Respirometric techniques coupled with laboratory-scale tests for kinetic and stoichiometric characterisation of fungal and bacterial tannin-degrading biofilms

Tanning is the process used to obtain leather from animal skins. Vegetable tannins have been the most used tanning agents for centuries. Historically, leather was tanned with tannins extracted from wood, leaves, fruits and roots. Nowadays, chrome and tannins are the most used tanning agents worldwide. The phenolic groups of tannins create hydrogen bonds with collagen proteins. The colour obtained depends on the mixture of tannins exploited (Mavlyanov et al. 2001). Tannins are also a portion of the chemical load of tannery wastewater, which is characterised by its low biodegradability and its relevant recalcitrant soluble chemical oxygen demand (COD). Moreover, the presence of tannins affects negatively the biological treatment of tannery wastewater because of their inhibitory properties for a wide variety of microorganisms (Field & Lettinga 1992; Mannucci et al. 2010). Tannins are usually removed from wastewater by means of chemical processes and, as a consequence, the development of a biological treatment to effectively remove tannins could lead to environmental and economic advantages (Giaccherini 2016). Despite the antimicrobial properties of tannins, several fungi and bacteria are quite resistant to tannins and are able to use tannins as carbon and energy sources (Scalbert 1991). The biodegradation of natural tannins in the environment is mainly associated with fungi rather than bacteria. Consequently, fungi are potential candidates for the bioremediation of wastewater streams generated in the tanning industry (Tigini et al. 2019; Singh et al. 2020) since tannins are treated either inefficiently, through activated sludge processes or expensively by physical–chemical treatments.

Long-term performance of fungal-based bioreactors, both under sterile and non-sterile conditions, is still a challenging task (Bardi et al. 2017). In fact, it is well-known that culture stability and degradation efficiency are frequently lost in fungal bioreactor cultures operated under continuous, non-sterile and long-term conditions (Svobodová & Novotný 2018). However, in a previous work, Spennati et al. (2019a) developed and operate a submerged packed-bed bioreactor (4 L volume) to remove Quebracho tannin (QT; the most used condensed tannin) in a continuous mode. This engineered ecosystem, based on fungi, was designed to be applied in typical tannery wastewater treatment trains, under non-sterile conditions. The bioreactor was inoculated with Aspergillus tubingensis attached as a biofilm to polyurethane foam cubes (carrier) and operated for 5 months. Different operational conditions with respect to hydraulic retention time and the organic loading rate were also set. A stable fungal biofilm was developed in the reactor fed with QT with a COD removal up to 53%. Spennati et al. (2019a) demonstrated that the above mentioned laboratory-scale bioreactor had a long-term and steady performance to biodegrade complex tannins under non-sterile conditions, offering the possibility to scale-up this novel technology to full-scale. Scaling-up this technology, therefore, represents a major advance in the field of biological tannery wastewater treatment. Even so, the key to succeed in a biological process scale-up relies in the availability of essential parameters inherent to the biocatalyst involved in the process and the bioreactions taking place (Xia et al. 2015). This means that the kinetic and stoichiometric characterisation of the biomass is a crucial step required before scaling-up the novel bio-based technology proposed in the previous study (Spennati et al. 2019a).

Fungal growth kinetics are widely studied in industrial processes implemented for the production of value-added products such as antibiotics or enzymes. The main aim of these types of industrial biotechnologies is the production of relevant metabolites where the substrate is usually glucose and fungal growth is performed in an axenic culture broth. However, the conditions adopted in the bioprocess targeting the treatment of tannery wastewater are generally more complex from those mentioned above. In this case, International Water Association models, describing microbial growth kinetics and stoichiometry, are usually applied to model wastewater treatment bioprocesses. Concerning bacteria, the use of Monod kinetics to describe growth and degradation of organic matter is widespread in the scientific community (Henze et al. 2000); however, it has also been demonstrated that such substrate-limitation kinetic equations could also describe the growth of filamentous fungi (Kelly et al. 2004), allowing the comparison between both microorganisms growth, bacteria and fungi. There are few examples in literature reporting kinetic equations describing tannin degradation by activated sludge in industrial wastewater treatment plants (WWTP) (Lu et al. 2009), and even fewer for fungal biomass (Wang et al. 2008). Literature reporting stoichiometric and kinetic characterisation of fungal biomass degrading tannin-type compounds is scarce. Only a few authors have reported kinetic and stoichiometric data related with the biodegradation of close-to-tannins substances. For example, García García et al. (1997) described the removal of phenols contained in vinasse as a sole carbon source, reporting a maximum specific growth rate of 0.06–0.047 h⁻¹, a growth yield of 0.38–0.39 g COD g⁻¹ COD and a semisaturation constant of 13,525–4,558 mg COD L⁻¹.

As mentioned above, there is a lack of information in the literature about the characterisation of tannin-degrading biofilms or biomass and its growth modelling (Spennati et al. 2019b). The performance of dedicated experiments, following a procedure to target the estimation of fungal kinetic and stoichiometric parameters, such as respirometry, becomes essential. Respirometry is the measurement and interpretation of the biological electron acceptor consumption (usually oxygen) under well-defined experimental conditions (Munz et al. 2008). Respirometry is a powerful tool that can be combined with other techniques to provide information relative to biological wastewater treatment (and in general environmental biotechnologies), including COD fractionation and the characterisation of wastewater, the estimation of kinetic and stoichiometric parameters, the monitoring of bioprocesses, and the evaluation of potential toxicity and inhibition effects over different suspended cultures, mainly in aerobic processes (Mora 2014). The principal output of respirometry is the oxygen uptake rate (OUR) measured as the consumption of dissolved oxygen (DO) per unit of time. The OUR is a function of catabolic and anabolic metabolisms of a biological system. It is possible to distinguish between the endogenous OUR (OUR_end), in other words the basal consumption of oxygen observed without the addition of any substrate, and the exogenous OUR (OUR_ex), the consumption of oxygen observed in the presence of a substrate and related to its oxidation. Respirometry has rarely been applied to study fungal biofilms or biomass. The majority of studies have applied the homogeneous respirometry under sterile conditions (Schinagl et al. 2016), on sterile solid matrices (Willcock & Magan 2001), on soil matrices to analyse some behaviours of fungi and bacteria (Boening et al. 1995), or with olive mill wastewater treatment (Caffaz et al. 2007). However, respirometric assays for fungal tannin-degrading biofilms under non-sterile conditions are still lacking.

The present study aimed at filling in the gap of information around the kinetic and stoichiometric characterisation of tannin-degrading fungal biofilms by coupling respirometric tests and mathematical modelling of experimental data. For this purpose, a classical respirometric set-up was especially adapted to obtain reliable outcomes to support further research and design on fungal-based technological solutions for tannery wastewater treatment facilities. The development of a mathematical model developed in AQUASIM was an essential step to successfully complete the overall study.

A respirometric vessel was used in this study to obtain OUR profiles from a fungal biofilm. The respirometer was equipped with probes (to monitor different variables such as temperature, pH, oxygen concentration), magnetic stirring and a pH control system. A liquid-flowing gas (LFS) configuration (flowing gas, static liquid) was selected to perform the respirometric tests. The respirometric vessel had a volume of 0.3 L and was provided with a gas diffuser. The temperature was controlled by recirculating water through the vessel jacket using a thermostatic water bath (Polystat24, Fisher Scientific, Spain). Temperature and pH probes (SenTix82, WTW, Germany) and a DO probe (CellOx 325, WTW, Germany) were connected to a benchtop meter (InoLab Multi 740, WTW, Germany) and a computer for data acquisition and process monitoring. The pH was controlled through the addition of sodium hydroxide (NaOH) and hydrochloric acid (HCl) solutions with a microburette (Multi-Burette 2-SD, Crison Instruments, Spain).

The biofilm to be characterised in the respirometer was obtained from the bioreactors mentioned above (see Spennati et al. (2019a) for detailed information), which contained the colonised polyurethane foam (PUF) cubes in a rotating cage. Bioreactors were provided with a pH control system and continuous aeration. Each respirometric test required withdrawing 10 PUF cubes from the bioreactors. The cubes were pre-washed with phosphate buffer and submerged in the respirometric vessel, containing mineral medium, to set an abiotic stage. The use of respirometry with submerged, immobilised biomass is innovative and there was not any conventional procedure stablished, thus requiring several attempts to identify a feasible procedure. Fungal pellets, fresh immobilised biomass in PUF cubes and immobilised biomass sampled from the bioreactor were used to test and define the respirometric procedure. Dry mass content in PUF cubes was analysed per triplicate (three PUF cubes) and each respirometric test (using a 0.3 L vessel) used 10 PUF cubes from the bioreactor (plus three additional cubes for dry mass determination). The PUF cubes were placed in sterile water for 24 hours under endogenous conditions. Before the test, the water was replaced. This procedure was done in order to reduce the bioreactor liquid and suspended biomass.

The pH set point in the respirometer was 5.8 ± 0.2 and controlled during the tests by dosing 0.05 M NaOH and HCl solutions. The air flow was regulated at 10 NmL min⁻¹ with a mass flow controller (TecFluid, USA). The PUF cubes were placed in the middle of the vessel, far enough from the surface and the bottom of the vessel, to avoid direct contact with both the magnetic bar and the pH and DO probes, thus minimising fluctuations.

The respirometer was spiked with multiple pulses of substrate. Before each pulse, the OUR_end and the oxygen mass transfer coefficient (K_la) were evaluated following the procedure reported by Mora (2014). In short, the colonised PUF cubes were maintained overnight under endogenous conditions (without substrate). The air flow was stopped to monitor the endogenous oxygen uptake. Once the OUR_end was calculated, the air flow was activated to calculate the K_la from the re-aeration DO profile.

Respirometric tests were performed with pulses of 1, 2, 5 and 10 mL of QT (10 g L⁻¹) in the vessel with 10 PUF cubes. The concentrations tested were within the range of 47 mg L⁻¹–476 mg L⁻¹ of COD. The QT pulses were chosen, as shown in Table 1 in order to obtain comparable experimental substrate/biomass (S/X) ratios in the respirometer and in the bioreactor.

Results obtained from respirometric tests were used to estimate kinetic and stoichiometric parameters through the calibration of the mathematical model and, afterwards, to simulate the continuous operation performance of the bioreactor. The model was designed with the software AQUASIM, which is specifically designed to simulate wastewater treatment processes and contains tools to describe dynamic processes in biofilm-based technologies. AQUASIM software includes a Biofilm Compartment option (consisting of a biofilm phase and a bulk fluid phase) that was used to simulate the biofilm growth and activity on the support media (PUF cubes), and the diffusion phenomena of substrate and oxygen within the biofilm (Boltz et al. 2010). The AQUASIM biofilm compartment supports the biofilm modelling for different types of biofilm reactors (Wanner & Morgenroth 2004). In AQUASIM, the definition of both biological systems, the respirometer and the bioreactor, was necessary before running the simulations. Specifically, for the simulation of the bioreactor, the first step was the definition of the boundary conditions, required for the biofilm compartment. The reactor volume was imposed to be 4 L and the reactor type was Confined with a constant total volume for biofilm and bulk fluid (no biofilm growth out of the PUF). The pore volume was defined as Liquid Phase only while the concentration of influent suspended solids was neglected since the inlet COD was soluble. For the same reason, the biofilm matrix was modelled as Rigid and the diffusive mass transport of solids was neglected. Among the main hypothesis of the model, adsorption was also neglected while the diffusion was considered for tannins and oxygen.

The biofilm density was calculated from samples of PUF colonised with pure fungi. The samples were inserted in a graduated volumetric vessel and the dry mass of the biomass was measured with the standard procedure. An experimental fungal biofilm average density of 12.5 g L⁻¹ as dry mass was used in the model. Nevertheless, a high variability in fungal biofilm density is reported in the literature (Spigno et al. 2003).

The diffusion coefficient in the biofilm may change with the biofilm density and the biofilm thickness; usually, the relative diffusivity (f_dif) ranges from 40% to 90% (Hibiya et al. 2004). Some authors have reported a correlation between the oxygen profiles and the biomass distribution in biopellets of Aspergillus niger (Hille et al. 2005) and others have proposed that the pellet density could allow predicting the steepness of oxygen concentration profiles. The biofilm density supports the adoption of a f_dif = 80% (Stewart 1998; Horn & Morgenroth 2006). The diffusion of phenols in water was estimated to be 8.47 × 10⁻¹⁰ m² s⁻¹ (2.35 × 10⁻¹³ m² h⁻¹) (Fan et al. 1990) and that of natural tannins to be 5 × 10⁻¹¹ m² s⁻¹ (1.38 × 10⁻¹⁴ m² h⁻¹) (Tzibranska 2000). Tannins diffusion was modelled with the same equation used for oxygen diffusion. However, dissolved compounds (oxygen, tannins) are transported first through the mass transfer boundary layer (external mass transfer) via convection and, then, through the biofilm matrix (internal mass transfer) via diffusion (Khabibor Rahman et al. 2009). Nevertheless, the external mass transfer was neglected in these simulations due to the mixing in the reactor.

As mentioned earlier, the development of a procedure to perform respirometric tests with immobilised fungi in PUF cubes (mixed fungal biomass or MFB) required several trials, also with pure fungal biomass (A. tubingensis immobilised in PUF cubes). After the endogenous phase, this pure fungal biomass (PFB) required multiple wake-up pulses (50 mg COD L⁻¹) before performing respirometric tests to recover 100% of the biological activity. Hence, this was the procedure followed to characterise the QT-degrading biofilm obtained from the bioreactor. Additionally, previous tests were also performed with PFB to verify the tannin biodegradation capacity of a selected fungal strain in the respirometer.

In Figure 1 the respirograms corresponding to the biodegradation of four pulses of QT by MFB sampled from the bioreactor are represented. An increasing COD concentration was tested and the OUR profiles were used for characterisation and modelling purposes. As preliminary assessment of the respirograms, it was observed that the OUR_max was still not reached with the highest concentration tested because the typical flattening of the OUR profile in the top region of the curve did not appear. Moreover, toxicity under 500 mg COD L⁻¹ also did not occur because the OUR_end monitored between each concentration tested was comparable (data not shown).

Experimental data obtained from respirometric tests was used to kinetically and stoichiometrically characterise QT degradation by MFB. As mentioned in previous sections, a simple Monod-type kinetic equation was used for a respirometric estimation of microbial parameters as a first approach, while a biofilm model was used to simulate the reactor performance.

A decay coefficient (b_f) of 0.22 d⁻¹, reported by Wang et al. (2008) for A. niger, was adopted in this study. The estimation of the yield coefficient (Y_f) was done through the calibration of the parameters with the respirograms. To estimate Y_h, the biodegradable COD was assumed as 17% of the total COD, extrapolated as the average during the steady-state performance of the bioreactor under similar conditions as those set during the respirometric tests (concentration and S/X ratio). The Y_f estimated was 0.45 ± 0.01 g COD g⁻¹ COD, which was similar to the value suggested in literature for A. niger degrading glucose (0.37 mg dry mass mg⁻¹ glucose or 0.51 mg COD mg⁻¹ COD) (Aguilar et al. 2001). The active biomass (X_f) was estimated from the bioreactor experimental data at two different HRT tested, once the steady state was reached. The estimated active biomass concentrations were 251 mg COD L⁻¹ and 465 mg COD L⁻¹ for HRTs of 52 and 28 h, respectively.

To estimate the affinity constant (K_sf) and the maximum specific growth rate (μ_f) of the MFB, a simple Monod-type kinetic equation, without diffusion within the biofilm, adsorption and inhibition was applied. The best estimation of K_sf and μ_f were 800 mg COD L⁻¹ and 2.88 d⁻¹, respectively, although the best fitting simulations over-estimated the measured OURs due to model simplifications. The use of simple kinetic equations, as a first approach towards the calibration of a more complex mathematical model, may be a smart shortcut that allows predicting preliminary values of some model parameters. However, after several attempts, respirometric profiles could not be described accurately. This result was expected because one of the limiting factors in aerobic biofilms performance can be the transport by diffusion of substrate and oxygen within the biofilm, especially when biofilms are thick and active. In this first calibration approach diffusion was not considered.

The QT-degrading bioreactor was operated for 5 months under different conditions. Figure 2 shows the profile corresponding to the organic loading rate experiments performed in the bioreactor (four different inlet COD concentrations ranging from 1,440 to 100 mg COD L⁻¹) and also the experimental data corresponding to the outlet COD concentration. These results were used to obtain a more accurate estimation of the kinetic parameters, μ_f and K_f, previously obtained from the calibration of a Monod-type equation with the respirometric profiles. The calibration of a non-steady-state mathematical model under shifting conditions allows avoiding the estimation of apparent kinetic and stoichiometric parameters (Süß & De Visscher 2019). Moreover, the effect of substrate inhibition was successfully described by switching from a Monod-type equation to a Haldane-type kinetic equation and implementing the corresponding equation in AQUASIM. Accurate description of experimental profiles was definitively obtained considering mass transport resistance phenomena (diffusion and K_la) through the Biofilm Compartment component of AQUASIM. In Figure 2, the simulated profile resulting from the mathematical model calibration with the bioreactor experimental data is shown. Table 2 reports the estimated values of the parameters considered in this research to characterise the QT-degrading MFB.

As can be observed from Figure 2, the calibrated model described most of the operating conditions set in the bioreactor, except for the first COD concentration tested, where only one experimental datum was available for model calibration. Regarding the estimated parameter values (Table 2), the obtainment of such a low inhibition constant highlights the negative effect that QT has on microbial activity at low concentrations. An inhibition constant of 13 g COD m⁻³ indicated that the lowest concentration tested in the respirometer was already causing an OUR reduction of 14%. Moreover, the high saturation constant indicated the low affinity that the MFB had for the QT, even showing a satisfactory performance in the bioreactor. Maximum biodegradation rates were obtained at 115 mg COD L⁻¹, presenting more than 50% of the activity at concentrations ranging from 25 to 400 mg COD L⁻¹.

Parameters estimated in this work showed some similarities with those available in the literature about the characterisation of tannin-degrading suspended biomass. In relation to the aerobic biodegradation of tannic acid by activated sludge, Li et al. (2009) reported a μ_max of 5 d⁻¹, which is comparable to that obtained in the present study. However, they also reported literature ranges for K_sf (from 10 to 866 mg L⁻¹) and K_i (from 54 to 680 mg L⁻¹) for tannins and phenolic compounds biodegradation, which highlights the importance of a direct experimental estimation of kinetic and stoichiometric parameters, instead of adopting the corresponding one reported in the literature.

The calibrated mathematical model and the estimated kinetic and stoichiometric values, together with the successful simulation of the bioreactor performance, would allow scaling-up of this novel technology and predict the optimal operating conditions to maximise the biodegradation of tannins contained in wastewater. The overall outcome from this research indicates that a fungal-based full-scale bioreactor would be a convenient process (after further engineering and optimisation) to deal with tannin-rich wastewater (such as tannery, oil mill, etc. wastewaters). Moreover, it could be used as a technology to produce fungal biomass for other purposes. As an example, it could be exploited for bioaugmentation in activated sludge processes treating wastewater with recalcitrant compounds or for bioaugmentation in ex-situ or on-site plants dedicated to the bioremediation of contaminated soils and sediments. Further research into the microbial ecology under different operating conditions would be also of interest to understand the interaction between mixed fungi biomass or biofilms and the bioreactor performance. The identification of microbes developed under the conditions set in this novel fungi-based technology would allow a faster start-up, and the adoption of optimal operational strategies by inoculating tannin-degrading bioreactors with tailor-made colonised carriers.

This research generates a potential interest for future investigation targeting the improvement of mathematical models describing tannin-degradation biofilms to understand the mechanisms and driving forces involved in such a complex bioprocess. In general, microorganisms can be observed in the environment as planktonic (free-swimming) organisms or as aggregated communities called biofilms. The transition between these two forms is a complex and highly regulated process characterised by a strong interaction between cells and various environmental signals. Multiple drives influence biofilm formation and structure, and biofilm detachment generally occurs when external forces (such as limitation nutrients diffusion, share stress etc.) overpower the internal strength of the matrix holding the biofilm together. The multiple simulations performed on respirometric tests showed that the limiting step was the diffusion in the biofilm. The chosen approach confirmed that this phenomenon played a crucial role. Further studies are required to achieve a better understanding of this and other mechanisms in order to improve the modelling of this specific immobilised biomass. As an example, coupling respirometry with microsensors would contribute to the development of a next generation procedure to characterise tannin-degrading fungal or bacterial biofilms. The combination of these two techniques would allow a better description and understanding of biofilm dynamics. Moreover, the diffusion of tannins and oxygen inside the fungal biofilm and the associated biological reactions could be studied, implementing biofilm profiling during respirometric tests with microsensors. This would improve the overall knowledge related to tannin-degrading biofilms, it would allow the development of more complex mathematical models and the development and application of online respirometry as a powerful tool to control tannery wastewater treatment.

In this study, respirometric tests have been coupled to mathematical modelling to describe the biodegradation of tannins by a fungal biofilm selected, under non-sterile aerobic conditions. Respirograms and experimental data obtained from a tannin-degrading bioreactor, operated for 5 months, were used to calibrate the mathematical model and estimate kinetic and stoichiometric parameters essential to scale-up this novel fungal-based technology. This study contributes to the validation of respirometric techniques applied to a fungal tannin-degrading biofilm and presents a modelling approach as a prediction tool to optimise operating conditions, whilst also filling the gap of knowledge on the stoichiometry and kinetics of fungi applications in environmental biotechnologies. The estimated kinetic parameters, μ_f, K_f and K_i, indicate that the mixed fungal biomass investigated can grow optimally at concentrations around 100 mg COD L⁻¹ of QT and at more than 50% of the maximum growth rate at concentrations ranging approximately from 25 to 400 mg COD L⁻¹.

The authors thank MANUNET III FUNCELL Project No. n. MNET17/ENER-1143.

All relevant data are included in the paper or its Supplementary Information.