Bioresources inner-recycling between bioflocculation of Microcystis aeruginosa and its reutilization as a substrate for bioflocculant production

Bioflocculation, being environmental-friendly and highly efficient, is considered to be a promising method to harvest microalgae. However, one limitation of this technology is high expense on substrates for bioflocculant bacteria cultivation. In this regard, we developed an innovative method for the inner-recycling of biomass that could harvest the typical microalgae, Microcystis aeruginosa, using a bioflocculant produced by Citrobacter sp. AzoR-1. In turn, the flocculated algal biomass could be reutilized as a substrate for Citrobacter sp. AzoR-1 cultivation and bioflocculant production. The experimental results showed that 3.4 ± 0.1 g of bioflocculant (hereafter called MBF-12) was produced by 10 g/L of wet biomass of M. aeruginosa (high-pressure steam sterilized) with an additional 10 g/L of glucose as an extra carbon source. The efficiency of MBF-12 for M. aeruginosa harvesting could reach ~95% under the optimized condition. Further analysis showed that MBF-12, dominated by ~270 kDa biopolymers, contributed the bioflocculation mechanisms of interparticle bridging and biosorption process. Bioflocculant synthesis by Citrobacter sp. AzoR-1 using microalga as a substrate, including the polyketide sugar unit, lipopolysaccharide, peptidoglycan and terpenoid backbone pathways. Our research provides the first evidence that harvested algae can be reutilized as a substrate to grow a bioflocculant using Citrobacter sp. AzoR-1.

In recent years, microalgal blooms have drawn substantial attention, particularly because of the threat they pose to human health and the environment 1 . Microcystis aeruginosa (M. aeruginosa) is a ubiquitous toxin-producing cyanobacteria present in the aquatic environment 2 . M. aeruginosa can reduce dissolved oxygen levels, cause discoloration of receiving water (red tide) and produce odors and toxins that pose hazards to human health and aquatic ecosystems 3 . Indeed, from another perspective, microalgae are promising and new biomass resources for lots of high-value applications, i.e. triacylglycerol, bioalcohols (e.g., ethanol and butanol), polyunsaturated fatty acids (e.g., eicosapentaenoic acid, and docosahexaenoic acid), and pigments (e.g., lutein and chlorophyll) 4 . Combining and considering above aspects, one of the promising ways to achieve microalgal biomass would be collecting the microalgal cells that bloomed and suspended in the aquatic environment 5 . To address this, flocculation technologies have to be applied for either waterbody remediation or microalgae biomass recovery 6 .
However, the high economic cost for separating microalgae biomass from water is still a bottleneck. Traditional bioflocculants, such as ferric salts or aluminum salts, have been studied to flocculate the microalgal cells 6,7 . Although the removal efficiency for traditional flocculants was quite high in many studies, these bioflocculants have certain disadvantages for long-term use. Residual aluminum in treated water, either in the supernatant or in sludge, is difficult to remove and sometimes exceeds the upper limit of water standards, causing a threat to human health 8 . In this context, a great deal of research has been devoted to the use of natural materials as flocculants, such as clays 9 , chitosan 10 , and cationic starch 11,12 , which can be biodegraded, and thus are safer for humans and the ecosystem. Especially, flocculants produced by microalgae, bacteria or fungus are more attractive for microalgae recovery, due to its safety, biodegradability and non-secondary pollution 13,14 . However, the high cost of bioflocculants takes the cost prohibitive issue, in which the high cost on the substrates (e.g., carbon sources and nitrogen sources) that are used for the cultivation of bioflocculants producing microorganism is the major challenge. Previous studies used crop stalks and kitchen wastes to produce bioflocculants, but the bioflocculant yield was unsatisfactory 15,16 , and the production conditions were critical 17,18 . Straw contains a large amount of cellulose, which can be used as a carbon source. However, during bioflocculant production, cellulose-degrading bacteria need to be added, which increases the competition between the microbial populations and causes an antagonistic effect 16 .
The microalgae, M. aeruginosa, we intend to flocculate are potentially a promising substrate for bioflocculant bacteria cultivation. The carbon and nitrogen levels in the M. aeruginosa biomass reached 22 to 28% and 4 to 5%, respectively 19 , indicating that the biomass can provide a potential nutrition source for bioflocculant production. Thus, we hypothesized a clue for the inner-recycling of biomass that could harvest the typical microalgae, M. aeruginosa, using a bioflocculant produced by bacterium (e.g., Citrobacter sp. AzoR-1 in this work). In turn, the algal biomass was utilized as a substrate for Citrobacter sp. AzoR-1 cultivation and bioflocculant production.
To confirm above hypothesis, we cultivated a bioflocculant-producing bacterium, Citrobacter sp. AzoR-1, which produces a bioflocculant with a high harvest efficiency for a typical microalga, M. aeruginosa. Next, we examined the flocculated M. aeruginosa for bioflocculant production and determined the optimal bioflocculant conditions and the mechanisms of flocculation using the bioflocculant. The metabolic and gene expression profiles of the microorganisms during the microalgal consumption were identified through transcription analyses, to reveal the pathways underlying the bioflocculant production. What is novel would be the bioresources inner-recycling between bioflocculation of Microcystis aeruginosa and its reutilization as a substrate for bioflocculant production. This work provided a promising and new clue for efficiently and economically treating microalgae blooms.

Results
Harvesting of M. aeruginosa by the produced bioflocculant. RSM provided response surfaces and contour plots to study the interactions between the operational parameters and removal efficiency. All of the selected optimum solutions retained the desired simulations of the removal efficiency (see Table S4). Predicted data achieved from the response surface methodology (RSM) under the optimal conditions were compared with the experimental proving results to validate the model. The experimental verification values observed for the removal efficiency ranged from 91.68% to 97.21% (see Table S4). Deviations between the experimental and the predicted values are all within 2%, indicating that the model fitted the experimental data well (see Table S4). Furthermore, to intestate the liner variable's impact on the flocculation efficiency with other variables fixed in median, the experimental verification values were also fixed the prediction curve as shown in Figure S1. The optimal M. aeruginosa flocculation conditions by MBF-12 were from 10 to 30 °C in temperature, 12.7 mg/L in dosage, 1.2 hours in settling time and pH < 8 (see Figure S1).
The highest removal efficiency of M. aeruginosa was 95.37%, and the lowest was 2.53%. Two linear terms, the pH and the MBF-12 dose, had significant effects on the flocculation efficiency (Fig. 1). The optimum MBF-12 dosage was 12.70 mg/L for 10 9 algal cells per liter. The removal efficiency began to slightly decrease at higher bioflocculant doses because the formation of negatively charged flocculants led to charge protection 19 . Furthermore, for a fixed MBF-12 dose of 10 mg/L, the removal efficiency decreased from 87.09% to 7.21% when the pH was increased from 8 to 13. Such significant effect of pH are related to the functional groups changing on bioflocculant biopolymers under alkaline conditions 20-24 , which will be discussed later. Flocculated M. aeruginosa was reused as the main substrate for bioflocculant production in Culture A-F to select an optimal media. Then, Culture-C was selected for all the subsequent bioflocculant (MBF-12) production protocols, as it achieved both a high removal efficiency and low glucose consumption (see Fig. 2a). As shown in Fig. 2(a), Culture-B and Culture-C achieved the highest bioflocculant production efficiencies, which reached 3.85 and 2.9 g L −1 (dry weight per Liter culture), respectively. In contrast, the production efficiencies was the lowest and no more than 1 g L −1 for both of Culture-A and Culture-F. Bioflocculant productivity in PT-1 and Culture-C at different growth stage were also studied, as is shown in Figure S3, the highest productivity started at 4th day, which suggest that, in both PT-1 and Culture-C, the synthesis of bioflocculant is mainly in the late stage of microbial growth. The maximum yeild of bioflocculant was 4.3 g L −1 in PT −1 and 3.8 g L −1 (dry weight per Liter culture) in Culture-C, respectively.
The biomass production (presented as the dry cell weight) and removal efficiencies obtained from Culture-C and PT-1 cultivation were compared. PT-1 was used as a background that was cultivated without using algae. Interestingly, even though the cell proliferation rate in Culture-C was 50% lower than that in PT-1, the removal efficiency of the bioflocculant produced in Culture-C was 15% higher than that of PT-1 (see Fig. 2b), at MBF-12 dosage of 12 mg/L; bioflocculation time of 0.1 hour and settling time of 1.2 hours. The two cultures apparently produced bioflocculants with different characteristics. M. aeruginosa possesses a high protein content and a low sugar content (18% w/w) 25 , which signifies that the extra glucose included in Culture-C was the main reason for the more efficient bioflocculant production in Culture-C.
MBF-12 is a long-chain biopolymer which molecular weight was 270 kDa (see Fig. 3c), so it is long enough to perform bridging between M. aeruginosa cells. Notably, MBF-12 was unfolded in solution and with a size larger than 10 μ m according to its SEM image (see Figure S2), which should contribute to a more effective flocculation by bridging the M. aeruginosa cells.
The Fourier-transform infrared (FTIR) spectrum of purified MBF-12, which was generated from Culture-C, exhibited a broad hydroxyl stretching peak at 3423 cm −1 and an amine band at 1630-1550 cm −1 , as shown in Fig. 3d. The adsorption peak at 2930 cm −1 indicated the C-H stretching vibration. Strong absorption peaks present in the range of 1100-1200 cm −1 are typical peaks for proteoglycans 25 . Bioflocculant metabolic pathways. Metabolism-related genes comprised a significant portion of the annotated non-redundant sequences, as found in other microbial transcriptomes [26][27][28] . More than 50% of the 4,961 non-redundant sequences were predicted to be associated with metabolism, and many of the metabolic genes were predicted to be associated with polysaccharide biosynthesis, reflecting the bioflocculant products of the strain.  Table S5). Because the genes for the final conversion of saturated bioflocculants to hydrocarbon end products remain unidentified, the transcriptome data included putative genes associated with unsaturated hydrocarbons. Such genes are related to bioflocculant biosynthesis, including the polyketide sugar unit, lipopolysaccharide, peptidoglycan, and terpenoid backbone (Table S5 and Fig. 4). Genes related to these bioflocculant biosynthesis pathways were classified into different functional categories using the Kyoto Encyclopedia of Genes and Genomes (KEGG) Automatic Annotation Server (KAAS) program, as shown in Fig. 4 and Table S5. The KEGG pathway database is copyrighted by Kanehisa laboratories [29][30][31] .

Discussion
To further understand the destabilization mechanisms of MBF-12, we need to discuss the charge of the functional groups on the bioflocculant as a function of pH. Only organic carboxyl groups and amine groups were especially discussed here, as other groups (e.g., phosphate groups) have much lower concentration on the bioflocculant. The pKa of organic carboxyl groups are 4.1 29 and the pKa of organic amine groups are 8.8 30 ; consequently, we need to examine the charge on the bioflocculant for pH less than pH 4.1, 4.1 to 8.8 and pH greater than pH 8.8.
The carboxyl groups would be fully protonated at a pH < 3.1, and we would expect to see the zeta potential to be positive depending on the surface concentration of amine groups. We do not see a positive zeta potential, but we do see that the zeta potential increased when pH decreasing from 4 to 1 (see Fig. 5). Perhaps there are uncharged oxygen containing functional groups that cause the Zeta potential to be always negative. Nevertheless, the M. aeruginosa cells bioflocculation efficiency increased in pH range from 4 to 1, which was consistent with the charge changing trends of the bioflocculant (see Fig. 5).
As the pH increases from 4.1 to 8.8, we do see the zeta potential becomes more negative as the carboxyl group ionize. Consequently, the flocculation mechanism of MBF-12 would be bridging for pH in this range. The amine groups would be protonated for pH values < 8.8, and we would expect to see that the flocculation efficiency to decrease with increasing pH from 4.1 to 8.8. However, charge neutralization would not occur because the charge on the algal cells is negative and the charge on the bioflocculant is also negative. Consequently, bridging would be the predominate flocculation mechanism for pH values from 4.1 to 8.8, and the effectiveness of flocculation was observed to decrease as expected as pH increased in this range because of repulsion between the bioflocculant biopolymers and the algae cells.
As the pH increases above 8.8, we see the zeta potential of MBF-12 increased obviously from ~− 16 mV to ~− 2 mV. A possible explanation is that at high pH values hydrolysis occurs which removes the carboxyl groups from peptidoglycans and lipopolysaccharides under alkaline conditions. And, amine groups, which may lose a proton then resulted bioflocculant become more negatively charged. Furthermore, M. aeruginosa cells became more negatively charged as the pH increases, i.e., reached ~− 40 mV when pH > 8.8 (See Fig. 5). Consequently, the flocculation efficiency decreased with increasing pH due to electrostatic repulsion between the bioflocculant and cells (see Fig. 2). For pH values greater than 8, bridging would still be the predominant flocculation mechanism but the flocculation effectiveness diminished, which was observed, because negative charge on the algal cells increased.
Besides interparticle bridging, biodsorption appears to be another main bioflocculation mechanisms of MBF-12. The basic components of microalgal cell wall are cellulose and glycoproteins 31 , which could strongly be adsorbed by the main components of MBF-12, lipopolysaccharide, peptidoglycan and Isopentenyl-PP (according to the transcriptome analysis). According to Filho 32 , low surface-charge nonionic polymers and the high-molecular-weight should be in the range from 10 5 to 10 7 g/mol for effective bridging. The molecular weight of MBF-12, 270 kDa, was just in this range. Furthermore, the Donnan potential on two sides of algal cell walls could adsorb inorganic metal ions for cell metabolisms [33][34][35][36] , which may also facilitate adsorption and interparticle bridging to occur between flocculant MBF-12 and M. aeruginosa cells.
Overall, MBF-12 was the biopolymer that with MW as much as 270 kDa. It should be originated from the extracellular substances of Citrobacter sp. AzoR-1, according to our extraction method as given in Supporting Information (SI-2). MBF-12 is comprised of high-molecular-weight polymers, with both of positive groups and negative groups, and they can flocculate M. aeruginosa cells mainly through interparticle bridging process and biosorption as shown in Figure S2 33 .
The vigorous bioflocculant synthesis in Citrobacter sp. AzoR-1, especially the prevalence of mature lipopolysaccharide synthase transcripts which has never been reported before in Citrobacteria transcriptomes, was considered to be resulted by feeding the bacteria with the microalgal biomass substrate. The transcriptome analysis of the microalgal substrate culture revealed that the genes related to the MBF-12 synthesis of Citrobacter sp. AzoR-1 were positively regulated. As expected, the Citrobacter sp. AzoR-1 transcriptome contained many expressed sequence tags (ESTs) associated with the polysaccharide biosynthesis system, which was consistent with previous research results 4,5,37 . Interestingly, the Citrobacter sp. AzoR-1 transcriptome data embraced many genes associated with terpenoid backbone metabolism.
In this work, the efficiency of MBF-12 for M. aeruginosa harvesting could reach ~95% under the optimized condition. Even though the commercial inorganic flocculants, such as Al 2 (SO 4 ) 3 , FeCl 3 and Polyaluminum chloride (PAC), showed similar harvesting efficiencies for M. aeruginosa (see Figure S4), however, they may cause micralgal cell rupture and drain, and thus increase algal toxins in the aquatic environment 38,39 . These commercial flocculants are usually artificial synthesized, in contrast, MBF-12 was produced by reutilizing the harvested microalgal biomass via a bioflocculant-producing bacteria cultivation. This implied MBF-12 was potentially more economical. Furthermore, MBF-12 was proved to be ion independent during flocculation (as confirmed in Fig. 1), whereas the previous reported bioflocculants were normally cation-dependent 40,41 . As a result, the metal cation could be saved when using MBF-12 for microalgae flocculation. Guo et al. 42 and Alam et al. 43 also reported the high efficiency of extracellular biopolymers on microalgae harvesting (90% in flocculation efficiency), but they did not reutilize the harvested biomass in their work 44,45 .
Bioresources inner-recycling, between bioflocculation of Microcystis aeruginosa and its reutilization as a substrate for bioflocculant production, was achieved in this work. By this means, the bioflocculation of Microcystis aeruginosa reached as much as 95%, with the bioflocculant produced from Citrobacter sp. AzoR-1 by the harvested Microcystis aeruginosa biomass.

Methods
The bioflocculant-producing strain was isolated from an activated sludge of a secondary wastewater treatment plant in Changchun, China. The 16 S rDNA sequence of the strain was 99.9% similar to Citrobacter sp. AzoR-1, as identified by Sangon Biotechnology Co., Ltd (Shanghai, China). The sequencing result is given in Supplementary Information SI-1.
When using microalgal biomass as substrates for bioflocculant production, we optimized the medium components by supplementing extra carbon and nitrogen to 10 g/L M. aeruginosa biomass (see details in Table 1). The biomass concentration refers to the wet weight concentration after centrifugation and before high-pressure steam sterilization. After sterilization and inoculation of the media, the strains were cultured in a rotary shaker at 30 °C and 150 rpm for 72 h.
The methods used to prepare the bioflocculant are given in Supplementary Information SI-2. The microbial flocculant of Citrobacter sp. AzoR-1 was termed MBF-12.
M. aeruginosa was cultured for 10 days in BG11 medium to a concentration of ~10 9 cells per liter for the flocculation study. M. aeruginosa cells were washed twice with distilled water prior to the flocculation experiments to remove the influence of ions from the various culture media. The initial M. aeruginosa concentration was determined from the optical density at 680 nm (OD 680 ) using a UV-vis spectrophotometer (Model-T6, PGENERAL Ltd., Beijing, China).
Beakers containing the mixture of M. aeruginosa solution and MBF-12 were shaken in an orbital shaker (Model-HZQ-X100, HDL APPARATUS Ltd., Hangzhou, China). After the addition of the bioflocculant, the biomass was stirred at high G values of ~350 s −1 for rapid mixing for 1 minute, and then low G value of ~19 s −1 for 5 minutes to promote the mixing bioflocculation of M. aeruginosa cells. Finally, the mixture was settled from 0 to 1.71 hours. Culture broth without bioflocculant was utilized as a control.
The removal efficiency of M. aeruginosa was determined using the following calculation, removal efficiency (%) = (A − B)/A × 100, where A is the OD 680 value of supernatant of the sample and B is the OD 680 value of supernatant of the control. The optical density of the supernatant was determined at the half height of the clarified culture.
A central composite design (CCD) and response surface methodology (RSM) were applied to optimize the MBF-12 flocculation conditions in this study (see details in Supplementary Informtion SI-3). Optimization of the operating conditions was conducted using the quadratic models of the experimental design. The removal efficiency of M. aeruginosa cells was determined by testing five optimum solutions provided by the models. The solutions were labeled from 1-5 and were selected and verified under selected optimum conditions (see details in Table S4).
The Zeta potential of MBF-12 was determined using a Malvern nano ZS90 analyzer (Malvern Instruments Ltd, Worcestershire, UK). A Fourier-transform infrared spectrometer (Nicolet 6700, Thermo Fisher Scientific,  Waltham, USA) was used for MBF-12 detection after vacuum drying pretreatment. Fluorescence excitation emission matrix (EEM) analyses were conducted with a fluorescence spectrophotometer (Hitachi F-7000, Hitachi Inc., Tokyo, Japan) at a scan rate of 240 nm/min and an excitation/emission slit bandwidth of 2.5 nm. The scanning field was set at excitation from 220 to 400 nm and emission spectra from 300 to 550 nm. Total RNA was used for the complementary DNA (cDNA) library construction and was extracted using the RNeasy MinElute Cleanup Kit (Qiagen Inc., Hilden, Germany) according to the manufacturer's instructions 46 . The Citrobacter sp. AzoR-1 was cultured axenically in Culture-C; cells were harvested five days after inoculation, and the samples were immediately frozen in liquid nitrogen. Then, the samples were ready for transcription analyses to reveal the sequences underlying the bioflocculant production (see detailed procedures in Supplementary  Information SI-4).