Collateral implications of carbon and metal pollution on carbon dioxide emission at land-water interface of the Ganga River

Atmospheric CO2 source and sink is among the most debated issues that have puzzled climate change geochemist for decades. Here, we tested whether heavy metal pollutants in river sediments favor preservation of organic matter through shielding microbial degradation. We measured CO2 emission and extracellular enzyme activities at land-water interface (LWI) of 7 sites along a 285 km main stem of the Ganga River and 60 locations up- and downstream of two contrasting point sources discharging urban (Assi drain; Asdr) and industrial (Ramnagar drain; Rmdr) wastewaters to the river. We found the lowest CO2 flux at Rmdr mouth characterized by the highest concentrations of Cu, Cr, Zn, Pb, Ni, and Cd. The fluxes were relatively higher at locations up- and downstream Rmdr. Substrate induced respiration (SIR), protease, FDAase, and β-D-glucosidase all showed a similar trend, but phenol oxidase and alkaline phosphatase showed opposite trend at the main river stem and Asdr. Sites rich in terrestrially derived organic matter have high phenol oxidase activity with low CO2 emission. The CO2 emission in the main river stem showed curvilinear relationships with total heavy metals (∑THM; R2 = 0.68; p < 0.001) and TOC (R2 = 0.65; p < 0.001). The dynamic fit model of main stem data showed that the ∑THM above 337.4 µg g-1 were able to significantly decrease the activities of protease, FDAase, and β-D-glucosidase. The study has implications for understanding C-cycling in human-impacted river sediments where metal pollution shields microbial degradation consequently carbon and nutrient release and merits attention towards river management decisions.


Introduction
The loading of terrestrially derived organic carbon (OC) influences ecosystem functioning in rivers and streams along hydrologic continuum, particularly through the processes controlled by microorganisms (Tezuka 1990;Taylor and Townsend 2010;Berggren and Giorgio 2015). Rivers and streams are the primary routes of delivering dissolved and particulate organic carbon emerging from terrestrial landscapes through inland waters to coastal areas and oceans (Cole et al. 2007;Park et al. 2018;Siddiqui et al. 2020).
Increased interest in understanding global carbon cycle, as it relates to climate change, has encouraged researchers to work on carbon capture, storage and release in rivers, oceanic and terrestrial systems (Jaiswal and Pandey 2019a). Land-derived OC supply is an important energy source for riverine food webs and ecosystem functions, including C and nutrient cycling (Boyero et al. 2011;Pandey et al. 2014;Gessner et al. 2010;Rosemond et al. 2015). A large amount of terrestrially generated OC enters into rivers and streams where it is decomposed/accumulate in bed sediment and transported downstream (Foulquier et al. 2015;Austin and Vivanco 2006). A substantial portion of terrestrial plant litter entering forested headwaters is processed immediately by heterotrophic microbes and invertebrate shredders, resulting in significant quantities of fine particulate organic matter which is transported downstream (Vannote et al. 1980). Furthermore, point sources releasing sewage and industrial effluents and non-point sources such as surface runoff are continuing to add massive amount of C and nutrients to streams and rivers (Bernot et al. 2010;Yadav and Pandey 2017). The presence of both the C-sources (allochthonous and autochthonous) allows uncoupling of primary and secondary C-production and has implications for food web dynamics, CO 2 emission, and riverine transport (Van den Meersche et al. 2009;Lau et al. 2009).
Being a conducive habitat for microbial metabolism, inland waters act as effective bioreactors, where terrestrial organic matter is rapidly decomposed (Richey et al. 2002). The organic matter is relatively short-lived in inland waters (mean residence time 2.5 ± 4.7 years), as compared to centennial to millennial-scale residence times in soils and oceans (Catalán et al. 2016). Because of multiplicity of factors that regulate organic matter turnover, such as C-chemistry, sediment characteristics, microbial composition, hydrologic control, and disturbance, forecasting long-term stability of carbon compounds in large rivers is difficult (Kandeler et al. 1999;Dai et al. 2002;Schimel and Weintraub 2003). The microorganisms regulate biogeochemical transformations including the release of carbon dioxide (CO 2 ) (Guérin et al. 2006;Cole et al. 2007). Microbial degradation of organic matter in oxic sediments produces mainly CO 2 . Because most of the inland waters are saturated with C gases, they constitute a net source of CO 2 to the atmosphere and are quantitatively important in ecosystem C-budgets (Butman and Raymond 2011;Aufdenkampe et al. 2011;Teodoru et al. 2015). Despite these facts, the C-metabolism and CO 2 emissions of large rivers remain poorly quantified. Surface waters receive about ̴ 1.9 Pg of carbon annually from different natural and anthropogenic sources. About 0.2 Pg of this C undergoes burial in sediments, and about 0.9 Pg is flushed to the ocean. The balance amount (0.8 Pg) returns to the atmosphere (Cole et al. 2007). Global emission of CO 2 from lotic waters is about 1.8 ± 0.25 Gt CO 2 -C yr −1 (Raymond et al. 2013). This amount is equivalent to about 70% of the oceanic C sink globally (Le Quéré et al. 2014). According to Cole and Caraco (2001), the world's 80 largest rivers add about 0.15-0.3 Pg CO 2 -C annually, whereas the inundated floodplains in humid tropics emit ̴ 0.9 Pg CO 2 -C annually. This amount is about 3-fold of global riverine emission (Richey et al. 2002). The US rivers and streams release about 2370 ± 800 g CO 2 -C m −2 yr −1 (Butman and Raymond 2011). Because of strong hydrologic control and heterogeneity of sub-habitats, quantification and prognostics modeling of CO 2 emission in large rivers are very difficult (Jaiswal and Pandey 2019a).
Our recent, multi-spatio-temporal scale studies show that, being a stable and biologically active zone, the landwater interface (LWI) of large rivers provides an imprint of hydrodynamic retention characterizing the processes and patterns in close correspondence with those occurring in the main river channel (Jaiswal and Pandey 2019a). The LWI can be used as a stable testbed and predictor of river ecosystem responses to human disturbances (Jaiswal and Pandey 2019a). The β-D-glucosidase, protease, and FDAase are widely used to address soil/sediment microbial processes. The β-D-glucosidase catalyzes the terminal step in cellulose breakdown with subsequent release of glucose monomers to the microorganism (Dick 2020). The activity of this enzyme varies with soil/sediment types and positively correlates organic matter supply (Sinsabaugh et al. 2008;Yan et al. 2010;Mariscal-Sancho et al. 2010). Understanding the variations in β-D-glucosidase activity can help understanding sediment/soil C-metabolism and C-cycling (Jaiswal and Pandey 2019a). The fluorescein diacetate hydrolytic assay (FDAase) is a widely accepted measure of total microbial activity in soils (Schnürer and Rosswall 1982;Battin 1997;Adam and Duncan 2001). The fluorescein diacetate hydrolysis requires three major group of enzymes, esterases, proteases, and lipases involved in the microbial decomposition of organic matter (Schnürer and Rosswall 1982). Also, it is used as a potential indicator of esterase activity of stream biofilms (Battin 1997). Phenol oxidase enters the environment through excretion or lysis and mediates lignin degradation and C-mineralization (Sinsabaugh 2010).
Aquatic ecosystems serve as reservoirs of terrestrial and anthropogenic-C accumulation. C-enrichment interacts with site-specific conditions, including occurrence of other contaminants. Thus, the rate of decomposition vis-à-vis CO 2 emission varies subject to the nature of C-source and sitespecific conditions, for instance, presence of other contaminants. Anthropogenic input of C is often accompanied by a large supply of nutrient and non-nutrient contaminants (Smith and Schindler 2009). Thus, the fate of C in such water bodies can be strongly influenced by the interaction of multiple pollutants. For instance, nutrient supply, chiefly N and P, can gear up autochthonous C-build-up and decomposition, whereas metal enrichment can provide countervailing effect. Thus, the complex interactions, especially those pertaining to countervailing effect, deserve scaled-up studies from a climate change perspective and for understanding ecosystem level effects of heavy metal pollution. This has concern for the Ganga River because sewage sources alone add 110 Gg of TOC, 13.28 Gg of DIN, and 5.29 Gg of DRP annually into the river. In addition, the basin receives about 1.81 Tg, 2.77 Tg, and 0.13 Tg of TOC, TIN, and P, respectively, through atmospheric deposition annually (Singh and Pandey 2018).
Quantifying CO 2 emission and C-sequestration in large rivers requires an understanding of factors that regulate microbial processes-linked C-metabolism. The processes which inhibit microbial activities reduce CO 2 emission. Metals, for instance, inhibit microbial activity (Ford 2000;Verma and Pandey 2019) and, consequently, the CO 2 release (Dobler et al. 2000). Because most of the human impacted rivers simultaneously receive C and metal pollutants, it is expected that C enrichment will enhance carbon dioxide emission, whereas metal pollutants will cause a stabilizing effect. Thus, countervailing effect of such disturbances is likely to create a complex pattern of C-dynamics challenging to predict CO 2 release/C-sequestration (Jaiswal and Pandey 2019a). A clear understanding of such countervailing effects leading to unanticipated ecosystem consequences is essential for policy decision in river management (Jaiswal et al. 2021). The purpose of this work is to fill the gap in our understanding of how point-and non-point sources of C and metal input affect C-metabolism and CO 2 emission at the LWI of the Ganga River. Here, we have attempted to establish mechanistic links between C and metal pollutants to develop a new context to understand collateral implications of human-driven stressors. The study was designed in two parts: first, to evaluate the effects of two contrasting point sources, one releasing metal-rich and the other releasing C-rich wastewater, on microbial extracellular enzyme activities and CO 2 emission at LWI, and second, to assess the collateral impacts and linkages with regulatory determinants at selected sites along the main stem of the Ganga River. We attempted to identify patches with variabilities and mechanistic regulatory control which can be used to quantitatively extrapolate system level CO 2 emission scenario for large rivers. The study has relevance in up-scaling system-specific scenario for improving prognostic models of CO 2 emission in addition to evaluating collateral implications of C and metal pollution in human-impacted large rivers.

Study area
A 2-year study (2017 to 2018) was conducted between Allahabad (25°28′N; 81°54′E) and Varanasi (25°19′N; 83°1′E) covering 285 km of the Ganga River. The sampling was done during low flow at six sites: Hatwa (Htwa), Sangam (Sngm), Sitamarhi (Stmh), Adalpura (Adpr), Ramna (Rmna), and Rajghat (Rjht). We consider the least-polluted site, Devprayag (Dvpg), ̴ 812 km upstream to Hatwa, as a reference site. Also, we selected two contrasting point sources in Varanasi region (Assi drain, which discharge 66.4 mld urban sewage and Ramnagar drain, which discharge 24 mld of metal-rich industrial effluents) for trajectory analysis. For trajectories, we collected samples from 30 locations (each 100 m apart), selecting 15 upstreams and 15 downstreams of each point source (Fig. 1). The Ganga basin, a mixed landscape with predominance of highly fertile agricultural plains (> 73% of the total area), support over 26% of India's population. The river experiences strong pressure from urban-industrial-agricultural releases and water regulation. The climate is tropical where March to June extends as a hot summer, July to October as a humid rainy season, and November to February as a cold winter. About 90% of the annual precipitation (870-1130 mm) occurs in the rainy season. The relative humidity ranged from 27 to 83% (summer) and 58 and 99% (rainy). Summer temperature maxima may reach 46°C.

Sampling and analysis
Triplicate samples were collected from land-water interface (LWI) of three sub-sites considered for each study site using sediment corers. Samples were brought to the laboratory and preserved (4°C) for the analysis of microbial variables. Samples were dried at room temperature, homogenized, and sieved (2-mm mesh) for heavy metals .After digestion in a tri-acid mixture at a ratio of 5:1:1 (HNO 3 /HCl/HClO 4 ) in a microwave digestion system (SINEO model MDS-6G), the samples were analyzed in a Perkin Elmer atomic absorption spectrophotometer (Analyst 800, USA). The detection limits were 0.0008 (Cd), 0.003 (Cr), 0.001 (Cu), 0.006 (Ni), 0.015 (Pb), and 0.001 (Zn) mg/l. The standard stock solutions and drift blanks purchased from Sisco Research Laboratories (India) were used to calibrate the AAS for respective metal. For spiking of samples, we prepare the solutions using nitrate salts of Cd, Cr, Ni, and Pb and granules of Zn and Cu (99% purity). To ensure recovery strength, we repeatedly tested ratios of tri-acids, and a 5:1:1 ratio showed ̴ 99% recovery.
The sediment samples were analyzed for granulometric fractions (Stemmer et al. 1998). Chlorophyll a, extracted in acetone, was determined spectrophotometrically (Lorenzen and Jeffrey 1980). Conductivity was measured using multiparameter tester (Oakton 35425-10, USA). The ammonium molybdate stannous chloride method was used to measure soluble reactive phosphorus (SRP; Murphy and Riley 1962). Total organic carbon (TOC) was quantified using a TOC analyzer (LOTIX TOC analyzer). Total nitrogen (TN) was analyzed using Kjeldahl analysis and water soluble organic carbon (WSOC) following by Ciavatta et al. (1991).
Following a chloroform fumigation extraction, microbial biomass C (C mic ), was measured as per Jenkinson and Powlson (1976), biomass-phosphorus (P mic ) following Brookes et al. (1982), and nitrogen (N mic ) following Shen et al. (1984). The substrate-induced respiration (SIR) was assessed following Wardle et al. (1993). The microbial metabolic quotient was calculated in terms of BR/SIR ratio. Protease activity was measured in terms of aromatic acid released during casein proteolysis and expressed as L-tyrosine equivalents (Ladd and Butler 1972). Phenol oxidase was determined as described by Sinsabaugh and Findlay (1995) and alkaline phosphatase (AP) by Tabatabai and Bremner (1969). The β-D-glucosidase was assayed following Eivazi and Tabatabai (1988). For fluorescein diacetate hydrolytic assay (FDAase), the sample filtrates were incubated with fluorescein diacetate, and the fluorescein formed was estimated spectrophotometrically (Schnürer and Rosswall, 1982).
The CO 2 at LWI was collected in steel chambers (23 cm height; 22 cm diameter). The chambers contain a small fan for distribution of air and a thermometer to record temperature. The chambers were fixed 5-cm deep to ensure a seal against ambient air, and the gas was withdrawn through a rubber fitted valve. For quality control, and to avoid changes in temperature and moisture regime, attempts were made to collect samples at an identical time. A polypropylene syringe with three-way stopper was used for withdrawing the gas from the headspace at 1-h interval. Samples were brought in an icebox and analyzed in a Trace 1110 Thermo Fisher Analyzer. The accuracy was validated by repeated calibration with standard gas samples.

Statistical analysis
Mean values of measurements are presented with standard errors (SE). Correlation coefficients and dynamic fit models were used to measure the extent to which variables covary. Principal component analysis was considered to verify major determinants. Confidence limit > 95 justifies significance. The Sigma plot (version 11.0) and SPSS (IBM SPSS statistics 20) were used for the statistical analysis. To find Euclidean distance and extent of dissimilarities, we used non-metric multidimensional scaling (NMDS) and Bray-Curtis neighbor-joining metrices.

CO 2 emission
The CO 2 flux at land-water interface (n = 24) varied between 83.9 mg m −2 h −1 (Rjht Site) and 339.2 mg m −2 h −1 (Sngm Site). The emission flux increased over time except at Rjht where it declined marginally on a temporal scale. At Sngm, the CO 2 emission was 4-fold higher than those at Rjht Site (Fig. 2).

Regulatory variables
Substrate-induced respiration (SIR) showed a trend similar to CO 2 emission. The enzyme activity-β-D-glucosidase, protease and FDAase did show maxima at Sngm Site and minima at Rjht Site with n = 24, whereas the microbial metabolic quotient (qCO 2 ) showed an opposite trend (Fig. 2). The phenol oxidase and alkaline phosphatase showed highest activity at Stmh Site and lowest at Rjht (Fig. 2).

Sediment characteristics
The concentration of metals in the riverbed sediment (n = 24) ranged from 42.2 to 161 µg g -1 (Zn), 35.2 to 188.2 µg g -1 (Cr), 10.2 to 170.2 µg g -1 (Cu), 10.5 to 82.4 µg g -1 (Pb), 0.5 to 7.8 µg g -1 (Cd), and 5.3 to 60.2 µg g -1 (Ni). The values were highest at Rjht followed by Sngm and lowest at Dvpg. The concentrations were highest for Cr and lowest for Cd (Fig 3). In the main river stem, the concentrations of total organic carbon (TOC), water soluble organic carbon (WSOC), and total nitrogen (TN) with sample size n =24 were high at downstream sites. At Rjht Site (n = 24), the concentrations of TOC, WSOC, and TN were 6-fold, 2.4-fold, and 4.9-fold higher compared to their respective values at reference site Dvpg. The C:N ratio also was highest at Rjht Site. Microbial biomass carbon (C mic ), nitrogen (N mic ), and phosphorus (P mic ) and Chl a biomass showed trends similar to CO 2 emission, with values being highest at Sngm and Lowest at Rjht Site. Conductivity and SRP (soluble reactive phosphorus) also followed trends similar to TOC. The proportion of coarse sand was relatively higher at upstream site (Dvpg), and percentage of fine sand was highest at Adpr while clay was highest at Rjht (Table 1). The CO 2 emission showed curvilinear relationships with ∑THM (R 2 = 0.68; p <0.001), TOC (R 2 = 0.65; p <0.001), and phenol oxidase (R 2 = 0.72; p < 0.001) and significant positive correlations with FDAase, protease, and β-D-glucosidase (R 2 = 0.89-0.92; p < 0.001) (Fig. 4).

CO 2 emission
The CO 2 emission showed longitudinally variable trends at both the point sources. The CO 2 emission was found highest at the mouth of the Assi drain (Asdr; 432 mg m -2 h -1 ; n = 24) and decreased towards downstream and upstream sites. At the mouth of the Ramnagar drain (Rmdr), the CO 2 emission did show the lowest value (10 mg m -2 h -1 ). At Rmdr, the CO 2 emission maxima at upstream sites of the drain was found 25-fold higher as compared to the mouth of the drain, whereas downstream maxima was over 38-fold higher. The CO 2 emission showed a gradual increase with distance upstream and downstream from mouth of the drain. The CO 2 efflux showed a wide variability and increased over time at Asdr but declined at Rmdr (Fig. 5).

Regulatory variables
To synchronize with CO 2 emission, we measured microbial/EE activities with same number of replicates/sub-sites (n = 24). At Asdr, the microbial/enzyme activities (SIR, FDAase, β-D-glucosidase and protease) all followed a trend similar to CO 2 emission with values being highest at the drain mouth. Metabolic quotient, phenol oxidase, and alkaline phosphatase showed opposite trends (Fig. 5). At Rmdr, microbial/enzyme (SIR, FDAase, β-D-glucosidase, protease, alkaline phosphatase, and phenol oxidase) followed a trend similar to CO 2 emission with values being lowest at the drain mouth and increased towards downstream. Metabolic quotient showed an opposite trend (Fig. 5).

Principal component analysis
Principal component analysis (PCA) showed three significant PCs for the main stem and Asdr and two significant PCs for Rmdr (Table 2). For the main stem, three PCs collectively explained 92.09 % of the variance. The first PC, which explains 62.12% of the variance, showed strong loading of heavy metals, qCO 2 , TOC, TN, WSOC, and sediment granulometry. Second PC explained 20.02% of the variance with strong loading of microbial activities, Chl a biomass, and CO 2 emission. Third PC have strong loading of alkaline phosphatase and phenol oxidase with 9.95% variance. The first PC at Asdr (66.12% variance) showed strong loading of microbial activities, TOC, nutrient, and granulometry while the second PC with 22.21% of variance showed strong loading of metals and qCO 2 . The third PC (8.9% variance) separates phenol oxidase and alkaline phosphatase. For Rmdr, two significant PCs were obtained. The first PC explained 67.13% of the variance with strong loading of metals, TOC, TN, C:N ratio, silt, clay, and qCO 2 , whereas PC 2 explained 27.04 % of the variance and separates microbial activities and silt/clay (Table 2).

CO 2 emission at land-water interface
The CO 2 emission reported here was significantly lower than those reported for a freshwater marsh (1456.60 ± 593.31 mg m -2 hr -1 ) and a brackish water marsh (1435.23 ± 689.71 mg m -2 hr -1 ) in the Min River estuary, Southeastern China (Hu et al. 2017). Mukhopadhyay et al. (2002) in their studies on monthly variations in the Hooghly River estuary found a CO 2 flux ranging from -5.5 to 154 mg m -2 hr -1 , which was lower than the values recorded here.  Selvam et al. (2014) reported that the Indian inland waters are an important source of CO 2 reaching to the atmosphere with CO 2 emission flux in all the 45 studied systems that ranged from -51.7 to 481.06 mg m -2 hr -1 . The upper limit of these values is comparable to the present study. Sawakuchi et al. (2017) reported an average value, including all seasons and sites, measured in the lower Amazon River and its tributaries to be 999.50 ± 896.5 mg m -2 hr -1 . The direct measurements in the lower Amazon River main stem as by Alin et al. (2011) show values ranging from 296.20 to 1682.20 mg m -2 hr -1 . Mann et al. (2014) found 572 to 2619.83 mg m -2 hr -1 at 25 sites during a single period (November 2010) from four major river basins in the Republic of the Congo (Alima, Lefini, Sangha, Likouala-Mossaka).These values are much higher than the values recorded in this study. Despite large spatial variation, we did not observe significant inter-annual difference in CO 2 emission. Sediment with favorable conditions tends to be substrate limited and shows high in situ activities and low organic matter accumulation. In contrast, sediments with stress conditions are activity limited and show low in situ activities, promoting C-storage (Sinsabaugh 2010). Also, the enzyme activity limited systems contribute to organic matter accumulation (Freeman et al. 2001(Freeman et al. , 2004. Earlier research shows that carbon quality and particle size distribution are an important regulator of microbial biomass distribution (Sinsabaugh and Findlay 1995). Small particle size support high concentration of TOC and TN, with concentrations generally high in clay, although the C:N ratio declines with diminishing particle size (Stemmer et al. 1998). Different soil particle size fractions possess different sizes of carbon pools (Von Lützow et al. 2007). Small clay and silt particles provide large specific surface areas with numerous reactive sites at which organic carbon can be sorbed by strong ligand exchange and polyvalent cation bridges (Feng et al. 2013). On the other hand, the large sand particles, which are dominated by quartz particles, exhibits only weak bond affinities to organic carbon (Balesdent et al. 1998;Christensen 1998;Kahle et al. 2002). The downstream sites, Sngm and Asdr, are rich in WSOC, TOC, and TN concentrations and also have high percentage of clay and silt at mouth of the drain. High substrate availability coupled with high percentage of silt and clay may invite the colonization of a large microbial population and create a suitable room for high microbial/ enzyme activities that trigger CO 2 emission. Because microbial extracellular enzymes (EEs) mediate the decomposition and mineralization of organic matter and heavy metals inhibit the EE activities, here we identified, in addition to other variables, two major determinants, chemical recalcitrance and heavy metal enrichment, to address changes in microbial extracellular enzyme (EE) activities vis-à-vis CO 2 emission along the river gradient.

Regulatory determinants: recalcitrance
Two different classes of mechanisms are used to explain selective preservation of organic carbon in soil and sediments. These include the following: recalcitrance, where carbon molecules are generally preserved owing to being unreactive, and interaction with minerals that help escape remineralization (Hemingway et al. 2019). Microbial EEs mediate decomposition and mineralization of organic matter (OM).Terrestrially derived OM often shows recalcitrance to degradation (Sinsabaugh 2010;Ward et al. 2013). Terrestrially derived OM, rich in leaf litter, are generally poor in N content (high C:N ratio) and are poorly decomposable relative to autochthonous C. Lignin and cellulose constitute the most abundant C fractions on land (Boerjan et al. 2003). Lignin accounts about 30% of organic-C in the terrestrial C reservoir (Boerjan et al. 2003), and ̴ 55% of land sequestered lignin is degraded in the river continuum (Malhi et al. 2008;Bose et al. 2009;Houghton et al. 2001). Roughly 80 Tg C in the Amazon terrestrial biosphere is fixed as lignin each year (Mayorga et al. 2005;Ward et al. 2013). Similar to CO 2 emission, microbial/EE activities, except protease, did not show significant difference between years 2017 and 2018. Protease showed a significant (p < 0.005) increase in 2018 at Sngm and Rmna Sites indicating input of some nitrogenous substrate at these sites. Sites with chemical recalcitrance were measured in terms of phenol/phenol oxidase activitya signature of terrestrially derived OC, and those rich in metal pollutants showed mismatch in CO 2 emission rates and C-inputs. Phenol oxidase is one of the key enzymes able to degrade recalcitrant phenolic materials such as lignin (McLatchey and Reddy 1998). Phenolic materials are highly inhibitory to enzymes, and their lower abundance at some sites allowed higher hydrolase activities. Sites, Stmh and Adpr, with high amount of terrestrially derived organic matter showed highest phenol concentration, and high C:N ratio Fig. 4 The dynamic fit models for CO 2 emission showing relationship with regulatory variables at land-water interface of the main stem of the Ganga River and CO 2 emission rates were not comparable to C inputs. The downstream sites, Sngm and Rjht, with easily utilizable soft-C, added from sewage for instance, showed less activity of phenol oxidase, but high activity of β-D-glucosidase. Sites with high concentration of Chl a (high autochthonous organic C) did show low phenol oxidase activity and high activity of β-D-glucosidase. The C:N ratio, a proxy of decomposability (Datry et al. 2018), was recorded lowest at highly productive site (Sngm). At the mouth of both the point sources (Asdr; Rmdr), the phenol oxidase activity was found to be the lowest which increased with upstream and downstream trajectories. Pandey and Yadav (2017) reported very low levels of DO close to the mouth of Assi drain. Thus, low amount of land-derived OC coupled with oxygen deficiency could decrease phenol oxidase activity, whereas the activity of hydrolases remained relatively higher even in an anaerobic environment (Lee et al. 1999). We also tested alkaline phosphatase (AP), an ectoenzyme of microbial origin, used as a measure of P status in soil and sediment (Sayler et al. 1979;Johnson et al. 1998;Sinsabaugh et al. 2009). The AP showed a trend similar to phenol oxidase, whereas sediment-SRP showed an opposite trend in the main stem as well as along point sources. Also, the sites with low productivity have high AP and phenol oxidase activities. Overall, these two enzymes can be used as an indicator of less polluted and low CO 2 emission sites, whereas high activities of other enzymes indicate sites acting as a hot spot of CO 2 emission.

Regulatory determinants: heavy metals, carbon and nutrients
We further tested whether heavy metal enrichment was able to reduce EEs activities and CO 2 emission in the main stem Point source up-and downstream trends in CO2 emission and regulatory variables at land-water interface of the Ganga River. Asdr, Assi drain; Rmdr, Ramnagar drain. The zero represents drain mouth, whereas -sign shows upstream influences and source orientated sites. Although we did not observe significant between year differences in heavy metal concentration, yet we found a large difference spatially. The countervailing effect of heavy metals at Rjht and Rmdr reduced CO 2 emission. The NMDS separated microbial biomass, EE activity, Chl a biomass, and CO 2 emission in one group between the Sngm and Rmna (Fig. S1). Asdr also showed the similar pattern (Fig. S2), whereas for Rmdr, the NMDS showed that the microbial activity follows a concurrence with downstream sites (Fig. S3).The granulometry and organic matter content affect the distribution of heavy metal (Farkas et al. 2007). Fine grained sediments tend to have relatively higher metal contents due in part to high specific surface of particles (Rubio et al. 2000). Positive correlations among metals, organic substrate, and sediment particle size validate this concept. A higher proportion of fine sized particles lead easy transportation of the sediment downstream (Bartoli et al. 2012). Accordingly, the river sites down the industrial and the urban areas were significantly more polluted than upstream sites. The Rjht and Rmdr Sites showed C-storage in bed sediment. Despite high substrate availability, there were relatively lower in situ microbial/ enzymatic activities. Many investigators have reported that heavy metal stressors can induce change in the microbial biomass and activity (Khan et al. 1998;Jaiswal and Pandey 2018). The microorganisms release extracellular enzymes (EEs) to activate the lysis of organic matter (Romaní et al. 2004). The EEs are produced as a result of cellular metabolism influenced by C and nutrient availability. The EEs are highly sensitive towards toxicants, including heavy metals, and respond sharply to even after small changes in sediment quality Jaiswal and Pandey 2018;Verma and Pandey 2019). The Rjht Site at the main river stem and the sites located downstream of Rmdr showed high concentrations of total heavy metals, nutrients, TOC, and high C:N ratios. Despite high TOC, an important resource/ substrate for microbial activity, reduced microbial biomass and enzyme activity at these sites indicated that, when metal concentrations exceed certain level (here, ∑THM > 337.4 µg g-1 ), the carbon favored microbial activity declines (Jaiswal and Pandey 2019a).
Nutrient supply stimulates microbial activity and decomposition of organic matter and, under appropriate temperature and water conditions, leads to high respiration and CO 2 emission (Liu et al. 2009). Because we did not observe significant inter-annual difference in carbon, nutrients, and other attributes, we presented data as means of 2017 and 2018. We found significant positive correlations between C availability and CO 2 emission at Asdr. However, at metal rich locations, this relationship implied up to certain levels only. As the concentration of metal increased, this relationship gone weaker. It seems that the microbial activities, and consequently the CO 2 release at LWI, are constrained by high concentration of toxic metals. Previous studies support these observations (Jaiswal and Pandey 2019a). At Rmdr, the CO 2 emission was found to be low at drain mouth followed by an increasing trend indicating that metal stressors inhibit enzyme activities and, thereby, the emission of CO 2 . At Asdr, the enzyme activities and emission of CO 2 showed a response almost synchronous to changes in the concentration of organic carbon. The C rich sites, characterized by faster rates of CO 2 emission, did show high activity of FDAase and β-D-glucosidase. The FDAase is used as a proxy of total microbial activity and organic matter turnover (Schnürer and Rosswall 1982), whereas the β-D-glucosidase is responsible for the hydrolysis of glycosidic bonds and used as marker of C acquisition (Sinsabaugh et al. 2009). The metal concentrations at Asdr did not appear high enough to induce inhibitory effects. On the contrary, the Rmdr Sites showed metal concentrations high enough to suppress enzyme activities and, consequently, the CO 2 emission. The carbon as a Fig. 7 The dynamic fit models for CO 2 emission showing relationship with regulatory variables at land-water interface of the two point sources of the Ganga River substrate stimulates EE activity, while metal stressors can suppress it. However, under high concentrations of C and metals, the inhibition by the latter might counterbalance the substrate (carbon and nutrient)-driven stimulatory effects Pandey 2018, 2019b). The dynamic fit model (Fig. 4) showed that total metal concentration above 337.4 μg g -1 induces adverse effects on CO 2 emission. In all the three subsets of studies, PCA separated metals opposite to CO 2 emission and microbial/enzyme activities. The NMDS yielded almost similar results. The Asdr Site showed low level of total heavy metal (ΣTHM < 337.4 µg g-1 ), with microbial activities relatively higher, and consequently the excess supply of C could assume importance triggering CO 2 emission. Previous studies have shown a marked decrease in C mic and enzymes activities in metal polluted soils and sediments (Abaye et al. 2005;Li et al. 2009). We found significant positive relationships between qCO 2 and metal concentrations. Earlier studies show that C mic declines and qCO 2 increases as metal concentration increases (Liu et al. 2012). Metal-associated changes in C-utilization have been considered a microbial response to stressors and are used as an indicator of metal pollution (Brookes 1995). Overall, the patterns of CO 2 emission shown here demonstrate that the distribution of co-occurrence of metal and C-sources ultimately regulate the pattern of carbon sequestration in the Ganga River. Similar spatial trajectories can be expected for other human-impacted rivers suggesting the need of caution for regional C-budgeting and modeling.

Conclusions
The auto-and allochthonous C input in human impacted large rivers is continuously increasing, and these increases are often paralleled by almost similar increase in metal pollutants. Many lines of evidence presented here indicate that although CO 2 emission in human-impacted large rivers is controlled by multiplicity of factors including chemical recalcitrance, high concentration of heavy metals suppresses the microbial extracellular enzyme activities responsible for organic matter decomposition and, consequently, the magnitude of CO 2 emission. Although we still have very less knowledge of how and to what magnitude metal enrichment will effect organic matter decomposition and CO 2 release, our study clearly indicates that at source oriented locations, total metal concentration was sufficient to suppress microbial/enzyme activities vis-à-vis CO 2 release. We conclude that the overall impacts of heavy metal pollution on large rivers are probably much larger than what is generally predicted. When riverine-C sequestration is considered, a completely different scenario emerges: A transition towards metal pollution leads to increase C storage relative to loss as CO 2 . This has implications for understanding C-cycling in human-impacted river sediments where metal pollution shields microbial degradation and, consequently, C and nutrient release. While this type of analysis is urgently needed from a climate change perspective, further studies need to include C-stabilizing effects for assessing ecosystem level impacts of heavy metal pollution for management and rejuvenation of large rivers.