Three original river sediments (Orig-SD1, Orig-SD2, and Orig-CH) were collected from creeks in Pearl River Delta, China. Sediments were passed through a 100-mesh standard sieve (0.15 mm opening size) and kept in a cold room (4°C) in darkness for later use. Orig-SD1 and Orig-SD2 were identified as useable sediments because fabrics coated with them met the color demand of Gambiered Guangdong silk (one side: shiny black; the other side: brown; Pan et al., 2016), while Orig-CH was unusable. The physicochemical characteristics and performance details of these three original sediments were listed in Supplementary Table S1. In order to test the bioaugmentation performance, three modified sediments with little ferrous were considered, including high temperature (121°C) and pressure (0.1 MPa) oxidized Orig-SD1 (SD1), air-dried Orig-SD2 (SD2), and unusable Orig-CH (CH).

The silk textiles used in this study were prepared in the ChengYi factory (Foshan, China), which has been dyed with the tannin extract of Ju-liang roots to form a brown color. The prepared textiles (Orig-textile, the color characteristics seen in Table 1) were cut into 2 cm × 2 cm square pieces for sediment coating.

Enrichment cultures were prepared in 20-mL serum bottles containing 15 mL ferric citrate medium (g L^-1: ferric citrate, 3.4; NH₄Cl, 1.0; KHPO₄, 0.25; K₂HPO₄.3H₂O, 0.72; CaCl₂.2H₂O, 0.07; MgSO₄.7H₂O, 0.6; and glucose, 10) and 1 g Orig-SD1 as the inoculants which were sealed with Teflon-coated butyl rubber septa and aluminum crimp caps (Wang et al., 2008). All culture bottles were incubated at 30°C in the dark. The consortium was subcultured in a new serum bottle with 15 mL fresh culture medium once the enrichment cultures turned light green or colorless from yellow color (2–7 days, given by ferric citrate redox indicator). The subculture was consecutively repeated five times. The IRB culture was washed twice and then suspended with sterilized deionized water (OD600 = 1.0) before it was used as bioaugmentation inoculants.

The bioaugmentation experiment procedures are follows: an aliquot (3 g) of the unusable sediments (SD1, SD2, and CH) transferred into the 10-mL vial. Three milliliters of enriched consortium were added into the three sediments, uniformly mixed and cultured at 30°C in the dark for 7 days (SD1S, SD2S, and CHS). Sediments added with sterilized water were used as controls (SD1W, SD2W, and CHW). Each treatment was with three replicates. After 7 days, all sediments with and without bioaugmentation were used to coat the textiles prepared as mentioned above. After reacting for 1 h, the rest sediment on the textile was washed away and the textiles were dried in the sun.

After incubation, a thin layer of sediment was used to coat the Orig-textile for 1 h according to our previous method (Pan et al., 2016) and then the rest sediment was washed away. The color characteristics of both sides of sediment-coated textiles were evaluated with the Commission International d’Eclairage (CIE) Lab coordinates by USPRO Colorimeter (Datacolor 110TM, USA). Optical source was D65, viewing angle 10° and measure diameter 10 mm. The CIELAB color system is organized with three axes in a spherical form: L^∗, a^∗, and b^∗. L^∗ is associated with the lightness of the color and moves from top (100, white) to bottom (0, black), whereas a^∗ and b^∗ are associated with changes in redness–greenness (positive a^∗ is red and negative a^∗ is green) and in yellowness–blueness (positive b^∗ is yellow and negative b^∗ is blue). The L^∗, a^∗, and b^∗ were calculated from three repetitive measurements for every sample. And color difference (ΔE) between textile samples was used as the indicator on judging the dyeability of sediments and calculated according to the Eq. 1. From the point of technical dyeing using natural raw dyes, a somewhat wider color difference of ΔE ≤ 3 could be permitted with the use of natural resources (Martínez et al., 2001).

L_i^∗, a_i^∗, b_i^∗: color values of the i textile;

L_j^∗, a_j^∗, b_j^∗: color values of the j textile.

The sediment pH and oxidation-reduction potential (ORP) were monitored with a S20 K pH meter (Mettler Toledo, Switzerland). The organic matter content was estimated according to the previous method (Haller et al., 2011) except extended to 4 h. HCl-extractable Fe(II) and total Fe (TFe) in the sediment were extracted with 0.5 mol/L HCl for 1 h (Lovley and Phillips, 1987). After centrifuged at 8000 rpm for 5 min, the supernatant was determined using the 1,10-phenanthroline colorimetric method at 510 nm on a full wavelength scanner (Thermo Scientific, MULTISKAN GO). TFe including Fe(III) and Fe(II) in the sediment was extracted with hydroxylamine hydrochloride and determined as the Fe(II) determination (Lovley and Phillips, 1987).

DNA was extracted from 250 mg of sediment samples using the PowerSoil^TM DNA Isolation Kit (Mo Bio Laboratories, Carlsbad, CA, USA) according to the manufacturer’s instructions. The bacterial 16S rRNA genes were amplified using the PCR primers 515f/806r targeting the V4 region (Pylro et al., 2014). To distinguish the different samples, a Barcoded-tag with six nucleotide bases was randomly added to the upstream of the universal primer. The primers which were added with Barcoded-tag sequences were Barcoded-tag fusion primers. After quantification and quality control, PCR products were gradually diluted and quantified. The V4 tag PCR products were pooled with the other samples and sequenced using 300 bp paired-end model with the Illumina MiSeq platform at Chengdu Institute of Biology (Chengdu, China).

After sequencing, the final V4 tag sequences were assembled through finding the overlap between paired-end reads by the FLASH software. Chimeras were identified via UCHIME algorithm on mothur platform. Low quality fragments were filtered out using QIIME software. Sequences were clustered to operational taxonomic units (OTUs) at 97% sequence similarity by using UCLUST software (Edgar, 2010). Singletons were removed from the whole sequence data set and each sample was randomly sampled and normalized at 11,000 sequences. The numbers of original reads and final OTUs are listed in Supplementary Table S2. The dissimilarity test [non-metric multidimensional scaling (NMDS)] based on Bray–Curtis similarity distance matrices were performed by the Vegan package in R 3.1.3. The dominant OTUs in each group were depicted in a heat map conducted with R 3.1.3, and canonical correspondence analysis (CCA) was used to analyze the relationship between these OTUs and sediment properties with Mantel test. The Illumina sequence raw data reported here was submitted to the NCBI Sequence Read Archive (http://www.ncbi.nlm.nih.gov/sra) under accession number SRP083001.

After 7-day incubation, all of the three bioaugmented sediments (SD1S, SD2S, and CHS) revealed higher dyeabilities than their control groups (Table 1). The bioaugmented sediments performed shiny black color on the sediment-coated side (L^∗_SD1S = 23.11; L^∗_SD2S = 24.91; L^∗_CHS = 24.86), comparable with the standard cloth dyed with Orig-SD1 (ΔE < 3). Furthermore, the bioaugmented sediments did not penetrate to the back and the backside still kept brown (ΔE_SD1S = 2.97; ΔE_SD2S = 1.41; ΔE_CHS = 0.84), which met the demand of two tone colors for Gambiered Guangdong silk. In contrast, the control sediments (SD1W, SD2W, and CHW) still kept the original brown color (L^∗_SD1W = 29.51; L^∗_SD2W = 26.11; L^∗_CHW = 28.52), near to the color of Orig-textile (L^∗_Orig-textile = 28.17).

In order to clarify the influences of IRB inoculation on sediment dyeability, characteristics of sediments were investigated (Table 2). The pH and LOI of the bioaugmented groups showed a slight increase and decrease, respectively, compared to the control group, except for the pH of SD1S. As expected, the concentration of HCl-extractable Fe(II) in bioaugmented sediments were significantly higher than those in controls (p < 0.01), increasing by 3.5-, 12-, and 83-fold, while there was no significant change in HCl-extractable TFe (Table 2). In addition, the ORP results were well consistent with the Fe(II) results: the controls had positive ORP with oxidizing conditions while the IRB bioaugmented groups had negative ORP with reducing conditions for iron reducing.

Eleven thousand effective sequences were re-sampled from each sediment sample, resulting in 1,440–3,759 OTUs at 97% sequence identify cutoff (Supplementary Table S2) and being assigned to different taxa (Figure 1). In the enriched iron reducing consortium, Firmicutes (77.4%) was the overwhelmingly dominant phylum, followed by Proteobacteria with 11.0%. Among the Firmicutes phylum, nearly 44.3% sequences was assigned to genus Clostridium. At the OTU level, one-third sequences were assigned to Clostridium sp. (denovo 173227).

The NMDS analysis was applied to analyze sediments microbial communities with and without bioaugmentation. Although the data points of sediments with the same source were adjacent, there was a distinction between bioaugmented and non-bioaugmentation samples. Moreover, the succession of microbial communities proceeded in different directions of community succession (Figure 2).

In details, in terms of the phylum level (Figure 1), Firmicutes and Proteobacteria dominated in SD1 and SD2, while the CH community was mainly dispersed by Firmicutes, Proteobacteria, and Bacteroidetes. After bioaugmentation, the average abundance of Firmicutes increased from 55.3% (SD1W), 4.6% (SD2W), and 6.2% (CHW) to 59.7% (SD1S), 35.6% (SD2S), and 8.3% (CHS), respectively, but total percentages of Proteobacteria in all sediments decreased, even by 8.2% for SD1S. At the genus level (Supplementary Figure S1), bioaugmentation with IRB consortium led to a significant increase in the relative abundance of genera Symbiobacterium (by 24.0%), Planctomyces (by 1.1%), Geobacter (by 0.2%), Novosphingobium (by 5.8%), and Nevskia (by 1.7%) in SD1; Clostridium (by 3.0%), Bacillus (by 13.9%), Brevibacillus (by 2.8%), and Janthinobacterium (by 1.6%) in SD2; Clostridium (by 0.9%), Anaeromyxobacter (by 0.6%), and Geothrix (by 0.5%) in CH. In addition, the dominant 38 OTUs (>1% total sequences for each sample) of all the samples were analyzed (Figure 3), which were mainly composed of Firmicutes (17 OTUs) and Proteobacteria (12 OTUs). The major species (e.g., denovo 173227) in the IRB consortium showed no dominant position in bioaugmented sediments. The number of the significantly increased OTUs in bioaugmented sediments were 8 (SD1S), 5 (SD2S), and 6 (CHS), respectively, and some of these OTUs were different among the three sediments. Moreover, some sediment-specific (indigenous) IRB increased and be only found in their corresponding sediment (Supplementary Table S3).

The relative abundance of potential iron reducers were listed in Table 3. The genus Clostridium was the most abundant genus in the IRB consortium, but just occupied 1.11, 3.55, and 1.20% in SD1S, SD2S, and CHS, respectively. Pseudomonas (11.64%) was the most abundant genus of IRB in SD1S. Bacillus (14.19%) and Brevibacillus (2.84%) were the predominant iron-reducing genus in SD2S and increased by 50- and 46-fold, respectively, compared with the SD2W. Differently, the iron reducing genera for CHS were dispersed and composed of Anaeromyxobacter (1.37%) and Geothrix (1.07%) both increased by about twofold. The significantly increased OTUs in the sediments but not in the IRB consortium (Supplementary Table S3), were classified to several genera which were reported to own the ability of iron reduction such as Anaeromyxobacter (Breidenbach et al., 2015), Bacillus (Alabbas et al., 2013), Clostridium (Hassan et al., 2015), Azospira (Peng et al., 2016) Paenibacillus (Petrie et al., 2003), Desulfosporosinus (Bertel et al., 2012), and Treponema (Baek et al., 2016).

Significant correlation was observed between the composition of the bacterial community (OTU level) and physiochemical properties [pH, ORP, Fe(III), LOI] of sediments with different sources through the Mantel test (p < 0.01) and canonical correspondence analysis (CCA) (Figure 4). CCA showed the first two components (CCA1 and CCA2) together explained 76.68% of the total variation of sediment microbial community. Although some parameters [e.g., Fe(II) and ORP] had significant changes after bioaugmentation (Table 2), greater sediment-to-sediment difference than treatment-to-control in CCA profile (Figure 4) implied that the whole physiochemical properties were the substantive influence on the bacterial community constitution.