Two cruises were conducted by the vessel with a gross tonnage of ∼100 tons from 15 to 18 August 2016 and from 9 to 12 January 2017, respectively. Water samples were collected near the surface (0.5 m depth) and bottom (0.5 m above bottom) layers in the PRE, using 5.0-L Niskin bottle (Figure 1). In situ environmental factors (salinity, pH, turbidity, and temperature) were measured using a YSI 6600 sonde (YSI Inc., United States). Chl a concentrations were determined after extraction in 90% acetone overnight (Parsons et al., 1984).

CSP_color concentration was the concentration of CSP determined spectrophotometrically after staining with a solution of 0.04% Coomassie Brilliant Blue (CBB, G-250) following the method of Cisternas-Novoa et al. (2014). Briefly, 10–50 mL water samples from the PRE were filtered onto 25 mm polycarbonate filters (0.4-μm) using low and constant vacuum (<150 mbar). The filters then were stained with 0.5 mL of CBB for 30 s. Then the filters were rinsed with Milli-Q water until the excess dye was removed. All filters were prepared in duplicate. Stained filters were transferred into 4 mL of extraction buffer (3% sodium dodecyl sulfate in 50% isopropyl alcohol) in 15 mL polypropylene tubes. Tubes were then sonicated at 37°C for 2 h. Absorbance of the solution was recorded at 615 nm using spectrophotometry. Milli-Q water was used for blanks in the same way. Bovine serum albumin (BSA) was used as standard (Cisternas-Novoa et al., 2014). As the measurement of CSP was calibrated using BSA, it could be effective way to convert CSP concentrations to nitrogen concentrations. The conversion assumes that CSP from the PRE bonds with Coomassie Blue in the same manner with BSA. Concentration of protein-N in CSP was calculated using the % N of the BSA (16%) supplied by the manufacturer (Sigma Chemical Co.) (Dortch et al., 1984), and the unit of CSP-N is μmol BSA-N-eq. L^–1.

CSP abundances and equivalent spherical diameter (ESD) were measured by microscopy (Engel, 2009). Two to thirty mL were filtered (<150 mbar) onto polycarbonate membrane filters (0.4 μm) with Milli-Q water as blanks. The filters were stained with 0.5 mL CBB working solution and rinsed with Milli-Q water to remove excessive CBB. Then, filters were placed onto Cytoclear slides and stored at –20°C until analysis. Images were taken at a magnification of × 200 with a light microscope (Olympus BX53, Japan), and CSP abundance and ESD were determined by using Image J software. The abundance of CSP adsorbed with silt and debris materials was determined using particle recognition on filters and image processing similar to the method used by Cisternas-Novoa et al. (2019). The blue channel pictures were used to quantify silt or detritus attached CSP after RGB split. In this way, silt, sand, and debris particles were readily distinguished from CSP stained blue.

A subsample of water (200–500 ml) was filtered onto Whatman GF/F filters in duplicate (<150 mbar) for PN and PHAA. For PN, carbonates were removed by exposing filters to fuming HCl for 12 h, and then filters were dried at 60°C overnight. Filters were put into tin cups and PN was determined using elemental analyzer (Elementar, vario EL cube) calibrated with an acetanilide standard according to Sharp (1974). For PHAA samples at S1, S4, S6, and S8, analysis was performed with some modifications following Fitznar et al. (1999). Duplicate GF/F filters were hydrolyzed with HCl (16%, 20 mL) and 100 μL of 1 mmol L^–1 ascorbic acid inside sealed glass ampoules at 110°C for 24 h. Thirteen derivative amino acids (AAs) serine (Ser), arginine (Arg), glutamic acid (GlX), aspartic acid (AsX), glycine (Gly), threonine (Thr), tyrosine (Tyr), valine (Val), alanine (Ala), phenylalanine (Phe), isoleucine (Ile), leucine (Leu), gamma-amino-butyric acid (GABA) were separated on a 1260HPLC (Agilent) with a Phenomenex™ Hyperclone column (BDS C18,150 × 4 mm) with guard column (4 × 2 mm) after derivatization with o-phthalaldehyde and mercaptoethanol. The internal standard was α-amino butyric acid (Aba). The relative standard deviation was <5%.

CSP spatial differences were measured by Paired-sample t-test. Difference of CSP between each station was determined by Tukey HSD test. Pearson correlation was used to analyze the correlations between CSP properties and environmental factors. Statistical analysis was performed using Origin 8.0 (OriginLab Corporation, United States) software.

The freshwater discharge from the Pearl River delta was about 552 × 10⁹ m³ in August 2016 and 89 × 10⁹ m³ in January 2017, respectively (Data from the Zhujiang Sediment Bulletin, 2016 and 2017)¹. Vertical profiles of salinity showed distinct seasonal pattern driven by river runoff and monsoon strength. Water was stratified in the mid- and lower- estuary during the summer cruise of 2016 (Figure 2A). During the winter cruise, however, the estuary was well mixed based on salinity profiles, due to the river runoff reduction and the increased northeast winds (Figure 2B). The environmental factors are given in Table 1. Low dissolved oxygen concentration occurred in the upper section during summer and winter. Dissolved oxygen concentration below 2 mgL^–1 at S1 in bottom during summer indicating hypoxia at bottom. S5 and S6 are in the edge of west channel of the estuary. High turbidity at S5 and S6 at bottom was closely related to the intrusion of salt water and resuspension of sediment.

Figure 3 shows the variation of Chl a and CSP_color concentrations during two cruises. In the summer cruise, CSP_color concentrations ranged between 73.7 and 685.3 μg BSA eq L^–1, with an average of 325 ± 227 μg BSA eq L^–1 (Figure 3A). Chl a concentration in the surface was reduced from 17.3 μg L^–1 at S2 in zone 1 to the lowest value 3.6 μg L^–1 at S5 in zone 2. High Chl a concentration was also found at more oceanic station (S8) with 15.6 μg L^–1 in the surface water. Similar to the distribution of Chl a, concentrations of CSP in surface were significantly higher in zone 1 (river section) and zone 3 than those in zone 2 (t-test; p < 0.05). CSP_color concentration in the bottom was significantly higher than that in the surface (Figure 3B) (t-test; p < 0.05).

During the January 2017 cruise, CSP_color concentrations ranged from 106.4 to 396.4 μg BSA eq L^–1, with an average of 246 ± 80 μg BSA eq L^–1 (Figures 3C,D). High concentrations in surface were found at the lower estuary (Figure 3C). CSP_color concentrations in the surface and the bottom was not significant different during winter (t-test; p < 0.05). The spatial variation of Chl a concentration during the winter cruise showed a similar pattern as in summer except for the sample at station S8, which showed a decrease from S7.

Although CSP showed a similar spatial distribution as Chl a, no overall significant correlation was observed between the CSP_color and Chl a concentration (p > 0.05, n = 32). CSP_{(ESD ≥ = 2 μm)} abundance in the surface was ranging from 6.4 × 10⁶ L^–1 to 9.2 × 10⁷ L^–1 and from 1.3 × 10⁷ L^–1 to 6.5 × 10⁷ L^–1 during summer and winter, respectively (Figure 4). The spatial distribution of CSP abundance was similar to the CSP_color concentration determined by the spectrometer.

Figure 5 shows pictures of CSP taken by microscopy. About 20% of the CSP adsorbed with silt and debris materials to form large-size flocs in the surface samples of the river reach (Figures 5A,B), while these large biomineral floc enriched in protein rarely appeared in the mixing zone and lower estuary. Moreover, in the river reach, aggregates of cyanobacteria, such as Spirulina spp., were stained by CBB (Figure 5A). In contrast, CBB-stained diatoms or diatom spines did not occur in zone 1. In the mixing zone (zone 2), the smaller-sized CSP <10 μm were most abundant (Figure 5D). At more saline stations, many diatom aggregates stained by CBB were observed, but free and non-attached CSP were less abundant than in the upper estuary (Figures 5E,F).

Figure 6 shows the distributions of PN and PHAA in the PRE. PN concentration varied between 3.1 and 23.7 μmol L^–1 in August 2016 and between 4.1 and 21.0 μmol L^–1 in January 2017, respectively (Figures 6A,B). High values of PN were found within the estuary relative to the water entering from land or seaward ends except for S7. The average percentage of CSP-N in the PN pool was 34.98% in summer and 30.8% in winter, respectively (Table 2). In the lower estuary, high percentages of CSP-N in PN were found from the samples in winter and from S7 samples during the August cruise, respectively (Table 2).

Particulate hydrolysable amino acid concentration was highest at S1 and showed lower concentrations in the middle of PRE (Figures 6C,D). CSP-N concentration was positively correlated with PN (p < 0.001, n = 16, r² = 0.44) and PHAA (p < 0.001, n = 16, r² = 0.56) (Figures 6E,F). The N content in CSP was comparable with the N content of PHAA. The dominant AAs in descending order were as follows: Gly, AsX, Ala, GlX, Leu, Thr, Ser, and Leu (Figure 7). Despite similar molar percentages of most AAs at different stations, some AAs showed slight differences along the estuary and between surface and bottom samples. AsX exhibited higher mole% in the mixing zone than those in in zone 1 and zone 3. Ala was high at S8 in surface (13.9%) where high Chl a concentration was observed in summer cruise. Higher distribution of AsX, Gly, and Ser appeared in the bottom than in the surface.

In this study, CSP abundance (ESD ≥ 2μm) in the PRE was comparable to those found in marine water ranging from 10⁶ to 10⁸ L^–1 (Long and Azam, 1996; Engel and Galgani, 2016; Busch et al., 2017; Engel et al., 2020; Table 3). Due to higher phytoplankton biomass and terrestrial input in the PRE, in particular in the upper estuary, CSP_color concentrations were much higher in the PRE than those in the oligotrophic seawater, i.e., a North Atlantic Bermuda Rise station (CSP_color concentration 0.532–18.20 μg BSA eq. L^–1) (Cisternas-Novoa et al., 2015). Compared to the Mississippi River plume (Ziervogel et al., 2016) and the Danube River (Luef et al., 2007; Table 3), CSP abundance in the PRE was orders of magnitude higher. In the study of Ziervogel et al. (2016), the stations were outside of the Mississippi River estuary (salinity ≥ 30) where Chl a concentration was low (average value of 1.4 μg L^–1) in the surface water and the removal of CSP by the sinking and dilution was significant in the Mississippi River plume (Ziervogel et al., 2016). In contrast, most of the stations in this study were in the inner of estuary or in the river sections, which had relatively higher concentration of freshly produced CSP due to high Chl a content in the PRE (average value of 7.1 μg L^–1), and received large amounts of terrestrial inputs within the inner parts, consequently yielding CSP accumulation in the PRE. Therefore, the large variability of CSP abundance among different rivers or estuaries could be attributed to the site-specific differences in environment and biological process.

Compared to TEP, the sources and fate of CSP in most natural ecosystem remained largely unknown (Long and Azam, 1996; Engel et al., 2015; Thornton, 2018). In general, CSP in the marine system are thought to originate from autolysis and viral or bacterial lysis, extracellular excretions, grazing and microbial degradation of detritus (Mannino and Harvey, 2000; Cisternas-Novoa et al., 2015). However, estuaries are highly dynamic environments with particles originating from both terrestrial and marine inputs; data on CSP characteristics are limited for estuaries so far (Ziervogel et al., 2016).

Pictures taken by microscopy indicated that the primary origins of CSP in the PRE is autochthonous primary production. In the freshwater of the upper estuary, much of living phytoplankton cells, such as cyanobacteria and chlorophyta were stained by CBB. In the more marine stations of the estuary, chain-forming diatoms were stained by CBB, such as Skeletonema costatum, Chaetoceros spp., and Pseudo-nitzschia pungens. Although the alga themselves should not be considered as CSP, it is generally observed that higher autotrophic biomass leads to more gel particles. However, the correlation coefficient between CSP and Chl a is too low to show a good relationship. Therefore, other processes could influence the concentrations of CSP in the PRE (see section “Discussion” below). Similarly, the slight decoupling of CSP from Chl a distribution also occurred in the Mississippi River Delta (Ziervogel et al., 2016). Concentrations of CSP decreased from S7 to S8 during summer, although phytoplankton biomass was high at S8, which could be related to the sampling period coinciding with the stage of bloom building up and/or to the difference of phytoplankton composition (Granum et al., 2002; Prieto et al., 2002; Lourenco et al., 2004; Grossart et al., 2006; Engel et al., 2015).

The discharge in August 2016 was 6 times that in January 2017 in the PRE. Therefore, higher freshwater runoff and terrestrial input could supply large amounts of CSP into the PRE in the wet season. At S1, CSP_color concentration in the wet season was more than twice that measured in the dry season, although the difference of Chl a concentration was not significant between the two cruises. It can be speculated that one source of CSP in the PRE is the input of terrestrial material. About 20% total abundance of CSP from the surface sample in the upper of the PRE strongly adheres to debris and soil silt. Although the number of CSP that adhered to silt was small numbers, they made a greater contribution (50.2%) to the total CSP in terms of area in the upper of estuary due to their large particle size. The abundance of these bio-mineral floc decreased with distance from the upper to the lower reach of the PRE, except for the bottom water samples at S5 and S6 where the turbidity was higher. In addition, their abundance at the surface (Figures 5A,B) was higher than that in the bottom waters (Figure 5C). It is therefore unlikely that such large CSP in the river reach originated from sediment resuspension. It has been suggested that the majority of OM introduced by rivers into estuaries is tightly associated with suspended minerals (Hedges and Keil, 1999). This kind of CSP that associated with silt may be bound to, or a constituent of humic substances or be Glomalin-related soil protein (GRSP) (Koide and Peoples, 2013; Wang et al., 2018). GRSP is produced by arbuscular mycorrhizal fungi and can be leached from soil and flushed into river and coastal sediment (Wang et al., 2018), and a recalcitrant pool of glomalin generally strongly adheres to soil particles (Koide and Peoples, 2013). Therefore, GRSP is likely to be one terrestrial source of CSP in the Pearl River.

Higher concentration in bottom than in surface waters in the middle and lower estuary [turbidity maximum (TM) zone] reflected that one source of CSP might be derived from resuspension of the bottom sediments. Wai et al. (2004) has found that there was one TM at S6 in the west channel. The contributions of the resuspension, Eulerian advection, Stokes drift and gravitational circulation to the total sediment transport at S6 were of 52.1, 16.2, 14.0, and 16.6%, respectively (Wai et al., 2004). Therefore, the CSP maximum at station S6 in bottom water may be mainly attributed to the resuspension process. This inference was supported by the observation that a large amount of TEP could be released from sediment during sediment resuspension in the turbid PRE (Sun et al., 2012), although TEP and CSP are not the same population of particles in the PRE (comparison between CSP and TEP in the PRE is available in the Online Supplement).

The spatial distribution of CSP showed higher concentration in the upper reach and in the lower part of the estuary compared to the mixing zone. CSP_color concentration dropped from the upstream to the Humen downstream by 80% in summer and 55.6% in winter, respectively. This suggested that the annual average loss of CSP was 67.8% from the river to the estuary, which is consistent with profound recycling of riverine POM (≥60%) (Hedges and Keil, 1999). This high CSP loss might be associated with bacterial degradation, dilution by the mixing water, lower phytoplankton production downstream, and/or deposition to sediment. In the PRE, particles are classed depending on the hydrodynamic condition. CSP aggregates of high density, i.e., CSP flocculated with mineral or fine sand, are more prone to deposition, whereas particles with low density, e.g., free CSP, may be transported downstream or upstream by the tidal current (Sun et al., 2012). During the transport process, CSP could be involved in a series of deposition-resuspension cycles along with the time scale of estuary mixing. The residence time of the PRE is about 5 days in the wet season and 30 days in the dry season (Han, 1998). As proteinaceous particles can be rapidly turned over, this scale of residence time likely complies with the degradation time of CSP. Microbial colonization and degradation reduce CSP size and decompose CSP into DOM (Ziervogel et al., 2016). For the same reason, higher biomass of bacteria in the upstream than the downstream in winter (Liu et al., 2007) might permit bacteria to degrade CSP more sufficiently in the upper of the estuary than more oceanic stations (S6, S7, and S8), which may explain that the concentration of CSP in the river section is lower than the mouth of the PRE. Therefore, it is unlikely that large amounts of riverine CSP survive the transport from the river to marine sediments (Chen et al., 2004; Zhao et al., 2009; He B. et al., 2010). This suggested that interplay between grain-size sorting due to hydrodynamic conditions and microbial degradation determined the fate and distribution of CSP in the mixing zone of the PRE.

Amino Acids are reliable indicators of OM degradation and bioavailability (Hedges et al., 1994; Dauwe et al., 1999). AAs, Gly, and Ser are well-enriched in old OM (Keil et al., 2000; Chen et al., 2004), while biogenic particles (fresh and liable organic material) contained more AAs GlX+Phe+Leu+Ala+Val in the PRE (Chen et al., 2004). The percentage of particular nitrogen attributed to phytoplankton is generally in a constant range given phytoplankton cells are sufficient in nutrition, and the ratio of Chl a to PN can reflect the fraction of phytoplankton nitrogen, non-phytoplankton and detrital nitrogen (Yentsch and Vaccaro, 1958). Therefore, the ratio of Chl a/CSP was taken as proxy for labile protein contribution during this study, assuming alga are the freshest source of CSP among POM sources. The sum of (GlX+Phe+Leu+Ala+Val) mole% was positively related to the ratio of Chl a/CSP (p < 0.001, r² = 0.46, n = 16) (Figure 8A). In contrast, there was a significant negative correlation between the ratio of Chl a/CSP and mole percentage of Gly (p < 0.001, r² = 0.56, n = 16) (Figure 8B). The spatial distribution of the ratio of Chl a/CSP was characterized with the higher values at S8 and S1 (average value of 0.040 and 0.039, respectively) compared with those at S4 (0.022) and S6 (0.019). Moreover, the ratio of Chl a/CSP was 48.67% higher in surface than that in bottom waters except for S1 during summer (Supplementary Table 1, the percentage of increase of the ratio of Chl a/CSP in surface over that in the bottom water is presented in the online supplement). This observation is in accordance with findings of higher moles% Gly and Ser in bottom than in surface waters. Our results showed that the contribution of freshly produced CSP were lower at S4 and S6 as well as in the bottom waters, which complies with previous observations indicated by AAs distribution and δ¹⁵N (Chen et al., 2004; Ye et al., 2016; Huang et al., 2020). The study by Chen et al. (2004) showed AAs concentrations were low in waters of Humen outlet downstream (zone 2), increased to the highest in the estuary plume (salinity 10–25) and decreased beyond the coastal plume during summer. In particular, the fresh organic material (biogenic particles, Chl a ≥ 5 μgL^–1 and SS < 10 mg L^–1) in the estuary plume contained more AAs (Glu+Phe+Leu+Ala+Val) and had higher AAs/HAs ratios, while OM were more degraded in the upper of the zone 2 indicated by lower Glc-NH₂/Gal-NH₂ (glucosamine/galactosamine) and AAs/HAs ratios (Chen et al., 2004). For the nitrogen isotopic composition of POM, owing to the preferential loss of ¹⁴N and subsequently leaving the remaining PON enriched in ¹⁵N during the degradation processes of OM, high δ¹⁵N in the bottom water were found in the PRE (Ye et al., 2016; Huang et al., 2020). It suggested that particulate OM in turbidity zone of the estuary and in bottom was well-degraded and relatively refractory, while particulate OM in high salinities waters and in surface water was more recently produced. It should be noted that the sources of CSP (allochthonous or autochthonous) are complicated concluding terrestrial soil, in situ freshwater and marine plankton production as well as their detritus, and the freshly produced CSP was quite reactive and easily degradable. Therefore, Chl a/CSP could vary spatiotemporally in estuary. Future studies may need to focus on the combined effects of physical and biogeochemical process on CSP formations and removals in estuaries with large freshwater input and biomass production.

Most previous studies for CSP were based on the microscopic method developed by Long and Azam (1996; Table 3). Cisternas-Novoa et al. (2014) proposed a spectrophotometric method to measure CSP concentration. Spectrophotometric analyses were validated by the microscopy, and allowed for estimation of CSP composition in terms of nitrogen. In this study, CSP-N was positively related to PN except for the upper river in the dry season and in bottom S6 during summer, which have an obvious deviation to fitted curve (Figure 6E). The deviation in the range of PN > 20 μmol L^–1 and 2 < CSP-N < 4 μmol BSA-N-eq. L^–1 may be attributed to the disproportional reduction of CSP and PN input in the upper river during the dry season being low discharge and the difference of lability between CSP and PN. The maximum of CSP concentration in the bottom water during summer was observed at S6. This deviation at S6 might be ascribed to two reasons: (1) the resuspension of sediment; (2) the estuarine circulation with CSP coming from sinking phytoplankton in the plume and being transported by the salt wedge advancing upstream near the bottom.

Estimated contribution of CSP to PN was comparable with that of PHAA, yielding average of 32.9%. A recent study has shown that the input of PN (allochthonous and autochthonous) in the PRE was about 2.0 × 10⁷ mol N d^–1 (Tan et al., 2019). Thus, CSP-N flux in the Pearl River Delta might be as high as about 6 × 10⁶ mol N d^–1. These results indicate that CSP is a potentially important PN pool in the Pearl River Delta.

Labile AAs are one important factor limiting OM degradation in anoxic waters (Pantoja et al., 2004, 2009). In the PRE, waters were characterized with a heavy load of CSP concentration in the fresh-water segment where oxygen depletion often occurs (Zhai et al., 2005; Dai et al., 2006); therefore CSP degradation may influence remineralization and nitrification, and potentially may contribute to hypoxia in the upper reach of the PRE (Xu et al., 2005; Dai et al., 2006; He B. Y. et al., 2010). Recently, Tan et al. (2019) found that ∼74% of sedimentary nitrogen removal (8.1 ± 2.0 × 10⁶ mol N d^–1) from the Pearl River Delta sediment is dependent on PN from water deposited onto sediments, and about 6 × 10⁶ mol N d^–1 of PN from water is mineralized to produce ammonium, and ultimately converted to N₂ or N₂O and released into the atmosphere in the PRE (Tan et al., 2019). Interestingly, the input flux of CSP-N (6 × 10⁶ mol N d^–1) from the PRE in our study is approximately equivalent to the sediment nitrogen loss observed by Tan et al. (2019). It is presumably assumed that denitrifying and anammox bacteria may preferentially utilize CSP over DOM and non-CSP PN sources in the PRE (Zeng et al., 2019).