0.57 g/L⁻¹). Butanol yields for crude glycerol (0.39 and 0.43 g_butanol g⁻¹_{glycerol consumed} for GLY1 and GLY2) were slightly higher than those obtained for pure glycerol (0.33 g_butanol g⁻¹_{glycerol consumed}). In the few reported studies on crude glycerol fermentation by C. pasterianum, butanol yields vary between 0.24 and 0.34 g_butanol g⁻¹_{glycerol consumed} using initial glycerol concentrations of from 25 to 50 g L⁻¹ [25,26,27]. The butanol yields achieved in this study thus corroborate the positive effect of crude glycerol pretreatment by GAC for industrial bioprocessing into butanol. Nevertheless, as part of the crude glycerol remained unreacted, other limiting factors such as the amount of undissociated acids was evaluated. In all cases, the pH fell to approximately 5.6 during the first 16 h (Figure 2) with small concentrations of butyric acid (<0.6 g L⁻¹, <1.5 mM of undissociated form) detected in the buffered medium. In Clostridium fermentations, the reduction in pH is associated with the accumulation of undissociated forms of volatile fatty acids, the presence of the undissociated form of butyric acid being necessary to produce butanol [36]. In this regard, the exogenous addition of 3 g L⁻¹ of butyric acid at a pH of 5.3 (8.5 mM of undissociated form of butyric acid) favoured the increase in the butanol yield without reducing the butanol production rate [34]. However, adding 4 g L⁻¹ at the same pH extended the lag-phase and reduced the metabolic rate (11 mM of undissociated form of butyric acid). A balance between the total amount of acids in the fermentation medium and the dissociated/undissociated ratio, which in turn links to pH, thus seems a key factor in enhancing cell growth and glycerol consumption. Implementing strategies for pH control could therefore be an alternative for improving glycerol utilisation when processing high substrate concentrations.

3.3. Effect of the Medium pH on Butanol Production

The effect of medium pH was assessed in 1 L reactors equipped with an on-line pH control using pure glycerol as the sole substrate. The pH control strategy consisted of controlling the minimum pH over the fermentation course, which has been shown to be a good strategy to promote butanol production for Clostridium acetobutylicum [37]. Fermentation profiles for the pH-setpoint varying from ≥4.5 to ≥6.5 are plotted in Figure 3. As can be seen, specific cell growth and glycerol consumption rates increased on increasing the setpoint for minimum pH. The data show that biomass concentration doubled on increasing minimum pH from 4.5 to 6.5 (2.02 and 4.12 g-dw L⁻¹ at 16 h at the end of exponential growth phase), and unreacted glycerol concentration fell from 42 to 25 g L⁻¹. The highest solvent production and glycerol consumption was achieved for the experiment at pH ≥ 6.5. A high biomass concentration is linked to the faster assimilation of butyric acid. Indeed, butyric acid was undetected after 17 h for the experiment at pH ≥ 6.5. The total concentration of butyric acid was 2.00 g L⁻¹ (at 16 h) for pH ≥ 4.5, which corresponds to 15.36 mM of the undissociated species, clearly exceeding the inhibitory concentration for cell growth of C. pasteurianum [34].

Table 5. The effect of pH on glycerol consumption and product formation. Initial pure glycerol concentration of 60 g L⁻¹. Production and product yields measured at 48 h.

pH Control	Glycerol Consumption (g L−1)	Product Formation (g L−1)			Product Yield (g g−1glycerol consumed)			Maximum Productivity (g L−1 h−1)
		Butanol	Ethanol	1,3-PDO	Butanol	Ethanol	1,3-PDO	Butanol	Ethanol	1,3-PDO
≥4.5	18.84	5.78	n.d.	1.16	0.31	--	0.06	0.55	--	0.32
≥5.5	29.32	7.51	2.05	2.17	0.26	0.07	0.07	0.76	0.39	0.31
≥6.5	38.18	9.09	2.11	3.80	0.25	0.06	0.10	1.20	0.32	0.31

n.d.: not detected.

shows the influence of pH control on product formation and glycerol consumption. Raising the minimum pH-setpoint from 4.5 to 6.5 doubled glycerol consumption, while butanol production increased by 57%. In the case of ethanol, it was not detected for the pH set point condition ≥ 4.5 and appeared at 2 g L⁻¹ when the minimum pH was raised to 5.5, coinciding with a decrease in butanol yield. This suggests a shift in metabolic flux among the metabolites of the pyruvate route (butanol and ethanol). A subsequent rise in pH did not affect butanol and ethanol yields, suggesting that the increase in butanol production may be related to enhanced biomass formation. The production of 1,3-PDO exhibited an increasing trend with rising minimum pH levels. Although there was a slight negative effect on butanol yield by increasing pH in two units, working at pH ≥ 6.5 improves butanol production as a result of the higher glycerol consumption. Raising the pH increased the maximum productivity of the metabolites of the pyruvate route, while productivity of 1,3-PDO barely changed. The maximum productivity of butanol more than doubled when the pH-setpoint was raised by two units (from 4.5 to 6.5), and even increased by 58% by raising the pH by one unit (from 5.5 to 6.5). Keeping the minimum pH of the fermentation medium above 6.5 thus not only favoured glycerol consumption but also butanol productivity, meaning that this operational condition was selected for further processing of crude glycerol under pH-controlled fermentation.

3.4. Fermentation of Crude Glycerol Under pH Control

The profiles of biomass growth, substrates, and metabolites (acids and alcohols) are shown in Figure 4 for the fermentation of pretreated crude glycerol (GLY1 and GLY2) at controlled pH ≥ 6.5. Profiles for pure glycerol at the same controlled pH are also plotted for comparison. It can be seen that crude glycerol negatively influenced the maximum biomass concentration. Values were of 2.77 ± 0.11 g-dw L⁻¹ for GLY1 and 3.00 ± 0.08 g-dw L⁻¹ for GLY2, while 4.10 ± 0.02 g-dw L⁻¹ was obtained for pure glycerol. Only GLY1 was found to have a negative effect on the specific growth rate: maximum specific growth rate of 0.116 h⁻¹ for GLY1 against 0.211 h⁻¹ and 0.200 h⁻¹ for pure glycerol and GLY2, respectively. The impact on biomass growth or rate can be attributed to the nature and/or the quantity of the residual impurities remaining in the crude glycerol, which, in turn is related to the industrial source. However, the reduced biomass concentration or growth rate did not adversely influence the final butanol production. In all cases, butanol concentrations exceeded 9 g L⁻¹ (9.09 ± 0.06, 9.35 ± 0.14 and 9.18 ± 0.56 g L⁻¹ for pure glycerol, crude GLY1, and crude GLY2, respectively). Whatever the glycerol source, maximum butanol concentration peaked at maximum biomass concentration. The increase in butanol concentration from nearly zero to 9 g L⁻¹ in a few hours (approx. from 8 to 20 h) stopped biomass growth, as the butanol inhibitory concentration for growth was greatly exceeded, so that a substantial amount of unreacted glycerol remained in the broth, especially for pure glycerol and GLY2. Butanol changes the composition of the cell membrane and affects both internal pH and substrate and product transport [38]. For example, studies show that maximum specific growth rate of C. pasteurianum falls to half at a butanol concentration of 7.64 g L⁻¹ [39], while the butanol production rate seems to play a key role in glycerol consumption and product formation. In the present results from two different sources of crude glycerol, some clear differences were found in terms of 1,3-PDO production. It seems that the slower biomass growth in GLY1 delayed butanol production. The lower toxicity at earlier fermentation times (<12 h) extended the period of bacterial metabolism to 40 h. Under excess glycerol, metabolism shifted to 1,3-PDO production, which is less toxic than butanol [14]. In contrast, high butanol concentrations were rapidly achieved for GLY2 (as in the case of pure glycerol), thus stopping biomass activity sooner and lower 1,3-PDO production, as compared with GLY1. The difference in the product formation between GLY1 and GLY2 can be related not only to the greater MONG concentration in GLY1 but also to the nature of the impurities. The interaction of fatty acids such as stearic, oleic, or linoleic acids with the membrane hinders the diffusion processes and causes inhibition [25]. As the effect is greater for unsaturated fatty acids, it should be expected to be greater in GLY1, which was obtained from the transesterification of UCO from vegetable oils only. As butanol concentration in the fermentation broth over time interferes with either glycerol uptake or the preferred metabolic pathways, in situ butanol removal is required for maximising butanol production when processing glycerol concentration as high as 60 g L⁻¹.

Pure and crude glycerol fermentations coupled with in situ gas stripping were carried out at a minimum pH set point of 6.5. Gas stripping was connected after 8 h of fermentation, when butanol production was still below the inhibitory level and was maintained until the end of fermentation. Butanol concentration was periodically determined in both the fermentation medium and in the stripping condensate. Total butanol production was determined from the mass balances. The fermentation profiles are plotted in Figure 5 for pure and both crude glycerol types. By removing butanol from the fermentation broth, the maximum cell concentration using pure glycerol (6.20 ± 0.26 g-dw L⁻¹) and GLY2 (3.70 ± 0.10 g-dw L⁻¹) was considerably greater and was reached later than in the assays without gas stripping. In contrast, a similar maximum cell concentration was found both with (2.75 ± 0.15 g-dw L⁻¹) and without gas stripping (2.77 ± 0.11 g-dw L⁻¹) for GLY1. The improvement in maximum cell concentration, especially in pure glycerol but also in GLY2 was associated with a considerable increase in glycerol utilisation. Total glycerol consumption was nearly complete (>99%) for pure glycerol and 83% for GLY2, which means an increase of 79% and 49% in comparison with those obtained for the glycerol sources without in situ butanol recovery. In the case of GLY1, glycerol consumption was already over 85% by solely keeping the minimum pH above 6.5, without the need for gas stripping. Nevertheless, a slightly better glycerol consumption (87%) was achieved with gas stripping, due to the improved butanol synthesis.

Butanol removal by gas stripping had a strong positive impact on metabolite production. Butanol recovery raised the total production of butanol by more than 20%, while the effect for pure glycerol was particularly outstanding, with an increase of 37% versus the assay without gas stripping. Regardless of the glycerol source, butanol production exceeded 11 g L⁻¹ (12.45 ± 0.83, 11.27 ± 0.12, and 11.43 ± 0.32 g L⁻¹ for pure glycerol, GLY1, and GLY2, respectively). The results obtained showed that in situ butanol recovery by gas stripping improved the conditions inside the broth and allowed fermentation to continue. Remarkably, this improvement took place despite the butanol concentration in the broth being over 9 g L⁻¹ for only a few hours. This was due to an imbalance between the butanol production rate and the stripping rate in the first 24 h of fermentation. Indeed, maximum butanol productivities in this period were higher than 0.9 g_butanol L⁻¹ h⁻¹, while butanol stripping rates were lower than 0.3 g_butanol L⁻¹ h⁻¹. In the final stage of fermentation, average stripping rates varied between 0.41 and 0.64 g_butanol L⁻¹ h⁻¹, being slightly higher for pure glycerol than for the two sources of crude glycerol. On the other hand, gas stripping enhanced the production of 1,3-PDO for pure glycerol and GLY2 (7.15 ± 0.06, and 6.78 ± 1.00, respectively). Other authors [22,40] have also reported increased 1,3-PDO production under gas stripping. Interestingly, in the case of GLY1, 1,3-PDO production was of the same order, or even slightly less (7.27 ± 0.14 g L⁻¹) than in fermentation without butanol recovery. The combination of minimum pH on the fermentation broth of 6.5 combined with gas stripping was thus shown to be an efficient technique for consuming high concentrations of pretreated crude glycerol by activated carbon with substantial butanol (>11 g L⁻¹) and 1-3 PDO productions (6–7 g L⁻¹).

Carbon mass balances accounting for glycerol consumption and formation of the three alcohols: butanol, 1,3-PDO, ethanol, and the acids: butyric and acetic (lactic acid was negligible), as well as biomass growth, assuming a cell mass formula of C₄H₇O₂N [14], resulted in overall yields of 0.70 ± 0.03 g C g⁻¹ C_{glycerol consumed}. The data confirms that the overall yield was practically constant for both sources of glycerol and process configuration. Mass balances were accomplished in all the experiments with 92% or greater mass closure. Although CO₂ production was not measured, the production of CO₂ calculated according to the stoichiometry proposed in [14] was 0.27 g C g⁻¹ C of glycerol consumed. The effect of gas stripping on the distribution of the major metabolites grouped by the number of C in the major products (C2, C3 and C4) was evaluated. Figure 6 shows the fermentation product distribution for the three sources of glycerol both with and without in situ gas stripping. In the case of the experiments with in situ gas-stripping, product formation was calculated as the sum of the concentrations in the fermentation broth and in the condensate from extracted gas with reference to broth volume. Experiments without gas stripping showed greater differences in the product distribution among the three types of glycerol than those carried out with gas stripping. The larger ratio towards C3 in case of GLY1 without gas stripping was the most remarkable difference. Gas stripping led to a nearly equal product distribution for both types of glycerol, while the in situ removal of butanol directed metabolism towards the formation of C4 products for all glycerol sources in detriment of C2 products. On the other hand, the incomplete assimilation of butyric acid (butyric acid %: 14.46 ± 0.20 for pure glycerol, 13.79 ± 0.07 for GLY1 and 12.39 ± 0.09 for GLY2) indicates that there is still room for raising butanol production by a better balance between stripping and glycerol uptake rates. The presence of impurities in the crude glycerol did not affect the predominance of the metabolic pathways, which means that the slightly lower uptake of crude glycerol than that of pure glycerol can be attributed to the limitation of substrate entering the cell in the presence of small amounts of fatty acids. However, the similar product distribution regardless of the source of the glycerol is important for future upscaling to industrial processes. In fact, the observed stability emphasises the robustness of the process, which could help to overcome the variability of the raw material derived from industrial biodiesel production.

4. Conclusions

A process consisting of activated carbon pretreatment followed by fermentation coupled to product recovery was shown to be a feasible configuration for the production of butanol from medium-pure industrial crude glycerol derived from two biodiesel plants with different fat sources (UCO or UCO plus animal fat). Irrespective of the crude glycerol source, the cell inhibitory compounds were successfully reduced by AC adsorption. As a result, a double substrate concentration (60 g L⁻¹) could be fermented further than that in the absence of pretreatment (30 g L⁻¹), which is of interest for the economics of the process. It was found that keeping the pH above 6.5 favours cell growth and the assimilation of acidic species to solvents, especially butyric acid to butanol. Butanol recovery by gas stripping is an easy technique for avoiding inhibition and guarantees practically the total use of the substrate (>83%) and a very high conversion to solvents (>18 g L⁻¹) from which the major compound was butanol (>11 g L⁻¹). The distribution of the synthesised products with in situ gas stripping was practically equal for the two sources of crude glycerol assayed, thus providing stability and robustness to the process. The results obtained in this work are expected to be useful in industrial applications and represent a contribution to biodiesel plants within the circular economy.