Introduction

Managed aquifer recharge (MAR) has been widely used to mitigate the worldwide water crisis, particularly in regions with water scarcity and uneven distribution of water resources (Dillon 2005; Kim et al. 2010). MAR has many advantages such as groundwater resource augmentation, seawater intrusion prevention, water storage, and water quality improvement through soil-aquifer treatment (Aoki et al. 2005). Virtually all engineered operations experience some degree of clogging that could markedly decrease the efficiency of injection wells by a drastic reduction in the saturated hydraulic conductivity (Ks) of the aquifer (Pavelic et al. 2011). Researchers have documented several causes of clogging, generally classified into physical, chemical and biological or a combination thereof (Hoffmann and Gunkel 2011; Pavelic et al. 2011). Physical mechanisms include sediment compaction (Pérez-Paricio and Carrera 2001), suspended solids in the recharge water infiltrating into the soil/sediments (Pavelic et al. 2011), and release of entrapped air bubbles (Rinck-Pfeiffer et al. 2000). Chemical clogging primarily results from precipitation of calcium carbonates (Baveye et al. 1998). Additionally, bioclogging can stem from the accumulation of bacterial cells and production of extracellular polymeric substances (EPS; Thullner 2010).

Previous studies that have focused on aquifer bioclogging mechanisms have demonstrated that the accumulation of bacterial cells in pore throats or at grain contacts causes a substantial reduction in the Ks of aquifer materials (Vandevivere and Baveye 1992). Studies have also reported that EPS produced by bacteria play a significant role in the clogging process. These exopolymers could affect Ks either by increasing the fluid viscosity or by decreasing the pore volume (Vandevivere and Baveye 1992). Xia et al. (2016) showed that two types of EPS—i.e., loosely bound EPS (LB-EPS) and tightly bound EPS (TB-EPS)—had different effects on Ks reduction in porous media and concluded that, compared to TB-EPS, LB-EPS led to a larger reduction in Ks during the bioclogging process. Additionally, poorly soluble gaseous compounds such as nitrogen and methane, can also cause bioclogging if the gas bubbles are sufficiently large to be trapped in the pore network (Kandra et al. 2015). Furthermore, other organisms such as fungi (Seki et al. 1996), have been observed in clogged porous media.

In most MAR systems, different water sources are used, e.g., river water, treated effluent, stormwater, desalinated seawater, rainwater, and even groundwater from other aquifers, depending on seasonal temperature variation. Temperature variation could cause a change in aquifer hydraulic conductivity because pore water viscosity is temperature dependent (Constantz 1998). Temperature also influences chemical reactions. Prommer and Stuyfzand (2005) demonstrated that, during managed recharge of aerobic water into an anaerobic aquifer, the oxidation of pyrite was temperature controlled. Additionally, temperature affects microbial growth and activity (Warren and Bekins 2015), hence the bioclogging in aquifers.

Previous studies have revealed that over 99% of bacteria present in subsurface environments cannot be cultured in the laboratory because artificial media fail to mimic completely the comprehensive abiotic and biotic conditions required for bacterial growth (Alain and Querellou 2009), and because bacteria cannot immediately adapt to the changes in conditions (Pham and Kim 2012). Therefore, culture-dependent bacterial populations likely underestimate the true diversity of the microbial community structure. Denaturing gradient gel electrophoresis (DGGE) of polymerase chain reaction (PCR)-amplified 16S rDNA gene fragments, a powerful molecular method, allows rapid detection of microbial community changes and provides accurate information regarding the distribution and composition of microbial species (Aydin et al. 2015). Combining culture-independent and culture-dependent methods to investigate bacterial communities offers an exact description and explanation of the bioclogging process in aquifers.

It is known that temperature can significantly affect microbial growth and activity (Warren and Bekins 2015); however, to the best of the authors’ knowledge, the quantitative effect of seasonal temperature variation on the bioclogging process and bacterial metabolism in aquifers has not been thoroughly elucidated to date. Therefore, this study attempts to investigate the effect of recharge water temperature on bacterial growth and metabolism and to understand the differences in the bioclogging process and temperature-variation-induced mechanisms in MAR systems by a laboratory culture experiment. In this study, a series of bacteria-inoculated sand columns, flowing with sterilized nutrient solution at different temperatures, were used in winter (i.e., March 2014) and summer (i.e., August 2015) to simulate the bioclogging process. The objectives of this work are: (1) to investigate the temporal variation of the relative saturated hydraulic conductivity (Ks′) in the sand columns during two seasons; (2) to investigate the bacterial growth and production of total EPS, LB-EPS and TB-EPS; (3) to compare the bioclogging process and the bacterial community in the two seasonal experiments (for a listing all terms and their definitions see Table 1).

Table 1 Abbreviations used in this paper
Full size table

.

Materials and methods

Porous media

Sand from a natural aquifer near the Dagu River (36.38024° N, 120.12089° E) in Jiaozhou city, Shandong province, China, was used as a model porous medium. After homogenizing, sieving and excluding the gravel fraction >2 mm, the sand was washed seven times with deionized Milli-Q water to remove clay or silt. Then, it was soaked in 0.25 M HCl for 24 h to remove trace metals, followed by treatment with 0.25 M NaOH and deionized Milli-Q water to adjust the pH to 7.0–7.2 (Yang et al. 2013). Between the acid and base treatment, the sand was washed using Milli-Q water until a neutral pH. Finally, it was baked at 550 °C for 2 h to remove organic matter according to a loss on ignition method (Qian et al. 2011). The particle size was analyzed according to the standard soil test method (Ministry of Water Resources of the People’s Republic of China 1999). The median diameter of the sand was 539 μm, and the particle-size-distribution curve is presented in Xia et al. (2016). The sand was analyzed for mineralogy using X-ray Diffraction (XRD, D/max-rB, Rigaku Co., Tokyo, Japan), and the result indicated it was dominated by quartz (86.4%), with minor compositions from feldspar (7.2%), dolomite (3.4%) and hornblende (3.0%).

Bacterial inoculum and nutrient solution

Bacterial inoculum was collected from a severely clogged recharge well near the Dagu River (36.38024° N, 120.12089° E) in March 2014 and August 2015, to closely simulate actual field aquifer bioclogging. The clogged recharge well was back washed by pumping at a rate of 833 L/min. Then, the discharge water was stored at 4 °C and immediately transported back to the laboratory. Before inoculation of each column, the discharge water was settled for 30 min, and the supernatant was used as the bacterial inoculum.

A synthetic nutrient solution containing glucose as the sole carbon source was used to stimulate bacterial growth. The nutrient solution consisted of (in milligrams per liter of deionized water): glucose, 58; NaCl, 5000 (0.5%, w/v); NH4+-N, 0.5; total phosphorous (TP), 0.01; MgSO4·7H2O, 45; CaCl2·2H2O, 20, and a trace elements solution, 1 ml/L (Xia et al. 2014a). The pH was adjusted to 7.0–7.2. Before being injected into the columns, the nutrient solution was autoclaved at 121 °C for 15 min and then oxygenated by fully stirring.

Column setup

Acrylic glass columns, each 22-cm long with an inner diameter of 5 cm, were used for the percolation experiments (Fig. 1a). Piezometers were installed at 0, 2, 4, 6, 12, and 18 cm from the inlet (i.e., P1 in Fig. 1a) of the columns to monitor the hydraulic gradient. The columns were wet-packed with the prepared sand in small increments (~2 cm) using mild column vibration to minimize any layer heterogeneity or air entrapment. Fine stainless steel screens were placed at each end of the columns to prevent sand grains from creeping out. The packed columns were carefully sterilized via exposure to ultraviolet light under the wavelength of 254 nm. Parameters such as column volume, soil porosity and pore volume are presented in Table 2. Soil porosity was calculated based on the following equation (Danielson and Sutherland 1986),

$$ {S}_{\mathrm{t}}=1\hbox{-} {\rho}_{\mathrm{b}}/{\rho}_{\mathrm{p}} $$
(1)
Fig. 1
figure 1

Experimental design of column setup: a schematic diagram of experimental setup; b schematic illustration of experimental design

Full size image
Table 2 Values for the sand column parameters in these experiments
Full size table

where St represents soil porosity, ρb the bulk density, ρp the particle density. In the study, the bulk density (ρb) and particle density (ρp) were measured to be 1.63 and 2.65 g/cm3, according to the core method and pycnometer method (Blake and Hartge 1986), respectively.

For the inoculating procedure, the packed columns were injected downward with five pore volumes of the bacterial inoculum and then left for 12 h to foster cell attachment. Afterward, all the tubing and other parts in contact with the nutrient solution were replaced by sterilized ones to avoid contamination. A constant-temperature water bath (DC-0530, Baidian Yiqi Co., Ltd., Shanghai, China) was used to maintain the temperature of the sterilized nutrient solution (in the header tank). A peristaltic pump (BT100-2J, Longerpump Co., Ltd., Baoding, China) was used to pump the sterilized nutrient solution (in the header tank) to the constant head tank. The flow was vertically run from the top to the bottom of the inoculated columns under a constant hydraulic head loss. After flowing through the entire column, the nutrient solution was discharged to avoid column contamination.

Experimental design

Identical columns, arranged as eight duplicates, were involved in two seasonal percolation experiments: 16 columns representing winter (i.e., inoculums acquired in March 2014) and 16 columns representing summer (i.e., inoculums acquired in August 2015) (Fig. 1b); thus, there were 32 columns in total. All the sand columns were flushed with the sterilized nutrient solution for the predetermined times, namely, 0, 48, 96, 144, 192, 240, 288, or 336 h in March 2014 and 0, 4, 8, 10, 12, 14, 16, or 22 h in August 2015. Note, all the experimental conditions were identical except that the temperature of the nutrient solution was controlled at ~5±2 and ~15±3 °C in winter and summer, respectively. After flowing for the predetermined time, the flow rates and hydraulic head of the two parallel columns were measured to calculate the Ks and Ks′ values and then the columns were dismantled and sand samples were collected for biochemical analyses. Each sampling was performed in triplicate.

Saturated hydraulic conductivity

The saturated hydraulic conductivity Ks (cm/s) of the different layers of the sand columns could be calculated based on Darcy’s flow equation (Darcy 1856):

$$ Q=A\times {K}_{i\mathrm{s}}\frac{{\left(\Delta H\right)}_i}{L_i} $$
(2)

where the subscript i refers to the i-th layer from the inlet, delimited by two piezometers. For instance, i = 1 represents the piezometers installed at the inlet and the first 2 cm of the columns (i.e., P1 and P2 in Fig. 1a). In this equation, Q equals the flow rate (ml/min), A the cross-sectional area of the columns (cm2), (ΔH)i the hydraulic head difference of the i-th layer (cm), and Li the distance between the two piezometers at which the hydraulic head was measured (cm). In this study, the results are shown in terms of the relative saturated hydraulic conductivity Kis′ = Kis/Kis0, which is the ratio of the saturated hydraulic conductivity at the predetermined time t in comparison to its initial value.

Bacterial cell enumeration

After flowing for the predetermined time, the feed water valves of the two parallel columns were closed, and the excess nutrient solution drained off. Then, all the tubing and other parts in contact with the columns were carefully removed. The sand in the first layer of each column (i.e., the inlet to 2 cm) was collected and gently vortexed to mix. Then, weighed sand was sampled for bacterial cell enumeration, EPS quantification and DNA extraction.

Small, 1-g subsamples of the sand were weighed and transferred aseptically with a sterile spoon to a 10-ml sterile centrifuge tube. Each sampling was performed in triplicate. The samples were mixed with 5 ml of sterile phosphate buffer solution (PB, 10 mM, pH 7.4) and 0.5-g sterilized glass beads and then sonicated at 40 kHz, 30 W for 30 s. Then, the mixture was vortex oscillated at 4 °C for 1 min. The bacterial cell numbers was measured by serial diluting the suspension and plating it on a nutrient broth (NB) agar medium (Skill Bio Co., Ltd., Beijing, China): gelatin peptone (10 g), beef extract (3 g), sodium chloride (5 g) and agar (15 g) in 1-L deionized water at a final pH 7.0–7.2 (He et al. 2014). The plates were incubated at 37 °C for 72 h, and then visible bacterial colonies were manually counted. The data were reported as colony-forming units (CFUs) per gram of sand samples.

EPS extraction and quantification

A heating extraction method (Morgan et al. 1990) was modified to include a mild step for extracting LB-EPS and a harsh step for TB-EPS. After flowing for the predetermined time, the columns were carefully dismantled and sand from the first layer of each column (i.e., the inlet to 2 cm) was sampled in triplicate. In brief, a 5-g subsample of the sand was weighed and transferred aseptically with a sterile spoon to a 250-ml sterile plastic bottle. Then, the sand was mixed with 100-ml sterile PB solution and 2-g sterilized glass beads and treated by ultrasound at 40 kHz, 30 W for 30 s. After that, the suspension was harvested by centrifugation at 550 g for 15 min. Then, the supernatant was discharged, and the pellet was washed twice with the sterilized PB solution. Subsequently, the pellet was resuspended with 30-ml PB and centrifuged at 4,000 g and 4 °C for 15 min. The organic matter contained in the supernatant was considered to be LB-EPS. The pellet after LB-EPS extraction was further used for TB-EPS extraction. The residual pellet in the centrifuge tube was resuspended using 30-ml sterilized PB and then heated to 70 °C for 30 min in a water bath. After heating, the suspension was centrifuged at 4,000 g and 4 °C for 15 min. The organic matter in the supernatant was assumed to be TB-EPS.

After the extraction step, the organic matter was measured for polysaccharides and proteins contents. The polysaccharides (PSs, expressed in μg/g sand) were analyzed using a UV/VIS spectrophotometer (SQ2800, Unico, USA) according to the phenol-sulfuric acid method (Frølund et al. 1996) with a glucose standard (Kermel Chemical Reagent Co., Ltd., Tianjin, China). The proteins (PNs, expressed in μg/g sand) were assayed using the UV/VIS spectrophotometer according to the Bradford assay (Sharma and Babitch 1980) with albumin from bovine serum (Sinopharm Chemical Reagent Co., Ltd., Shanghai, China) as a standard. The sum of the PS and PN contents were assumed to be EPS.

Bacterial community analysis

DNA extraction

After flowing for 336 h in March 2014 and 22 h in August 2015, two parallel columns were carefully dismantled and sampling of each column was performed. Then 0.25-g subsamples of the sand in the first layer of each column (i.e., the inlet to 2 cm) were transferred aseptically with a sterile spoon to a 15-ml sterile centrifuge tube. Next, the samples were mixed with 10-ml sterile PB, and 0.5-g sterilized glass beads; these samples were then sonicated at 40 KHz, 30 W for 30 s. Next, the mixture was vortex oscillated at 4 °C for 1 min. Then, the suspension was transferred to a sterilized centrifuge tube and centrifuged at 550 g for 15 min. The supernatant was discarded, and DNA was extracted using the Power Soil DNA Isolation Kit (MO BIO Laboratories, Inc., Carlsbad, CA, USA) following the manufacturer’s instructions. The pellet was transferred to the supplied Power Bead tube with 60 μl of the kit-supplied solution C1. The tube was incubated at 60 °C for 10 min and then shaken horizontally at maximum speed for 10 min using the MO BIO vortex adapter. The remaining steps were performed following the kit’s “Experienced user protocol.” The extracted DNA was stored at –80 °C until used.

PCR amplification and DGGE

The V3 region of the 16S rDNA gene was amplified by PCR using a universal bacterial primer set (forward primer, 5’-ATTACCGCGGCTGCTGG-3’; reverse primer, 5’-TGGCGGACGGGTGAGTAA-3’). PCR was performed in thin-walled tubes with a GeneAmp PCR system 9700 (Applied Biosystems, USA). One microliter of the template DNA was added to a reaction mix (final volume, 50 μl) containing dNTP (2.5 mM, 4 μl), 10 × PCR buffer (2.5 mM, 5 μl), forward and reverse primers (10 pmol, each 2 μl), Tap DNA polymerase (5 U/μl, 0.25 μl, TaKaRa, Dalian, China) and dd H2O (35.75 μl). The PCR reaction was performed as follows: (1) initial denaturation at 94 °C for 5 min; (2) 35 cycles of denaturation at 94 °C for 40 s, annealing at 55 °C for 40 s, and extension at 72 °C for 1 min; (3) final extension at 72 °C for 5 min. The DGGE of PCR products was carried out by using a D-Code System (Bio-Rad, USA). Gels were prepared and run in the following conditions: 8% (w/v) polyacrylamide, 1 × TAE buffer, linear gradient from 40 to 60% denaturant (where 100% denaturing agent was a mixture of 7-mol/L urea and 40% v/v deionized formamide). Electrophoresis was performed for 9 h at 150 V and 60 °C. After that, gels were stained with ethidium bromide and visualized with UV transilluminator (UVP, Inc., San Gabriel, CA, USA).

Analyses of DGGE spectra and bacterial diversity

DGGE banding patterns show the number and intensity of individual bands in each lane (Huang et al. 2012). Each DNA band represents a specific bacterial species in the community. The number of the bands directly reflects the genetic diversity of the bacterial community in samples. The higher the band number is, the more diverse the bacterial community is. In this study, Quantity One image analysis software (version 4.6.2, Bio-Rad Laboratories, Inc., USA) was used to quantify the number and intensity of the bands according to the manufacturer’s instructions.

The Shannon-Wiener index (H′) has been commonly used to characterize bacterial diversity in a community (Deng et al. 2012). In this study, the index was calculated based on the number and intensity values of the individual bands to compare the species richness and evenness of the dominated bacteria community in different seasonal experiments. The equations are as follows (Gong et al. 2012):

$$ {H}^{\prime }=-\sum \limits_{i=1}^S\left({p}_i\right)\left({\log}_2{p}_i\right)=-\sum \limits_{i=1}^S\left({N}_i/N\right){\log}_2\left({N}_i/N\right) $$
(3)
$$ E=\frac{H^{\prime }}{H{\prime}_{\mathrm{max}}}={H}^{\prime }/\ln S $$
(4)

where H′ represents the Shannon-Wiener index, pi the relative abundance of the i-th species, Ni the intensity of band i, N the sum of the intensities of the bands in a lane, E the evenness of the dominated bacterial community, and S the richness, as determined from the number of the dominant bands in each lane.

Cloning, sequencing and phylogenetic analysis

The dominant bands in the DGGE gels were excised. Each excised piece was washed twice with 1 ml of sterilized deionized water and then placed into a 1.5-ml sterilized centrifuge tube to reclaim the DNA using SK1135 kit (Shanghai Sangon Biotech Co. Ltd.). The PCR products were sent to Sangon Biotech Co. Ltd. (Shanghai, China) for DNA sequencing. Then, the sequences were compared to the stored sequences in the National Center for Biotechnology Information (NCBI) database (NCBI 1988) using the BLAST algorithm. The homology of the bands was analyzed, and the neighbor-joining (NJ) method was used to establish a phylogenetic tree with the computer program MEGA version 6.0 (MEGA 1993).

Nucleotide accession numbers

The 17 nucleotide sequences were deposited in the GenBank database (NCBI 1988) under accession numbers KF680873-KF680878, KF680880, KF680882, KF680884, KF680887-KF680888, KX658645-KX658647, KF680877, and KX658648-KX658649.

Data sources

The experimental data for the winter test (i.e., March 2014) came from previous work (Xia et al. 2016), and data for the summer test were obtained from the percolation experiments conducted in August 2015. Several figures from Xia et al. (2016) are redrawn to show a better comparison between the two seasonal tests in this study.

Results and Discussion

In this study, only the first layer of the sand columns (i.e., the inlet to 2 cm) was investigated, as it was where the largest reduction in Ks occurred—see Tables S1 and S2 of the electronic supplementary material (ESM). This result was consistent with the previous findings (Xia et al. 2014b). Previous researchers revealed that the first 10% of the post-inlet length of the column is the principal bioclogging zone (Yang et al. 2013).

Temporal variation of the relative saturated hydraulic conductivity

The K1s′ value was used to compare normalized clogging of the first layer (i.e., the inlet to 2 cm) for all the sand columns (Fig. 2); note, Fig. 2, as in Figs. 3– 6, presents the results of one of the duplicate experiments to avoid cluttering the figures, whereby the data showed that, although certain deviations occurred due to porous media heterogeneity, the variation patterns were consistent (data for the duplicate experiments are presented in part II of the ESM). Figure 2 indicates that declines in K1s′ occurred in both the winter and summer experiments. In winter, a mild 5% reduction in K1s′ was observed within 48 h, followed by a sharp 94% decrease in the next 192 h. Toward the end of the experiment, the decreasing trend of K1s′ lessened again. In the whole test, the K1s′ was a 99.8% reduction to the initial value. In comparison, the summer test curve displayed a similar trend with a much faster reduction in K1s′. The K1s′ value slightly decreased by 5% over the initial 4 h, followed by a large 81% reduction over the next 10 h. At the end of the percolation test, the Ks′ was reduced to 1% of its initial value. While the overall behavior was similar, the sand columns were entirely clogged within 22 h in summer compared to 336 h in winter, which indicated that rapid clogging occurred in summer.

Fig. 2
figure 2

Temporal variation of the relative saturated hydraulic conductivity in the first layer (i.e., the inlet to 2 cm) of the sand columns in the winter and summer tests

Full size image
Fig. 3
figure 3

Temporal variation of the bacterial cell numbers in the first layer (i.e., the inlet to 2 cm) of the sand columns in the winter and summer tests

Full size image

Bacterial growth and EPS production

Number of culturable bacteria

The variations in culturable bacteria of the first layer (i.e., the inlet to 2 cm) in the seasonal experiments are shown in Fig. 3. Although the culturable cell counts do not reflect the actual numbers of bacteria attached to the porous media, they are considered indirect estimates of both the abundance and activity of the attached bacteria. As shown in Fig. 3, the culturable cell numbers in both the winter and summer experiments reached maxima of ~×108 CFU/g sand, while the growth and decay periods in winter were significantly longer than those in summer. In the 336-h percolation experiment in winter, the numbers of culturable bacteria increased gradually within the initial 144 h, followed by a slightly decreasing trend in the next 192 h. The cell numbers reached a maximum value of 10.2 × 108 CFU/g sand, which corresponded to 500 times its initial value. Figure 3 also shows that the growth curve of the culturable bacteria was much shorter in summer compared to the experiment conducted in winter. After reaching a peak value of 5.4 × 108 CFU/g sand at 14 h, the total number of culturable bacteria gradually decreased as the percolation experiment progressed. This decrease implies that the higher temperature in the summer test (e.g., 15 °C) could accelerate bacterial growth rate and metabolic activity. This result agrees with Nagata et al. (2001), who concluded that the increasing temperature could favor microbial activity. Temperature governs biological metabolism by its effect on rates of biochemical reactions (Aledo and Jiménez-Riveres 2010); therefore, a favorable temperature could increase microbial activity (e.g., enzyme production, biodegradation) and membrane permeability. As the temperature decreases, cell membranes become increasingly viscous with decreasing membrane fluidity. The reduction in the fluidity of the membrane reduces transporter protein efficiency in the membrane and further influences transfer of oxygen and nutrition. This reduced efficiency finally leads to a decreased bacterial growth rate (Smet et al. 2015); however, under certain circumstances, the biochemical reaction rate may decline with increasing temperature (Aledo and Jiménez-Riveres 2010; Silverstein 2012). Aledo and Jiménez-Riveres (2010) demonstrated that there was a temperature switch point (e.g., ~30 °C for lactate dehydrogenase-B4a; ~20 °C for lactate dehydrogenase-B4b), above which the enzyme works slower due to the negative activation energy. Silverstein (2012) considered that this phenomenon occurred because of a dramatic conformational change of the enzyme (i.e., denaturation) in the temperature range of 20~30 °C.

Concentration and proportion of total EPS, LB-EPS and TB-EPS

Figure 4 demonstrates that there was a seasonal variation with generally increased LB-EPS, TB-EPS and total EPS in lower-temperature winter. In the winter test, the concentrations of LB-EPS and TB-EPS reached peak values of 540 and 155 μg/g sand, respectively, these values were much higher than those in summer (Fig. 4a,b). For the total EPS, the concentration increased sharply during the initial 192 h but fell slightly in the subsequent 144 h, with a peak value of 694 μg/g sand at 192 h in winter (Fig. 4c); however, total EPS reached a maximum of 417 μg/g sand within the initial 10 h in the summer experiment, which was lower than that in the winter experiment (Fig. 4c). This observation is in agreement with results reported by Wilén et al. (2008), who found that the total amount of EPS in flocs was significantly higher in the cold winter compared to the warm summer. Zhang et al. (2015) obtained negative correlations between the temperature and content of LB-EPS (R2 = −0.794, p < 0.01) and TB-EPS (R2 = −0.596, p < 0.01), which indicated that lower temperatures led to more LB-EPS and TB-EPS.

Fig. 4
figure 4

Temporal variation of the concentrations and proportion of different EPS in the first layer (i.e., the inlet to 2 cm) of the sand columns in the winter and summer tests: a–c the concentrations of the LB-EPS, TB-EPS and total EPS, respectively, and d the ratio of the LB-EPS to the total EPS

Full size image

In this study, the reduced bacterial activity coupled with stimulated EPS production occurred in the winter test. This inconsistency may occur for two reasons. First, EPS production is a microbial response to their living environment (Li and Yang 2007). Microorganisms tend to produce more EPS to protect themselves from harsh environments such as low temperature or the presence of toxic compounds (Lyko et al. 2008). Although lower temperature inhibits a large number of bacteria, the surviving ones produce more exopolymers (Yang et al. 2013). Secondly, the seasonal differences in the proportion of polysaccharides and proteins in EPS may be responsible, as explained in the following section.

Figure 4d shows that the proportion of LB-EPS varied considerably in different seasons. The LB-EPS accounted for 31–80% of the total amount of EPS in winter, while they were the predominant component of the total EPS in summer (e.g., 91–95%). These results indicated that, compared to the TB-EPS, bacteria were inclined to secrete a significant amount of LB-EPS in a favorable living environment such as suitable temperature in summer; however, when encountering adverse conditions, bacteria altered their double-layer structure and tended to produce more TB-EPS. Compared to the LB-EPS, the TB-EPS were reported to be the primary surface by which cells protect against dewatering, low temperature, and nutrient shortage by strong interactions (Priester et al. 2006).

Concentration and proportion of major components in the total EPS

Polysaccharides and proteins are two major components of EPS (Chen et al. 2013; Han et al. 2017). Figure 5 shows the quantitative differences in polysaccharides and proteins in the total EPS in different seasonal experiments. It is apparent that the production rate of both proteins and polysaccharides was faster in summer than in winter, which caused the same characteristic of EPS variation. The proteins concentrations were similar in summer and winter, and both of the maxima reached ~100 μg/g sand (Fig. 5a). Zhang et al. (2015) noted that extracellular proteins secreted by bacteria were enzymes, and the slightly higher concentration in summer corresponded to the enhancement of microbial activity; however, the concentration of exopolysaccharides in the winter, which reached ~600 μg/g sand, was nearly twice that in the summer (~320 μg/g; Fig. 5b). This difference is likely because the polysaccharides, which are highly biodegradable polymers, are more likely to be easily degraded in the presence of high exoenzymes in the summertime (Zhang et al. 2015).

Fig. 5
figure 5

Temporal variation of the a proteins (PNs) and b polysaccharides (PSs) concentrations in the total EPS, and c the PSs/Total EPS ratio, in the first layer (i.e., the inlet to 2 cm) of the sand columns in the winter and summer tests

Full size image

Figure 5c shows further differences in the proportions of polysaccharides to total EPS. Proportions in both the summer and the winter are above ~58%, especially in the winter, when they are over 90% on average. This result indicates that the polysaccharides are dominated by the production of total EPS. During the early stage of biofilm formation, bacterial cells tend to produce more exopolysaccharides than other exopolymers, which would promote cell adhesion onto the surfaces of porous media (Lopes et al. 2000). According to Vandevivere and Kirchman (1993), the polysaccharides/proteins ratio could be five-fold greater for attached cells than for free-living cells. Bacteria mediate cell cohesion and adhesion and maintain the structural integrity of the biofilm matrix by the production of exopolysaccharides (Tsuneda et al. 2001); therefore, polysaccharides are predominant in total EPS as the experimental results show.

In the winter test, the proportions of polysaccharides to total EPS were approximately 82–96%. The accumulation of the polysaccharides led to a significant amount of the total EPS. However, the proportions in the summer test (58–85%) had a ~20% reduction relative to those in the winter (82–96%). This difference may occur because more proteins were secreted during the period of increasing microbial activity, and the polysaccharides were biodegraded in the presence of exoenzymes (i.e., proteins) in the summer (Zhang et al. 2015).

Seasonal variations in the bioclogging process of the sand columns

The variations in the bacterial cell numbers, TB-EPS and LB-EPS concentrations versus the relative saturated hydraulic conductivities in the two seasonal experiments are shown in Fig. 6. The bioclogging processes were divided into four stages, regardless of the experiment season. In winter (Fig. 6a–c), the bacterial cell numbers and the two EPS concentrations changed with the decreasing relative saturated hydraulic conductivities. During stage I (i.e., 0–48 h), the K1s′ value decreased by 5%, with the bacterial cell numbers increasing from 6.39 to 7.10 log10 CFU/g sand, while the concentrations of TB-EPS and LB-EPS had almost no change. This outcome implies that the initial decline in K1s′ resulted from the bacteria growth. For stage II (i.e., 48–144 h), the K1s′ decreased by 67%, with increased in both the cell numbers and the two EPS concentrations. However, during stage III (i.e., 144–192 h), as the bacterial cell numbers and the TB-EPS began to decline after reaching their maxima, the concentration of LB-EPS further increased from 314 to 540 μg/g sand, which could be responsible for the further decrease in K1s′ (17%). At stage IV (i.e., after 192 h), complete clogging occurred with decreases of all three factors (the cell numbers, the TB-EPS and the LB-EPS), suggesting there might be other unknown influences as well.

Fig. 6
figure 6

Relationships between the K1′ value of the sand columns and the bacterial cell numbers, TB-EPS, and LB-EPS concentrations in two seasonal percolation experiments: a−c represent the winter test and d–f the summer test

Full size image

The summer bacterial cell numbers were not substantially different to those in winter. However, compared to winter, the EPS concentrations in summer changed significantly, especially the TB-EPS, which was relatively stable and close to zero during the summer bioclogging process. The differences in the clogging process between summer and winter mainly occurred in stage III (Fig. 6d–f). At stage I (i.e., 0–4 h), the K1s′ decreased by 8% due to the increasing cell numbers from 6.15 to 6.82 log10 CFU/g sand, even with no obvious rise of the TB-EPS or the LB-EPS. During stage II (i.e., 4–10 h), the growth of bacterial cells and production of LB-EPS led to a 39% reduction in K1s′, with no obvious increase of the TB-EPS, a result different from the pattern in the winter test. Then, a continuous reduction in K1s′ was observed at stage III (i.e., 10–14 h), with the cell numbers increasing from 7.63 to 8.73 log10 CFU/g sand. There was no increase in the TB-EPS or the LB-EPS, again somewhat different from the winter results. Correspondingly, the K1s′ value further decreased by 10% due to other unexplained reasons in stage IV (i.e., after 14 h).

The results demonstrate that, although the variation patterns were similar in winter and summer during stages I and II of the bioclogging process, they were somewhat different in stage III. The major contributors in stage III were the bacterial cell numbers in summer and LB-EPS in winter. In the summer experiment, the concentration of TB-EPS leveled off to an almost constant value during the entire bioclogging process and contributed little to the reduction of K1s′, which was also different from the situation in the winter test.

Seasonal variations in the bacterial community

The bacterial community structure

The intensity of the DGGE bands can indicate the abundance of the bacterial community (Huang et al. 2012). Figure 7 and Table 3 both show that there were some differences in the bacterial community structure between the two seasonal experiments. Figure 7a demonstrates that the dominant bacteria in the winter experiment were at bands 3, 4, 5, 7 and 9, which corresponded to Janthinobacterium sp., Rahnella aquatilis, Tolumonas auensis, Acidovorax radicis and Aquabacterium commune, respectively (Table 3). By contrast, four predominant bands (i.e., bands 12, 15, 16 and 17) were present in the summer test (Fig. 7b), corresponding to Brevundimonas diminuta, T. auensis and Acidovorax radicis (Table 3). The taxonomic assignment was also confirmed using the phylogenetic tree in Fig. 8. By phyla taxonomic level, Proteobacteria dominated the bacterial community in both experiments, with the average relative abundances of 91.6% in winter and 84.2% in summer, respectively (data presented in Tables S3 and S4 of the ESM). At the level of class, Betaproteobacteria accounted for 51.4% of the classes presented in the winter experiment, following by Gammaproteobacteria, Alphaproteobacteria and Deltaproteobacteria, with relatively low proportions of 26.5, 7.3 and 6.7%, respectively. Similarly, Betaproteobacteria was the dominant class in the summer experiment with the proportion of 37.9%, following by Gammaproteobacteria and Alphaproteobacteria, which accounted for 30.9 and 15.7%, respectively (data presented in Tables S3 and S4 of the ESM).

Fig. 7.
figure 7

Denaturing gradient gel electrophoresis fingerprints (i.e., DGGE) of the V3 variable region of the bacterial 16S rDNA gene for bacteria of the two seasonal experiments

Full size image
Table 3. Strains identified using 16S rDNA sequencing fragments from dominant DGGE bands of total bacterial community DNA extracted from the experimental samples
Full size table
Fig. 8.
figure 8

Evolutionary relationship of the bacterial 16S rDNA gene sequences for the bacteria of the two seasonal experiments. The phylogenetic tree was constructed with the neighbor-joining method. Bootstrap values less than 50% (based on 1,000 bootstrap re-sampling) were already hidden at nodes

Full size image

Note that several dominant bacteria, including Rahnella aquatilis, T. auensis in winter and Brevundimonas diminuta, Tolumonas auensis in summer, were anaerobic gram-negative bacteria and primarily presented in urine. This implies that the clogging induced bacteria that originated from anthropogenic sources, since the sampling site was located downstream of the Dagu River, which flows through many cities (e.g., Zhaoyuan, Laixi, Pingdu, etc.) and has been severely impacted by human activities (Zhang and Xu 2009).

Diversity of the bacterial community

The diversity of the bacterial community in the seasonal experiments was analyzed based on the DGGE analysis of the amplified partial 16S rDNA genes. The Shannon-Wiener index (H′) provides the basic information on the shift in the bacterial community structure in regard to bacterial diversity. Table 4 shows a 55% increase of H′ in winter, which indicates a higher bacterial diversity under the colder temperature; meanwhile, the winter samples showed a larger number of dominant bands (S score) and higher evenness (E score) compared to summer. It was expected that although the favorable summer temperature enabled drastic short-term bacterial growth (Fig. 3), it would lead to a less diverse and less evenly distributed bacterial community structure. An increase of bacterial diversity was observed in the winter experiment, which revealed that the low temperature tended to promote diversity and a more even distribution of the bacterial community.

Table 4. Bacterial diversity indices calculated from the dominated DGGE band patterns
Full size table

Implication of the seasonal bacterial community

The results indicated that some differences occurred in the bacterial community structure and diversity during the winter and summer experiments, particularly the dominant bacteria. The differences may stem from two possible reasons. On the one hand, the bacterial composition in the seasonal inoculums could be different due to in situ sampling, which were induced by complex field conditions (e.g. temperature, recharge water, etc.). On the other hand, a dynamic shift in bacterial community structure could occur due to changes in recharge water temperature during the bioclogging process in the seasonal experiments. Since the temperature-induced change of the bacterial community structure is a long-term process (He et al. 2014), it is inferred that the differences in the bacterial community in the clogged columns primarily resulted from the former reason.

The differences in the seasonal inoculums might also have a potential influence on the bacterial growth and EPS production; however, the influence is inferred to be weaker than that of temperature. Figure 6 shows that the bacterial growth pattern and the bacterial cell numbers were similar in the two seasonal experiments, while the difference in the growth rates was a multiple of ten. As discussed in section ‘Number of culturable bacteria’, the higher temperature could significantly accelerate metabolic activity and induce the much faster bacterial growth rate in the summer test. In addition, the dominant bacteria in both experiments had the ability to produce high amount of EPS (e.g., Badireddy et al. 2008; Douterelo et al. 2013; Liao et al. 2017), whereas EPS in summer were much lower than those in winter especially for the TB-EPS. It also implies that the differences that occurred in EPS production might be mainly induced by recharge water temperature. The aforementioned qualitative analysis inferred that the temperature could have a dominant influence on the bacterial growth and EPS production. However, to quantify the respective influences of temperature and the bacterial community, there needs to be further controlled experiments, which is the goal of the authors’ future research.

Conclusions

A series of bacterial inoculated sand columns flowing with sterilized nutrient solution at different temperatures was performed in March 2014 and August 2015 to simulate bioclogging processes during seasonal MAR. Based on the comparison of the bioclogging process in both the winter and summer experiments, the present study confirmed that, during seasonal MAR, the recharge water temperature could significantly affect the bacterial growth and metabolism behavior, and further induce differences in the bioclogging rate and process. The sand column clogging was ~10 times faster in summer than in winter, because of a faster culturable bacteria growth rate and EPS production in the favorable summer temperature. However, the maximum concentrations of total EPS in winter were about twice those of summer, a phenomenon that is primarily caused by a ~200 μg/g sand increase of both LB-EPS and TB-EPS. Although the low temperature decreased the bacterial growth rate, the surviving bacteria could increase production of exopolymers in order to protect them from harsh environments. The TB-EPS concentration remained low (<50 μg/g sand) during the entire bioclogging process in summer and contributed little to the total EPS (<10%). For the bioclogging processes, the behaviors were similar in stages I and II in both summer and winter, and the differences mainly occurred in stage III. In the third stage, the major contributor to the reduction of the relative saturated hydraulic conductivity was bacterial cell numbers in summer, while the LB-EPS led to the decreasing Ks′ in winter. The dominant bacteria in the winter experiment were T. auensis, Acidovorax radices, Janthinobacterium sp., Rahnella aquatilis, and Aquabacterium commune, while T. auensis, Acidovorax radices, and Brevundimonas diminuta dominated in summer. For the bacterial community diversity, a ~50% increase occurred in the winter experiment. The differences in the bacterial community structure and diversity could mainly result from the seasonal inoculums. Although the different inoculums could also have an impact on the bacterial growth rate and metabolism behavior, the recharge water temperature was considered to be the dominant factor in the seasonal bioclogging process. Quantitative study on contributions of inoculum and recharge water temperature on seasonal managed aquifer recharge will be undertaken in the future research.