1 Introduction

Hydrolysis poses one of the main challenges during anaerobic digestion (AD) of residual biomass, in particular when lignocellulosic biomass is used. Pretreatment methods to improve hydrolysis efficiency are often energy intensive and account for up to 40% of process costs during AD [1, 2]. Lignocellulosic residues like straw, agricultural waste, or sawmill residues are abundant, but recalcitrant for AD due to their crystalline lignocellulosic structure. Thus, efficient methods to improve lignocellulose digestion are required [1]. Naturally occurring microorganisms developed complex enzyme systems to break down lignocellulose. However, they might need an acidic environment in order to achieve maximum hydrolysis rates; thus, a separate reactor stage may be needed for a good process performance [3]. In a hydrolysis stage, process conditions can be adjusted to fit the needs of hydrolytic and acidogenic microorganisms better than in a main digester where process conditions have to meet requirements of the more sensitive methanogenic organisms. The separation of hydrolysis and acidogenesis from other phases of full AD has already shown to improve the energy balance of the whole process, which results in a higher net energy production finally [1, 4].

Plug-flow reactors (PFRs) have been applied as reliable AD reactors for dry AD for some time [5]. Higher stability, high removal rates, and low acid contents make them an attractive alternative to a conventional stirred tank reactor [6]. Lately, their characteristic to form gradients along the reactor due to a laminar flow field has got into focus: It has been shown that phase separation into acidogenesis and methanogenesis inside a PFR allows higher substrate flexibility, higher resistance against a low pH value [7], and higher loading rates compared to traditional stirred tanks [8]. The development of certain distinguished process phases inside a PFR has been explored for efficient biogas production of maize silage [8], pineapple peel and pulp [9], food waste with municipal waste water [6], and cattle manure [7], among others. Enrichment of different microbial subpopulations with hydrolytic bacteria in the front and methanogenic bacteria in the late part of a PFR has been shown by Dong et al. [7]. There, recirculation within the acidogenic phase further enriched hydrolytic and cellulose-degrading bacteria and resulted in improved chemical oxygen demand (COD) removal of 25% compared to AD without recirculation. This microbial enrichment helped to digest recalcitrant feedstock-like lignocellulosic residues. However, to the authors’ best knowledge, a PFR has not been explored as sole hydrolysis stage yet. In literature, the focus is put on high biogas and methane yields, while pure optimization of hydrolysis and acidogenesis yields remain rare. Operating a PFR separately as hydrolysis stage furthermore opens a variety of biorefinery pathways from intermediate metabolite into valuable products like long-chain fatty acids, biofuels, bioplastics, among others, which can be synthesized from short-chain carboxylic acids [3].

Bioaugmentation is an option to increase yields of certain products in anaerobic processes, especially during AD for methane production. Pure cultures and mixed microbial consortia have been successfully explored to increase hydrolysis, hydrogen production, or the methane yield mostly during batch processes in stirred tank reactors [10,11,12,13,14]. Fewer studies focused on continuous processes, where survival rates and a potential washout of the bioaugmented organisms was investigated. Martin-Ryals et al. studied the effect of a daily bioaugmentation with a cellulolytic consortium on the continuous digestion of sweet corn processing waste in a stirred tank reactor which led to an increase of soluble COD and methane production by 29–68% and 31–34%, respectively, compared to the non-augmented reaction [10]. Comparing one-time and repeated bioaugmentation in a sequencing batch trial, they achieved a 25% higher net soluble COD generation with daily addition while the effects of one-time addition were insignificant after 14 days of fermentation. Similarly, the positive effects of bioaugmentation of a thermophilic consortium on the digestion of corn stalks and cow dung were only notable in the first days, but almost insignificant after 9 days of digestion [15]. The bioaugmentation strategy of Tsapekos et al. of adding Clostridium thermocellum over 6 days to the continuous digestion of cow manure and wheat straw enhanced the methane yield significantly during addition (26–33%), but became insignificant after two hydraulic retention times (HRTs) in a steady-state operation mode [16]. In most cases, the reactor conditions are unfavorable to support further growth of the augmented bacteria [16]. The addition of the anaerobic fungus Orpinomyces joyonii to the inoculum during AD of rice straw resulted, however, in a long-term increase of cellulose and hemicellulose and reduction of volatile solids, and thus in an increased methane yield (38%). During this continuous digestion with 15 days of HRT, the fungi were able to survive for 122 days inside the reactor [17].

In this study, two Paenibacillus spp. were chosen based on their hydrolytic activity under acidic, anaerobic, and mesophilic conditions. Paenibacillus glucanolyticus is a facultative anaerobe, Gram-positive, rod-shaped, spore-forming, and mesophilic bacteria that has optimal growth conditions at pH 5.7 [18]. Hydrolysis of various β-glucans like carboxymethylcellulose, hemicellulose, cellulose, starch, arbutin, and cellobiose has been confirmed [18, 19]. Aerobic and anaerobic growth of two strains of P. glucanolyticus on lignocellulosic biomass without nutrient supplementation was confirmed by Mathews et al., where anaerobic growth yielded higher cell densities and better lignin degradation at pH 9.0 [19, 20]. Paenibacillus macerans exhibits hydrolytic activity on hemicellulose under anaerobic conditions and produces α- and β-glucan-degrading enzymes, including cellulase. It was shown that a hydrolytic activity was maintained until a pH value of 5.5 [21]. The ability to utilize residual oxygen from the substrate and their ability to sporulate probably ensure the survival of both organisms in the reactor under unfavorable conditions.

Maize silage (MS) is one of the most common substrates for AD and provides comparably high yields of biogas [22]. Santi et al. [23] showed, however, that the lignocellulosic components in MS are often incompletely fermented, with a digestion efficiency of only 40% and 29% for cellulose and hemicellulose, respectively. The application of MS in AD is controversial, due to the competition for arable land for food and feed production and due to soil erosion, nitrate leaching, or pesticide use [22, 24]. For the expansion of a circular economy, the utilization of locally available residual biogenic resources is envisaged. Straw is a lignocellulosic residue connected with grain production and offers potential for AD, but digestion is limited due to its recalcitrant structure [25]. For the investigation of the effect of bioaugmentation on feedstock with different contents of lignocellulosic substrates, MS and local bedding straw were used in different proportions.

To the authors’ knowledge, an application of a PFR as a hydrolysis stage in an anaerobic process, combined with bioaugmentation and thin-slurry recirculation, has not yet been investigated. Thus, the aim of our study was, in particular, to (i) describe hydrolysis in a PFR that was operated with MS and straw, (ii) investigate the effect of bioaugmentation with hydrolytic bacteria, and (iii) examine the impact of thin-slurry recirculation on the bioaugmentation efficiency.

2 Materials and methods

2.1 Feedstock

The feedstock that was used in this study was whole plant MS obtained from a farm in Bavaria, Germany, and mixed straw used for bedding in the horse stable of uFA Fabrik Berlin. Due to the continuous operation of the reactors, different batches of feedstock were used throughout the study. Characteristics of the feedstock were analyzed for each batch separately. Data were used for individual calculations correspondingly (Table 1). Upon arrival, both feedstocks were dried and sieved with a particle size exclusion of 50 mm. Bigger parts were crushed with a blender to reach the cutoff size. As storage of the feedstock had to be conducted for several months with limited cooling capacity, feedstock drying was assumed to result in the lowest change of the organic material, although loss of volatile components from the MS could have occurred. The bedding straw also contained small amounts of horse manure, sand, and residues of feed (apples, grains, etc.). Inert residues like plastics and stones were removed, and the residual composition mixed into the substrate. All samples were measured in triplicates. Total solids (TS), volatile solids (VS), and the ash content of the substrate were measured by APHA standards [26]. The soluble and the insoluble acidic lignin content were measured with the two-step acid hydrolysis method as described by Bhagia et al. [27] and Sluiter et al. [28]. For the measurement of non-structural carbohydrates, dry samples were milled for 2 min with the IKA A10 laboratory mill (IKA®-Werke GmbH & CO. KG). Fifty milligrams of the milled sample were extracted in 5 mL of 80% ethanol for 10 min according to [29]. The supernatant was removed, and extraction repeated for two times. The supernatants were combined and diluted 1:10 with double-distilled water. The carbohydrate solution was analyzed with the sulfuric-acid-UV method for lignocellulosic feedstock as described by Albalasmeh et al. [30]. For the measurement of the total chemical oxygen demand (tCOD), 50 mg of dried and milled feedstock were mixed with 1 mL double-distilled water in an Eppendorf tube. The sample was vortexed, soaked for 1 h, and afterwards diluted 1:10. Half of the sample was centrifuged for 5 min at 7.5 × g and filtered through 0.45 µm regenerated cellulose filters for the determination of soluble COD (sCOD). Both tCOD and sCOD samples were additionally diluted by a factor of 10, so that the samples were finally diluted by a factor of 100. Samples were analyzed with the Hach Lange Kit LCI 400 (Hach Lange GmbH, Berlin, Germany) for COD content and double-distilled water as reference solution. The tCOD solution was also used to analyze total LATON nitrogen with the Hach Lange Kit LCK 338.

Table 1 Average substrate characteristics of maize silage and bedding straw. Characterization has been conducted for different substrate batches (see supplementary material, Table S1)
Full size table

2.2 Plug-flow bioreactor operation

Anaerobic microbial hydrolysis was performed in a PFR with a maximum working volume of 14 L. The PFR was equipped with three sample ports horizontally distributed along the reactor, respectively, at the inlet, center, and outlet, each combined with sets of online sensors for conductivity, pH value, oxidation–reduction potential (ORP), and temperature for online gradient monitoring. Online data points were recorded every 10 min. The reactor scheme is shown in Fig. 1. The reactor was kept at mesophilic temperatures by external hose-heating and isolation. The pH value remained uncontrolled. Continuous stirring at 5 rpm was applied to prevent blockage. Gas composition measurement with BCP-CO2, BCP-H2, and BCP-CH4 sensors (BlueSens gas sensor GmbH, Herten, Germany) was installed in the last experimental phase with a feedstock share of 66% (w/w) of straw. The reactor was inoculated with 1.5 L of sludge from another PFR which was operated with MS in a dark fermentation process mode. Eight hundred grams of dry MS and 12 L of water were added for start-up. The reactor was operated in batch mode and afterwards semi-continuously under dynamic conditions of the HRT and thin sludge recirculation with a constant organic loading rate of 4 kgTVS m−3 d−1 to achieve stable microbial conditions. From week 15 of the experiment, a constant HRT of 14 days was maintained to achieve comparable conditions for the bioaugmentation trials. Samples from each port were taken two times per week and additionally before and after bioaugmentation. During the bioaugmentation experiments, a HRT of 14 days and a TS content between 12 and 16% (w/w) was maintained. Twenty percent of the weekly harvest of thin sludge was recirculated at each feeding/harvesting event to prolong the retention time of microbial biomass within the reactor. Harvesting and feeding of substrate and water was performed 4 times per week right after sampling. The reactor was operated anaerobically; however, residual oxygen from the dry feedstock entered during the feeding events. The harvest was sieved, and residual solids quantified by dry weight determination (mHarvest). Regularly, the solids content of the thin sludge was determined to set up the mass balance of the reactor (cHarvest). The total solids content of the PFR was calculated as follows:

$${TS}_{PFR}\left[\%\right]=\frac{{m}_{solids\left(PFR\right)}}{{m}_{total\left(PFR\right)}}\cdot 100$$
(1)
$$\frac{A\cdot {m}_{Feed}\left[kg\right]-{m}_{Harvest}\left[kg\right]-{C}_{Harvest}\left[\frac{kg}{L}\right]\cdot {V}_{Harvest}\left[L\right]}{{m}_{solids}\left[kg\right]+{m}_{{H}_{2}O}\left[kg\right]}$$
(2)

The mass flow of gas was not measured in this setup. In order to account for the organic material that was converted to gas, a percentage of the feedstock was deducted (A). These factors for MS (AMS = 0.85) and the mixed straw feedstock (AStraw = 0.75) were determined with the average difference of mass inflow and harvest during the total time of operation with the same feedstock. The reactor was operated on three different feedstock mixtures with increasing straw content being (i) MS, (ii) 30% (w/w) of bedding straw with MS, and (iii) 66% (w/w) of bedding straw with MS (see Table 2). After a substrate change, the PFR was allowed to stabilize over 3 HRT before bioaugmentation experiments were started. The benefits of online monitoring of gradients in the PFR will be discussed in a subsequent publication. In this report, relevant gradients as in pH value and conductivity are considered briefly. If not otherwise stated, average values of the measurements from all three ports were applied for all figures and calculations.

Fig. 1
figure 1

Schematic design of the plug-flow reactor with sensor triplets and sampling ports at the inlet, center, and outlet

Full size image
Table 2 Operational parameters of the PFR
Full size table

2.3 Bioaugmentation culture conditions

Two bacterial strains were chosen for bioaugmentation to the microbial hydrolysis process. Robust strains were chosen for hydrolytic activity on lignocellulosic substrates under mesophilic, anaerobic, and acidic conditions. Paenibacillus glucanolyticus (DSM 5188, formerly classified as Bacillus glucanolyticus) and Paenibacillus macerans (DSM 24, formerly Bacillus macerans) were both obtained from the DSMZ-German Collection of Microorganisms and Cell Cultures GmbH (Braunschweig, Germany). The two strains were grown in full medium containing 10 g L−1 yeast extract, 3 g L−1 meat extract, 5 g L−1 peptone, 10 mg L−1 MNSO4, and 10 g L−1 glucose, which were set to a pH value of 6.0. As P. macerans showed very quick acidification in batch culture, a phosphate buffer was added by amounts of 1.079 g L−1 KH2PO4 and 3.844 g L−1 K2HPO4 and the media pH value was adjusted to 8.0 by the addition of 30% NaOH. For cultivations of P. glucanolyticus, 25 mL of preculture was inoculated from cryoculture and grown aerobically in batch culture in Ultra-Yield™ flasks with Aerotop™ membranes (Thompson Scientific) at 30 °C and 200 rpm. After 24 h, between 4 and 6 cultures with a volume of 250 mL were inoculated from the preculture and cultivated for 2–3 days until no further growth was observed. Cultivation for P. macerans followed the same workflow, while cultures were grown at 37 °C and 200 rpm.

The bacteria were added to the reactor twice, namely at the onset of the cultivation on day 1 and on day 8, which corresponds to t = 0 and t = 0.5 of the total HRT. The conditions were maintained between 2 (MS and 30% of straw) and 3 full HRTs (66% of straw), wherein the first HRT is referred to as “addition phase” and the following as “recirculation phase.” All bioaugmentation experiments were conducted sequentially in the same reactor. To reduce the mutual influence of bioaugmentation experiments, at least one full HRT of 14 days without recirculation was maintained to wash out residual bacteria. The first addition of P. glucanolyticus represents an exemption, as the organism was added three times at t = 0, 0.5, and 1 HRT due to insufficient growth during the second addition. Detailed information to the bioaugmentation is shown in Table 3.

Table 3 Conditions at bioaugmentation with Paenibacillus spp. to plug-flow-based hydrolysis
Full size table

2.4 Analytical methods

For the measurement of sCOD, samples were centrifuged for 5 min at 7.500 g and filtered through 0.45-µm regenerated cellulose filters. One sample from each port was measured once per week with the HACH Lange Kit LCI 400 at a dilution of 1:100. Short-chain carboxylic acids (SCCAs) were measured by HPLC (1200 series Agilent Technologies, Waldbronn, DE) with the HyperREZ XP Carbohydrate H + 8 µm column (Thermo Fisher Scientific Inc., Waltham, MA) at 65 °C and 0.5 M H2SO4 as fluid phase and refractive index detection. HPLC samples were prepared by leaving 2 mL samples at 4 °C overnight for crystallization of lignocellulosic residues. Samples were then centrifuged at 16.000 × g for 10 min, sterile filtrated (0.2-µm nylon filter, Carl Roth GmbH, Karlsruhe, Germany) and clarified with the Carrez clarification kit (Merck KGaA, Darmstadt, Germany). The physiological state of the mixed culture was determined through the measurement of frequency-dispersed polarizability anisotropy (FDAP) with Elotrace (EloSystems GbR, Berlin, Germany). Firstly, particles were removed by centrifugation for 5 min at 650 rpm. Two milliliters of supernatant were transferred into 18 mL of dH2O and centrifuged for cell separation (2 min, 6000 rpm). The cell pellet was resuspended in 30 mL of dH2O in order to achieve an adequate level of initial electrical conductivity (< 40 µS cm−1) and measured at frequencies of 200, 400, 900, and 2100 kHz. This was proven to be appropriate for anaerobic fermentation samples [31]. Sugars were determined in selected samples via GC–MS measurement as described by Kielhorn et al. [32].

2.5 Calculation of metabolic activity

Hydrolysis is commonly determined by the measurement of the soluble organic matter versus the total organic matter. This is performed via COD or total organic carbon determinations. Here, the ratio of soluble to total COD was applied (Eq. 3) [33, 34]. During continuous digestion, however, this calculation is not accurate, as the tCOD in the reactor is changing over time and can only be determined by a mass balance. In order to account for that, a COD balance in relation to the feed was applied on a weekly basis. The specific hydrolysis rate (SHR) (Eq. 4) describes the soluble metabolites released from the substrate into the fluid phase (adapted from [35]). Organic material from the recirculation or bioaugmentation was subtracted to obtain the actual release (Eq. 5). The net acid production was calculated correspondingly (Eq. 6). To determine the acidification, the theoretical COD of the acids was calculated and put into proportion with the sCOD (Eq. 7, adapted from [36, 37])).

$$Hydrolysis\;\left[\%\right]=\frac{sCOD\left(t\right)}{tCOD\left(t\right)}$$
(3)
$$spec.\;hydrolysis\;rate\left[\frac{{g}_{O_2 released}}{{kg}_{vs fed}}\right]=\frac{{sC\dot OD}_{released}}{{VS}_{in}}$$
(4)
$${sC\dot OD}_{released}\left[\frac{{g}_{O_{2}}}w\right]={sC\dot OD}_{out}-{C\dot OD}_{in}-{C\dot OD}_{Recire}$$
(5)
$$Acid\;production\;rate\;\left[\frac{g_{SCCA}}{{kg}_{vs fed}}\right]=\frac{{SC\dot CA}_{produced}}{{VS}_{in}}$$
(6)
$$Acidification\;\left[\%\right]\;=\;\frac{{COD}_{tSCCA}}{sCOD}$$
(7)

3 Results and discussion

In order to study the performance of microbial hydrolysis in a PFR, MS mixed with different ratios of bedding straw was digested in a continuous fermentation mode. Bioaugmentation with two Paenibacillus spp. and partial thin slurry recirculation was applied to examine the contribution to an improved hydrolysis.

3.1 Influence of bioaugmentation on online measurement of pH value, conductivity, ORP, and gas

During the start-up period under dynamic conditions with MS as single feedstock, gradients of the pH value and conductivity developed. This indicates that the reactor operated in plug-flow mode and as a result, different microenvironmental conditions are present. In this paper, the focus of investigation was put on the influence of bioaugmentation onto hydrolysis efficiency and acid production. However, the possibility of process monitoring and control with online gradient monitoring of pH value, OPR, and conductivity in the PFR will be discussed in a separate publication.

After the dynamic operation with MS, a rather stable pH value between 3.8 and 4.0 developed, ensuring the inhibition of methanogenesis in the PFR [38, 39]. Higher straw content in the feedstock increased the pH value and shifted the pH gradient towards higher values at the inlet and decreasing pH along the reactor. Bioaugmentation with P. glucanolyticus temporarily increased the pH value during the first addition with MS digestion, whereas bioaugmentation with P. macerans increased the pH value in all ports. Estimations with the Henderson-Hasselbach equation showed that the amount of buffer salts in the P. macerans medium is too low to shift the pH in 13 L of liquid volume with a SCCA concentration of about 9 g L−1. During all bioaugmentation events, the pH value rose during the addition phases—likely by the higher solubilization of substrate where also alkaline substances were released—and decreased during the recirculation phase of thin sludge due to acid accumulation (see Fig. 2). A pH value between 5.0 and 7.0 is considered best for hydrolysis in a separate digestion stage while acidogenesis is active in a wide range between 4.0 and 8.0 [3, 40, 41]. Hydrolytic bacteria show higher activity and growth rates at higher to neutral pH value; e.g., degradation of MS was limited below pH 5.5 for Clostridium spp. common in AD [42]. While a pH value of 4.0 to 5.0 may inhibit the bacterial growth, enzyme activity of fungal and bacterial cellulases was found to be highest at in this pH range. However, hemicellulases have a high activity between pH 5.0 and 6.0 but possess a higher activity also over a wider pH range of 3.0 to 10.0 [43]. In our study, the pH value was between 3.5 and 4.8, so secreted cellulolytic and hemicellulolytic enzymes should be active at considerably high rates. Moreover, a low pH value below 4.0 can also contribute to hydrolysis of cellulose and hemicellulose as it acts as acid pretreatment [44]. Small organic acids released during AD like acetic acid, butyric acid, and propionic acid have been shown to promote the degradation of lignocellulosic biomass as they act as natural catalysts for the rupture of lignin-carbohydrate complexes (autocatalysis) by increasing the porosity and surface area of the feedstock [45, 46]. Pretreatment of rice straw with 2% acetic acid for 24 h at 80 °C reduced the cellulose and hemicellulose contents by 8.9 and 18.1%, respectively, while the combination of mechanical treatment and 2% acetic acid resulted in up to 29% lignin removal in corn straw [45]. In this study, SCCA concentrations of about 1% were reached (see Section. 3.2) that could have contributed to the rupture of the feedstock.

Fig. 2
figure 2

Data from continuous hydrolysis in a PFR under operation with MS (a), 30% straw (b) and 66% straw (c). Left: Online measurement of pH value, conductivity, and oxidation–reduction potential (ORP) at the inlet (blue squares), center (red circles), and outlet (black triangles). Weekly averages are depicted and the periods of bioaugmentation with P. glucanolyticus and P. macerans are marked in gray. Right: Profiles of soluble COD (black diamonds), total SCCA (green stars), specific hydrolysis rate (SHR), and acid production over time. Shown are the weekly averages, the shades indicate the deviation between sampling ports

Full size image

Conductivity was stable during the reference cultivation with MS (W 15–16/27–30), 30% (w/w) of straw (W 33–36) and 66% (w/w) of straw (W 51–54) but increased significantly after bioaugmentation. A higher straw content increased the baseline conductivity in the reactor, probably due to additional ions that are brought in by the concomitantly added manure in the straw. Conductivity measurements of the different batches of pure feedstocks showed a basal conductivity of 520–540 µS cm−1 for MS and 440–621 µS cm−1 for the bedding straw. Bioaugmentation with the Paenibacillus spp. led to an increased conductivity, especially at the center part of the PFR, likely due to maximal microbial activity there. In any AD process, conductivity is mainly influenced by bicarbonate/carbonate concentrations. During acidic digestion, however, bicarbonate concentration is negligible and linear correlations of conductivity with SCCA production have been found [47, 48]. The same correlation was observed during bioaugmentation in this study, where the total SCCA concentration increased concurrently with the increase in conductivity. From week 60 to 66, a drastic increase in conductivity in the center part of the reactor can be seen. This trend, which starts in the second recirculation period of P. glucanolyticus bioaugmentation, is most probably not associated with bioaugmentation or increasing acid concentrations—in fact, SCCA levels decreased from week 60 till 62—but to a new batch of straw substrate with higher basal conductivity of 621 µS cm−1 that was fed from the end of week 59 on. In the online measurement of conductivity, prompt decreases of conductivity after feeding events are measurable in MS digestion due to dilution with the added water (900–1000 µS cm−1, see supplementary material, Figure S1). However, this effect was quickly compensated by the quick solubilization of easily digestible matter into the liquid phase. The online measurements also showed a higher instability of conductivity in the inlet with MS feedstock, but a rather stable signal in center and outlet. We assume that the balance of salt uptake for growth and organic solubilization in the inlet is causing this, while the acidogenic processes at the center and outlet are more stable. With higher straw content, larger fluctuations in the center and outlet conductivity can be seen independent of feeding.

In this study, under all conditions of digestion, ORP remained in a favorable range for hydrolysis of lignocellulosic biomass; high ORP would result in complete substrate oxidation and carbon loss [49]. The ORP is naturally low in AD with baselines at − 480 to − 460 mV, whereas an ORP between − 100 and − 300 mV is considered optimal for methane production [40, 49, 50]. No direct influence of the bioaugmentation periods on ORP could be seen in our study. In the recirculation phase of P. glucanolyticus (weeks 18–20), the ORP increased abruptly from around − 410 to − 360 mV possibly due to oxygen added during feeding and recirculation or due to the disruption of the phase formation in the PFR. Chetawan et al. [49] found optimal hydrolysis efficiency in piggery waste at ORP of − 420 mV. They hypothesized that a controlled ORP level through micro-aeration leads to increased activity and enzyme production of facultative bacteria, while carbon loss due to respiration and complete oxidation to CO2 is diminished. The low ORP in the inlet during MS digestion (Fig. 2(a)) thus could have had positive effects on hydrolysis. The center and outlet ports with higher ORP indicate butyric type fermentation that has been associated with ORP levels between − 300 and − 250 mV [51]. Changing ORP levels in this trial were not related to bioaugmentation but rather to the shifting balance in the redox reactions and the concentration and state of the electron carriers due to the adaption of hydrolysis processes.

Occasional off-gas measurement during the digestion with MS showed expectedly low methane concentrations in the off-gas, namely below 2%. Online measurement of the gas phase was conducted from weeks 46 to 63 (Fig. 3). Bioaugmentation with P. glucanolyticus in week 55 led to a lasting increase of CO2 and CH4 concentration, due to higher metabolic rates in the reactor. Hydrogen concentration was decreased, possibly indicating the substrate concurrence of P. glucanolyticus with hydrogen-producing bacteria in the first 2 HRTs after addition. Gas concentration of especially hydrogen might be underestimated, due to measurement interferences in mixed gas and its’ high volatility. The residual gas volume is largely made up by nitrogen and traces of hydrogen sulfide and ammonia, which both usually appear in AD. Oxygen that might have entered during feeding is consumed immediately by aerobic bacteria (as measured in previous experiments). The low ORP measurement during the experiments confirms the very reductive environment, which is typical for a process with a very low oxygen abundance.

Fig. 3
figure 3

Average values gained during the online measurement of the gas phase of the PFR: CO2 (blue squares), H2 (red circles), and CH4 (black triangles)

Full size image

3.2 Effect of bioaugmentation on metabolite accumulation

Under non-augmented operation with 30% and 66% (w/w) of straw, stable concentrations of sCOD and total SCCA were measured in between 1 and 2 HRT of stabilization, indicating a steady state in the reactor (weeks 33–35, 51–54). For MS digestion in the PFR, the more dynamic state before and after bioaugmentation was used as reference. Highest sCOD values with MS were reached under bioaugmentation during the recirculation period being 27.8 gO2 L−1 in week 19 with P. glucanolyticus and 30.8 gO2 L−1 with P. macerans in week 26 (see Fig. 2(a)), which are accordingly 38.9 and 53.9% higher than under non-augmented reference conditions. Correspondingly, higher concentrations of total SCCA with an increase of 47.3 and 43.5% for P. glucanolyticus or P. macerans addition in the recirculation phase were detected. With the change to 30% (w/w) of straw, sCOD stabilized at 20.2 gO2 L−1 in the same range as during the non-augmented cultivation with MS. The bioaugmentation increased the sCOD concentration with P. macerans by 23.0 and 24.3% during the addition and recirculation period, and for P. glucanolyticus by 36.2 and 30.5%, respectively (see Fig. 2(b)). Total SCCA increased accordingly from the non-augmented average of 7.4 g L−1 to 8.6–10.0 g L−1 (+ 17–36%) for P. macerans and to 10.7–10.8 g L−1 (+ 45–46%) for P. glucanolyticus. Differently than with MS, P. glucanolyticus showed a faster increase in acid production with 30% straw (w/w). While soluble COD measurements shortly after bioaugmentation showed a higher concentration due to residual sugars in the cultivation medium, this effect did not last long and the weekly average sCOD did not show significant changes within the addition phase. GC–MS analysis of samples taken 4 h after bioaugmentation showed a glucose concentration in the range of 10–90 mg L−1. As shown in Fig. 4, GC–MS measurements of selected samples showed detectable glucose concentrations only in the bioaugmentation samples. A short-term availability of free sugars could thus boost the microbial activity and release of acids. Very small levels of arabinose and xylose between 10 and 40 mg L−1 were detected throughout the cultivation, indicating the breakdown of hemicellulosic structures. No other sugars were detectable. During the whole cultivation, hydrolysis was still the limiting factor, as all released sugars remained in a range of the typical saturation constant (the so-called KS value) of many bacteria with respect to the detectable carbohydrates, and thus were limiting the bacterial metabolism. In the recirculation phase, the sCOD increased significantly for both Paenibacillus spp., indicating an increased hydrolytic activity where a higher amount of particulate feedstock is solubilized in the medium. Feedstock adaption and growth of the Paenibacillus spp. could explain the delayed increase of sCOD in the recirculation phase.

Fig. 4
figure 4

Left: Soluble sugars of selected samples in the fermentation broth measured by GC–MS in the inlet (blue squares), center (red circles), and outlet port (black triangles) of the PFR. Right: Distribution of SCCAs when different feedstocks and bioaugmentation scenarios were applied

Full size image

For both feedstocks, MS and 30% of straw (w/w), the elevated metabolite concentrations decreased, when recirculation was stopped; thus, the benefit of recirculation is directly observable. Therefore, the bioaugmentation tests with 66% (w/w) straw were operated for 3 HRT, including 2 recirculation periods to assess the duration of positive effects on hydrolysis. The change of feedstock to 66% (w/w) of straw decreased the sCOD significantly to an average of 14.2 gO2 L−1 (Fig. 2(c)). The higher share of recalcitrant substrate reduced the hydrolysis efficiency. Bioaugmentation with P. glucanolyticus led to an increase of the sCOD (23.9%) and total SCCA (13.9%) during the addition period but both decreased rapidly in the recirculation periods. The addition of P. macerans resulted in an average higher sCOD concentration (16.0 gO2 L−1) that stayed above the reference level for the two recirculation periods, and correspondingly increased the total SCCA (18–34.2%). The applied bioaugmentation/recirculation strategy was successful in maintaining positive effects for 3 HRT when adding P. macerans with recalcitrant feedstock, but P. glucanolyticus was less effective. Bioaugmentation, in all feedstocks, increased the acid concentration compared to the prior state (Fig. 4). Judging over the whole test period, this was most significant for MS (+ 29.7 to 31.3%) and 30% straw (w/w) (+ 26.5 to 45.6%). Even with the higher content of recalcitrant feedstock in the 30% (w/w) straw digestion, bioaugmentation contributed to higher acid concentrations than during the digestion of pure MS. With 66% straw (w/w), only P. macerans addition led to an average increase of total SCCAs of 26.5%.

At all different feedstocks, butyrate (31.8–69.7%) and acetate (22.0 to 45.2%) were the dominant SCCAs. Also small proportions of other SCCAs were detected, but their share remained mostly below 5% of all measured components (Figure S2). Bioaugmentation with P. macerans increased the concentrations of pyruvate, succinate, citrate and lactate, and both Paenibacillus spp. increased the production of butyrate for all feedstock compositions. Although it has been described that Paenibacillus spp. do not produce butyrate in high amounts [19, 21, 52], we assume that the hydrolytic enzymes produced by the Paenibacillus spp. increased the substrate availability in the PFR and thus contributed to butyrate production by providing the respective precursors. This is in agreement with results from Esquivel-Elizondo et al. [53], who investigated butyrate production in controlled fermentation systems via metagenome prediction analysis and found that non-butyrate-producing bacteria contributed to butyrate production through the generation of acetate, lactate, and succinate; these acids are then converted to butyrate via interconversion reactions. In our study, especially with the addition of P. macerans, increased levels of lactate, succinate and pyruvate were detected. Also, concurrent acetate and butyrate production can take place via the butyrate fermentation pathway from sugars released from polysaccharide digestion by microorganisms of the genus of Clostridium, Bacillus, Bacteroides, Syntrophobacter, and Methylobacterium [54]. Hydrogen is produced as secondary metabolite in this pathway. The drop in hydrogen production during P. glucanolyticus bioaugmentation with 66% of straw (w/w) could indicate the metabolic switch from butyrate production from carbohydrates towards the conversion from intermediates to butyrate without hydrogen production. Formerly, it was found that butyrate accumulation is favored at a pH range of 5.0 to 6.0 [3]. In our case, however, a stable production of butyrate was achieved at pH values that were between 3.8 and 4.8. Generally in AD, a low pH value between 4.0 and 4.5 is associated with ethanol and lactate production by Lactobacilli at these lower pH values [51, 55]. In the cultivations studied here, an increased lactate accumulation occurred at a pH value between 3.5 and 3.6 (unpublished data). In dark fermentation, the balance between lactate- and butyrate-producing bacteria highly depends on the pH value. Detman et al. [55] showed that microbial communities from dark fermentation are able to convert acetate and lactate to butyrate. Such communities showed a higher biodiversity than those dominated by lactic acid bacteria. Likely, the gradients in the PFR created microenvironments which contributed to a higher biodiversity at lower pH values; thus, no metabolic shift towards lactate production was observable. Symbiotic cross-feeding on lactate of butyrate producers like Clostridia seems to be common in dark fermentation of complex feedstocks. It is even believed that this is the route that provides major carbon fluxes towards butyrate and hydrogen production [55]. Moreover, in our continuous PFR system, butyrate production might have been favored by several factors, which are a low hydrogen headspace pressure, longer HRT [53], and the availability of pentose sugars [56] (as detected by GC–MS measurements in this study). To the authors’ knowledge, no production of butyrate from lignocellulosic residues has been reported at this low pH value so far. Synergistic effects between different microbial species are known to enhance growth and substrate degradation in harsh conditions. Although partnerships of microbial communities producing butyrate are not well understood [53], these could have ensured a stable production with only minor shifts of the SCCA’s accumulation profile in the PFR.

3.3 Effect of bioaugmentation and recirculation on hydrolysis

Hydrolysis (Eq. 3) was about 20% on average during the reference cultivation of MS and 30% straw. A changed composition of the feedstock to 66% straw decreased it to 15.8%, due to the higher amount of recalcitrant material. Bioaugmentation increased the hydrolysis for MS and 30% straw to a maximum of 25.2% for 30% straw using P. macerans (Table 4). In case of a digestion of 66% straw (w/w), an increased hydrolysis was only determined for P. macerans addition with 19.7%, reaching similar values as with pure MS. It must be noted, however, that using this hydrolysis formula, mass transfer to the gas phase is not included. As mentioned before, there was a decrease in mass flow through the reactor of 15% for MS and 25% for straw substrate caused by substrate conversion to gas, so the hydrolysis rate might have been underestimated.

Table 4 Average measurement of FDAP, hydrolysis and acidification depending on substrate and bioaugmentation conditions. Included are all measurements at all ports
Full size table

During MS cultivation, an acidification between 64 and 68% was achieved, which was similar for reference and augmented conditions. For MS, maximal acidification at controlled alkaline conditions reached up to 71% (61% in acidic conditions) as measured by Jankowska et al. [57]. Thus, plug-flow-based digestion in acidic conditions as conducted in our study achieved comparable values as reported in literature. Bioaugmentation increased acidification only in the case of 30% straw (w/w). It can be assumed that the addition of Paenibacillus spp. contributed to the solubilization and degradation of polymeric substances, making them better available to the acidogenic bacteria. No significant influence of the Paenibacillus spp. could be found for MS and 66% of straw (w/w). Likely, the already-high acidification of MS under non-augmented conditions caused acidic stress and prevented higher acidification by bioaugmentation. Chen et al. [51] only reached a maximal acidification of 36% in the acidogenic digestion of rice straw and food waste; thus, we conclude that our PFR setup is better suited for acid production from lignocellulosic residues like straw. In total, addition of Paenibacillus spp. contributed to higher hydrolysis, which was more effective with a higher amount of MS and only slightly contributed to a higher acidification.

In the case of MS, the SHR stabilizes at 258 gO2 kg−1VSfed in weeks 15 and 16. An SHR of 264 gO2 kg−1Vsfed was reached when 30% of straw (w/w) was applied. Addition of P. glucanolyticus to MS digestion increased the SHR during the addition and recirculation phase by 7.2% and 42%, and by 4.0% and 26.4% for 30% (w/w) of straw accordingly (Fig. 5, Table 5). Recirculation prolonged and increased the positive effects on SHR. Similarly, acid production was enhanced by up to 25% for MS and 42% for 30% (w/w) of straw digestion. SHR with 66% (w/w) of straw was generally lower than that for the other feedstock mixtures, reaching an average value of 195 gO2 kg−1Vsfed under non-augmented conditions. During the digestion of 66% (w/w) of straw, the addition of P. glucanolyticus did not increase the SHR, but decreased it by 15%. This might be attributed to a higher share of carbon sources used for cell maintenance following the bioaugmentation. In the two recirculation phases, SHR fluctuated close around the level under reference cultivation conditions. The addition of P. glucanolyticus, however, led to an increased acid production during the addition phase and recirculation phase I. Bioaugmentation with P. macerans showed similar effects on metabolic rates. While minor changes in SHR occurred in the addition phase, the SHR was increased by 63.2% and 21.4% for MS and 30% straw (w/w) accordingly during recirculation, while net acid production was increased by up to 23% and 32%, respectively. Equally as with P. glucanolyticus addition, P. macerans addition to 66% straw (w/w) resulted in a decreased SHR over the whole trial period; however, the net acid production was increased. This indicates that the added bacteria contributed to the degradation of soluble, polymeric, and oligomeric carbohydrates, making them available for the acidogenic microorganisms.

Fig. 5
figure 5

Average values of SHR (top) and net acid production (bottom) before, with, and after bioaugmentation with P. macerans and P. glucanolyticus. The columns of bioaugmentation refer to the different phases: addition phase (I), first recirculation phase (II), second recirculation phase (III)

Full size image
Table 5 Percentual change of metabolic rates caused by bioaugmentation. A non-augmented cultivation under the same conditions is serving as reference state. An average of the values of the reference states with MS as feedstock before and after bioaugmentation was used here for easier comparability
Full size table

Maximum values of the net acid production from MS were reached under bioaugmentation (see Fig. 5). Bioaugmentation of both organisms yielded a very similar average acid production of 139.1 and 140.4 gSCCA kgVS−1 for P. glucanolyticus and P. macerans correspondingly. Acid production of 310 gSCCA gCOD−1 was reached by Li et al. [58] digesting MS in a leach-bed reactor at controlled pH 8.0. It seems acid production was limited in our study. Since a high acidification of over 65% was achieved, this is likely caused by limited hydrolysis efficiency. Higher straw content decreased acid production to 97.2 and 70.6 gSCCA kgVS−1 at 30% or 66% (w/w) of straw, respectively. The addition of Paenibacillus spp. to 30% straw (w/w) increased the acid production to an average of 116.4 and 136.3 gSCCA kgVS−1 for P. macerans and P. glucanolyticus, respectively; the latter being only slightly lower than the corresponding yield reached during MS digestion. As discussed before, P. macerans showed a higher impact on the 66% straw digestion lasting for the 3 HRT tested and increased the average acid production by 16.1%.

3.4 Effect of bioaugmentation on culture viability

The electro-optical measurement of cell polarizability by FDAP is an indirect measurement of the cultures’ physiological state. High levels of FDAP in a dark fermentation process have been associated with the hydrolytic activity, while a decrease in FDAP might indicate acidic stress conditions [31]. Higher FDAP usually corresponds to cells with a high energy level, a pre-requisite for fast synthesis rates. However, an increased FDAP with higher straw content (see Table 4) is related to a reduction of acidification stress as indicated by the lower acidification, total SCCA concentrations, and higher pH value. In contrast, an increased FDAP during bioaugmentation correlates with a higher metabolic activity due to an increased substrate availability and concomitantly higher metabolite concentration in the media. In the actual study, the average FDAP increased for all bioaugmentation trials in all feedstocks. As it can be seen in Fig. 6, bioaugmentation increased FDAP levels after addition especially when applying 66% straw (w/w). It also contributed to a stabilization of the FDAP values. Positive effects on cell viability of bioaugmentation can be seen during addition and the first recirculation phase, while in the second recirculation phase in weeks 59 and 60 with 66% straw (w/w) stronger fluctuations appeared again, which may be explained by the observed lower effect of P. glucanolyticus addition on the hydrolysis due to washout or limited survival. On the contrary, P. macerans addition still showed increased FDAP levels in the second recirculation phase (weeks 68 and 69), suggesting that it had a longer lasting effect on the cultivation. Our suggestion is supported by the higher increase of sCOD and SCCAs with P. macerans addition in 66% straw (w/w). Thin-sludge recirculation apparently also supports cell viability in the PFR, as immediate decreases in FDAP can be seen when it was stopped.

Fig. 6
figure 6

Measurement of FDAP during the digestion of MS (a) and 66% straw (b). Sampling points after bioaugmentation are marked with arrows and the phase without recirculation is indicated by “No Rec.” The average measurement between the ports of the PFR is depicted

Full size image

The slope of polarizability across frequencies can also be attributed to performance changes within the culture. Haberman et al. [59] measured FDAP in AD fermenting MS and found a decreasing signal at higher frequencies. A similar pattern was measured in our current study, indicating that this slope is typically for anaerobic, mixed microbial communities. In contrast, Junne et al. [60] found that the increase of polarizability at low frequencies is typical for the acidogenic growth phase under substrate excess in Clostridium acetobutylicum. The results of our current study indicate that, in most phases, cells are in a vital state, but not under substrate excess conditions or close to maximum growth rates. This is also not to be expected under the given cultivation conditions anyway. Noteworthy to mention is that bioaugmentation with Paenibacillus spp. increased the viability and stability of the whole microbial community, which was further enhanced also over longer time by the application of thin-sludge recirculation.

4 Comparison of effects on the overall process performance

Bioaugmentation studies in AD focus mainly on the direct effect on methane yield, while bioaugmentation of the first stage has been rarely investigated. It was shown that an increase of sCOD and SCCAs in the first-stage results in higher and more stable methane production in AD [10, 61]. As discussed before, bioaugmentation in continuous processes often shows only short-term effects for a few days after addition due to washout and limited survival of the organism. Using the bioaugmentation strategy and recirculation pattern in the PFR as in our current study, positive effects of two microbial additions were seen for 2 HRT with pure MS and 30% (w/w) of straw as feedstock and possibly longer, if recirculation was not stopped. Martin-Ryals et al. [61] achieved 25–38% higher hydrolysis and 25–55% higher acid production with daily bioaugmentation into the acid phase of sewage sludge digestion. Similar results for MS and 30% straw (w/w) could be shown here (26–63% higher hydrolysis rate, 25–42% higher acid production), using only two additions and the recirculation strategy. However, bioaugmentation with any of the Paenibacillus strains could not increase hydrolysis of 66% (w/w) of straw with high recalcitrance. The higher pH value during 66% straw digestion should provide a better environment for Paenibacillus spp. and enhance activity of extracellular hydrolytic enzymes, but no significant advantage of bioaugmentation was seen at a high straw content in the feedstock. Growth of different P. glucanolyticus strains on pure lignin compounds was shown to be possible [20]. However, growth was very slow with generation times exceeding 100 h. Moreover, cinnamic acid, a degradation product of lignocellulose, was suspected to be toxic for P. glucanolyticus [20]. Also in the current study, hydrocinnamic acid was detected especially after bioaugmentation by GC measurements (see supplementary material, Figure S3). Hydrocinnamic acids are hydroxy derivatives of cinnamic acids that can be released during lignocellulose hydrolysis, of which many are known to have inhibiting effects on AD [62]. Possibly, higher concentrations of released lignocellulosic compounds at 66% of straw, a low pH value, and very slow growth inhibited further hydrolytic activity of the Paenibacillus spp. It has been found that easily degradable polysaccharides like starch and hemicellulose are metabolized before recalcitrant crystalline lignocellulosic compounds in mixed cultures [42]. The same applies for the Paenibacillus spp. in our fermentation: a rather increased hydrolysis of MS occurred in parallel with a higher amount of hemicellulose compared to the lignin-rich straw feedstock. Hemicellulose can have a rather small digestion efficiency of 29% in AD of MS [23]; our current study indicates that bioaugmentation may be a possibility to improve this. The hydrolytic activity of both studied organisms was confirmed on various pure soluble polymers from lignocellulose in laboratory trials. Little is known, however, about their ability to grow on natural crystalline lignocellulosic structures as straw and MS. Decrystallisation of the substrate is a major bottleneck in cellulose hydrolysis and can only be achieved by truly cellulolytic microorganisms like Clostridium spp., among others [63]. Clostridia like C. thermocellum or C. cellulolyticum are ubiquitous and common in AD processes. Their multi-enzyme complex, the cellulosome, can efficiently degrade crystalline lignocellulosic structures, releasing the formerly bound polymers of hemicellulose, lignin, and cellulose into the media. Cellulose degradation through Clostridia, however, is severely inhibited by pH values below 6.0–6.5, mainly due to reduced growth rates [42, 64]. Since SHR decreases at a high straw content and cannot be enhanced by bioaugmentation, it seems that cellulose hydrolysis is limited by an unfavorable pH value below 5.0 in the PFR. While the total hydrolysis and acid yields from the PFR are in a mid-range and could be further optimized, it is likely that destabilization and structural changes occurred in the feedstock making it better accessible for further degradation afterwards.

5 Conclusion

The aim of this study was to evaluate plug-flow-based hydrolysis and the effect of bioaugmentation with Paenibacillus spp. The PFR showed to be a suitable system for hydrolysis and acidogenesis of MS with straw addition reaching SHR of 258–264 gO2 kg−1VS and acid production of 97–135 gSCCA kg−1VS in continuous operation. Even without pH control, the PFR showed a stable metabolic profile of acid production yielding mainly butyrate and acetate. With a high ratio of 66% straw, an SHR of 195 gO2 kg−1VS and an acid production of 70.6 to 85.9 gSCCA kg−1VS were reached. Bioaugmentation with Paenibacillus spp. proved to be beneficial for the digestion of MS and 30% straw (w/w), increasing SHR by up to 63% and SCCA production by up to 42%, while minor effects could be seen in the digestion of 66% (w/w) of straw. The net acid production was increased among all feedstock mixtures. Overall, from our results, we can conclude that bioaugmentation with Paenibacillus spp. enhanced the fermentation process majorly by three mechanisms:

  1. i)

    Enhanced solubilization of particulate substrate as shown by higher sCOD concentrations and increased SHR

  2. ii)

    Enhanced degradation of formerly unused soluble (hemicellulose) polymers resulting in better substrate availability, higher acid production, and higher culture vitality as measured by FDAP

  3. iii)

    Enhanced butyrate production by the generation of precursors for butyrate-producing bacteria.

Thin-slurry recirculation further increased and prolonged the effects of bioaugmentation for the 2 HRT that were tested and longer lasting effects are probable for MS and 30% straw (w/w). These longer positive effects of bioaugmentation indicate a suitability of the chosen microorganism for the survival and enzyme production in the PFR, because hydrolytic enzymes that were released from the previous Paenibacillus sp. Monocultivation would have mostly been deactivated by endogenous proteases in the hydrolysis’ culture broth after 24 h [65]. However, long-term survival of the augmented organisms in the PFR with lignin-rich feedstock is not probable, as the wear-off of positive effects can be seen by the time-dependent decrease of FDAP and conductivity measurements, which correlated well with the total SCCA concentration, within 2 HRT after bioaugmentation. Nonetheless, Paenibacillus sp. addition boosted microbial hydrolysis and acid production in the PFR by enhancing the indigenous microbial community. Compared to routine bioaugmentation, significantly less culture had to be used to achieve comparable results. This PFR setup can serve as a first stage for two-stage AD processes, in order to broaden the applicable feedstock spectrum and stablize methane synthesis, and thus make methane production more flexible. Bioaugmentation, together with the proposed reactor system, can enhance yields of recalcitrant biogenic residues, and thus pave the way for a good integration of AD and other bioprocesses into a circular bioeconomy.

It remains to be investigated what frequency of addition would be required to maintain enhanced hydrolysis and acidogenesis on a long term and whether bioaugmentation with a mixture of different species would lead to a further improvement when a high content of straw is applied as feedstock. For better insights into microbial changes during bioaugmentation and lignocellulose digestion at a low pH value, a microbial consortium determination via high-throughput 16 s RNA gene amplicon sequencing should be conducted at selected time points.