Quantification cycle (Cq) values were obtained after completion of the fluorescence quantitative PCR reaction. The relative abundance of the target gene in each sample and calibrator (blank group) was calculated using the Cq value of the internal reference gene gapA as a control and the relative expression level of the target gene = 2^-ΔΔCq (where ΔCq = Cq_{(target gene)} - Cq_{(reference gene)}, and ΔΔCq = ΔCq_(test) - ΔCq_(calibrator)) as described previously (Livak and Schmittgen, 2001). Statistical analysis was performed using SPSS software (version 10.0).

In the control group, no resistant E. coli strains were obtained (Table 4). The multidrug resistance rate was highest in the medium N and P dose group; the rate was second highest in the high-dose group. The resistance rate was lowest in the lowest- and low-dose groups. After 1 day, high multidrug resistance rates were found in the lowest, medium, and high dose groups. In the medium dose group, there were multiple multidrug-resistant E. coli strains isolated on selected days after the addition of N and P. In all treatment groups, there were high levels of multidrug-resistant strains from day 29 to the end of the experiment.

There was no biofilm formation in the control group or at day 0 in any of the treatment groups. After adding N and P nutrients, biofilm formation in the medium-dose group was the highest, followed by the high-dose and two low-dose groups (Figure 1). The biofilm that formed on the glass tube of the middle-dose group was visually apparent from the 1st day onward; meanwhile, the biofilm in the other groups was visually apparent from the 29th day onward. There was a positive correlation between the multidrug resistance rate and biofilm formation (Figure 2).

There was no mqsR gene expression observed throughout the study (samples collected on day 0 as controls). Five genes (i.e., the self-inducible gene luxS, motif flhD, fliA, motA, and fimH) were not expressed in the control group.

The gene expression level and duration trends of flhD, fliA, motA, and fimH in different treatment groups were consistent. The order of duration and expression level from highest to lowest was: the medium-dose group, the high-dose group, the low-dose group, and the lowest-dose group. The expression levels of the genes peaked after 8 days in the medium-dose group and high-dose group, and reached minima at the 15th and the 22nd day, respectively. The expression levels of the genes in the low-dose group and lowest-dose group peaked after 15 days, and reached minima at the 29th and 36th day, respectively. In the lowest-, low-, and medium-dose groups, fliA had the highest expression level, followed by motA, fimH, and flhD. The overall expression levels of the four genes were similar in the high-dose group.

The luxS expression levels were similar in all groups except the high-dose group (Figure 3) and the hierarchy of luxS expression duration and level in the other three groups was the same as for the other genes tested. We found that luxS began to be expressed in the medium- and high-dose groups after 1 and 8 days, respectively; the expression level was significantly higher than that of the other four genes. Meanwhile, luxS was expressed on the 1st day in the lowest-dose group at a higher level than all of the other genes except fliA.

Thevenon et al. (2012) investigated multiple antibiotic resistance (MAR) of the fecal indicator bacteria (FIB) E. coli and Enterococcus spp. from sediments of Lake Geneva (Switzerland) over multiple decades. Their study showed that after 1970 sediments had deposited much higher levels of FIB with MAR than those before 1970. This is due to the fact that, after 1970, eutrophication greatly increased around Lake Geneva. Thus, the trophic state of the lake and municipal wastewater treatment plants had substantial effects on FIB and on the distribution of multidrug resistance genes in the sediments. In addition, certain areas of the bay constitute a reservoir of FIB and FIB with MAR in the contaminated sediments. Therefore, the suspension of FIB and FIB with MAR can impact health risks and water quality during recreational activities.

Our previous studies also found that FIB from microcosms quickly acquired resistance when subjected to different concentrations of cornmeal, glucose, or N and P (Yu et al., 2009, 2012). Here, the concentrations of N and P were lowered; however, they were still equivalent to a medium eutrophication level (the N and P levels in the control group were equivalent to a mesotrophic level). We have found that even small additions of any type of nutrient salt to a microcosm can alter the resistance characteristics of FIB (Ali et al., 2016; Zhang et al., 2016). The resistance mechanism(s) of FIB seem to be sensitive to environmental changes, and we believe that these resistance mechanisms play a key role in bacterial self-defense systems in changing conditions, and are even more important than antibiotics themselves (Rochelle-Newall et al., 2015). Thus, it would be of great interest to develop a new system capable of detecting the no observed effect concentration (NOEC) of glucose or N and P on the formation of resistance in FIB. Furthermore, these NOECs will be critical in evaluating the threat of FIB resistance in water sources.

Real-time fluorescent quantitative PCR showed that E. coli mqsR was not expressed (Figure 3). Although studies have shown that mqsR can regulates the motility master regulon flhD and play an important role in biofilm development (González Barrios et al., 2006), our results show that the formation of E. coli biofilms was not related to mqsR expression following the addition of low levels of N and P.

Quantitative fluorescence PCR showed that expression levels of luxS, flhD, fliA, motA, and fimH, genes involved in biofilm formation, increased to a certain extent in each treatment group (Figure 3). This provides further confirmation that E. coli uses biofilm formation to adapt to changes in the external environment, in this case, in response to the addition of low levels of N and P.

Niu et al. (2013) proved that luxS influences E. coli biofilm development independently of autoinducer-2 and can assist in adaptation to varied conditions. In our study, the expression of luxS was induced rapidly in response to the addition of nutrient salts; it was detected in the medium- and high-concentration groups on day one. Pratt and Kolter (1998) used E. coli as a model organism to examine the initiation of biofilm development. Their study showed that motility is critical for normal biofilm development. Furthermore, it has been presented that type I pili (harboring the fimH gene) is required for initial surface adhesion (Pratt and Kolter, 1998). Niu et al. (2013) also showed that fimbriae may promote surface colonization and other aspects of initial biofilm development by E. coli.

Based on the role of pili and kinetic (or motility) genes in biofilm formation, as well as the results of this study, we infer that the lack of flhD, fliA, motA, and fimH expression suggests that the mature biofilm has begun to form. After completion of the adhesion and microcolony formation stages, in which the flagella and pili play an important role, the bacteria begin to gradually enter the mature biofilm stage (Berne et al., 2015). The shorter the duration of gene expression and maintenance of these four genes, the earlier E. coli enter the mature biofilm forming stage.

We showed that the expression of the four genes was maintained for the shortest time in the medium-dose group, followed by the high-dose group, the lowest-dose group, and the low-dose group. Thus, the medium-dose group was the earliest to form a mature biofilm; this agrees with previous analyses of biofilm formation abilities (Abberton et al., 2016). In addition, judging from the overall gene expression levels, gene expression is positively correlated over time, indicating that the medium-dose group was the best-adapted to the external environment. It formed a mature and stable biofilm in the shortest time and at the lowest gene expression levels, which ensured the survival of the bacteria in a changing environment.

The expression levels and trends of the three motor genes and pilus were consistent (Figure 3). The expression level of fliA was the highest, followed by motA, fimH, and flhD. In E. coli, flg, flh, and fli are the flagellum biosynthetic genes, and mot are motor genes (Chen and Helmann, 1992). The fimH adhesin is the tip protein of E. coli type I fimbriae (Lee et al., 2005). At the top of the hierarchy is the flhD master operon, which is consists of the flhD and flhC genes: expression of this regulon is essential for the expression of all of the remaining genes (Supplementary Figure S3). Class II operons consist of operons under direct control of the master operon. The genes contained in these operons encode a flagellum-specific sigma factor (σ²⁸ or FliA) that is required for the transcription of class III operons, many structural components assembled in the early and middle stages of flagella synthesis, and some proteins of unknown function. Class IIIa regulons are under the dual control of fliA and flhD/flhC; meanwhile, class IIIb operons (such as motA) are under the control of fliA alone (Liu and Matsumura, 1994). Claret et al. (2007) showed that sigma factor FliA, transcriptional activator FlhD₂C₂, and flagella regulators are involved in the coordination of flagella and type I pilus synthesis. FliA is also a key regulatory component linking synthesis of the flagella and type I pili; its effect on type I pili is mediated via a c-di-GMP-dependent signaling pathway. Additionally, fliA, the key gene connecting flagella and pili, had the highest overall expression in our experiments, followed by motA, the motor gene (Belas, 2014). This is important since motility plays a critical role in surface contact and biofilm expansion during the initial stages of biofilm formation.