δ15N Values in Crassostrea virginica Shells Provides Early Direct Evidence for Nitrogen Loading to Chesapeake Bay

Crassostrea virginica is one of the most common estuarine bivalves in the United States’ east coast and is frequently found in archaeological sites and sub-fossil deposits. Although there have been several sclerochronological studies on stable carbon and oxygen isotopes in the shells of this species, less is known about δ15N values within their shells, which could be a useful paleoenvironmental proxy to assess estuarine nitrogen dynamics. Modern C. virginica samples were collected in Chesapeake Bay for comparison with archaeological shells from nearby sites ranging in age from ~100 to 3,200 years old. Left valves were sampled by milling the hinge area and the resulting powder was analyzed for %N and δ15N values. Comparison of δ15N values between C. virginica shells shows relatively constant values from ~1250 BC to ~1800 AD. After ~1800 AD, there are rapid increases in 15N enrichment in the shells, which continue to increase in value up to the modern shell values. The increase in δ15N values is evidence of early anthropogenic impact in Chesapeake Bay. These results corroborate the observation that coastal nitrogen pollution occurred earlier than the 19th century and support the use of oyster shell δ15N values as a useful environmental proxy.

preservation over time. Mollusk shell material accretes sequentially, sometimes in annual growth layers, which could potentially be used as a high resolution temporal proxy for historical nitrogen levels, as well as creating a time-series of nitrogen loading over the life of the organism 11,26,27 . Furthermore, the shell material is never metabolized, so it can record the environmental conditions over a longer period of time than soft tissues 12,26,27 . It has previously been established that the calcite shell organic fraction chemistry reflects that of the organisms' soft tissues, with minimal variability in δ 15 N values (~0-2‰) between the soft tissues and shell material 27,28 .
Here we determine if C. virginica shells record historical evidence of N levels in the Rhode River basin in Chesapeake Bay during the late Holocene to assess when anthropogenic N loading began. In doing so, we seek to better validate and establish this novel environmental proxy. To test for Chesapeake Bay nitrogen in historic and prehistoric mollusk shells, we sampled archaeological and modern (living) oysters from the Rhode River Estuary at the Smithsonian Environmental Research Center in Edgewater, Maryland (38°53′N 76°32′W) (Fig. 1). Our archaeological samples date to the Woodland Period (including the Early [1200-500 BC], Middle [500 BC-AD 900], and Late [AD 900-1600]), and the 18 th and 19 th centuries 21 . During the Woodland period in the Chesapeake, Native American peoples practiced a largely hunting and gathering lifestyle and their populations appeared to expand during the Middle Woodland and especially during the Late Woodland, when maize agriculture appeared in parts of the region [29][30][31] . European colonization intensified in the area in the 17 th century, with colonial settlements clearing the forests and establishing large plantations on pristine agricultural land 32,33 . During the 19 th century, industrialization rapidly increased in the eastern US and continues today 33 .
Although Native Americans practiced agriculture and cleared forests in the region, these impacts were more localized and less widespread than the major landscape clearance efforts of Euro-American colonists and later peoples from the 17 th century onward. Historic period land clearance resulted in significant sedimentation 34 , and likely nutrient loading, into the Bay. These data lead Kirby and Miller 34 to hypothesize that an increase in observed growth rates of oysters from archaeological sites on the Potomac and Patuxent rivers during the 18 th and early 19 th century may have been due to an increase in planktonic primary productivity from nutrient loading during early eutrophication. The initial period of faster growth was later followed by a size decrease that they speculate was from intensive eutrophication and environmental deterioration, such as increased occurrences of harmful algae and parasitic disease after AD 1860. They speculate that other environmental changes, such as changes in water chemistry and an increase in the degree of fishing disturbances, in Chesapeake Bay during this time could affect the growth rate of C. virginica, but these environmental conditions are unlikely to cause as significant of an effect in shell growth rates as eutrophication. Our δ 15 N data provide a means to test this hypothesis and to help develop a novel proxy for assessing N in marine shellfish.

Results
One-way ANOVA tests for δ 15 N values between time periods produced an F-value of 297.3 and a significance of p < 0.001 at α = 0.05 (data summarized in Table 1). This indicates that the sample variance between different time periods was greater than the variance within time periods. A Tukey Post Hoc test yielded results that show that archaeological samples from the Early Woodland (1250-800 BC) are significantly different from Middle Woodland (500 BC-AD 900), 19 th century, and modern samples. Archaeological samples from the Middle Woodland were significantly different from Early Woodland, latter 19 th century, and modern samples. Archaeological samples from the Late Woodland (AD 900-1600), 18 th and early 19 th century were significantly different from latter 19 th century and modern samples. Later 19 th century and modern shells were significantly different from all other time periods (Fig. 2).
The F-value calculated by a one-way ANOVA for percent nitrogen (%N) was 34.8 with a significance of p < 0.001 at α = 0.05. A Tukey Post Hoc test yielded results that show archaeological samples from the Early, Middle, and Late Woodland periods were significantly different only from the modern samples (data summarized   in Table 2). The 18 th and 19 th century samples were only significantly different from the modern samples and the modern samples were significantly different from all other time periods (Fig. 3).

Discussion
Our results indicate increased N loading to Chesapeake Bay started as early as the 19 th century and substantially increased to the present. Time-series comparisons of the sampled shells display distinct trends, with both %N and δ 15 N values following a roughly exponential curve with higher concentrations of N and more 15 N enriched shells occurring in the 19 th century and modern samples. The %N remains more constant than the δ 15 N values, but increases ~0.07% from the 18 th century to modern shells (Fig. 3). The δ 15 N values remain relatively constant from the Early Woodland until the first part of the late 18 th to 19 th century (i.e., 1750-1800) but then increase ~3‰ from the late 18 th century to mid late 19 th century (i.e., 1850-1900) (Fig. 2). There was another substantial increase in δ 15 N values (~1‰) between the mid to late 19 th century shells and the modern collected shells, which is correlated to a 10-fold increase in reactive N species globally from 1860 to 1990 35 . δ 15 N values in mollusk shells have been shown to faithfully track the values in soft tissues and ambient POM 26,27,36 . In contrast, variation in mollusk shell %N has not been validated as a meaningful proxy for ambient N concentrations and additional analysis of modern samples is required before interpretation of such data is possible. Although the relatively constant %N suggests little to no N loss from the shells over time, our discussion will focus solely on δ 15 N values.  We interpret this substantial increase in δ 15 N values to be increased anthropogenic inputs into Chesapeake Bay during the 1800s. During the Woodland Period, it is unlikely that Native Americans 37 had any significant impact on the N loading of the bay. While there was urban development during the 18 th century, the impacts were likely more local in scale, as population levels were comparatively low in much of the region.
Our results are similar to several sediment nutrient loading studies in Chesapeake Bay over the last ~3000 years, with the exception of when δ 15 N values began to rapidly increase within the Bay 38 . Bratton et al. 35 observed significant increases in δ 15 N values from 1750-1800 AD, whereas we determined a later time period of 15 N enrichment. This is likely due to sampling location, however, since Bratton et al.'s study locations were closer to the Susquehanna River, which is a larger source of N for Chesapeake Bay 35 than our study location. Zimmerman and Canuel 38 also show increases in δ 15 N values beginning between the 18 th and 19 th centuries depending on study location within Chesapeake Bay, with their more northerly collected sediment core showing 15 N enrichment ~100 years before their southern core. A study location farther north in the bay would also not have as long of a period to mix with estuarine waters before deposition, so it could potentially record higher δ 15 N values at an earlier time period. Another possibility for the timing differences between this study and Bratton et al. 35 is due to the dating methods used. While our study solely used 14 C from corrected and calibrated archaeological oysters 21 , Bratton et al. 35 used pollen stratigraphy, 14 C of shell material within the cores, and the short-lived radioisotopes 137 Cs and 210 Pb within recently deposited sediment 4 . Since pollen was primarily used to relatively date the sediment, it is possible that the absolute dates obtained by their 14 C analysis could permit larger error in the dating due to reworking and bioturbation of the sediments.
During the 19 th century, population sizes and industrialization in the northeast US expanded exponentially 39 , which likely had a profound impact on the ecological health of the Bay. Between 1830 and 1880, over 80% of the forest surrounding Chesapeake Bay had been cleared, which, in combination with plowing for agriculture, greatly increased the sediment accumulation in the bay 4,5,20,35,[39][40][41][42] . During this time, the population size nearby Chesapeake Bay increased threefold 39 , and the increased amounts of sewage discharge and erosion caused by plowing likely increased the N inputs and 15 N enrichment in the Bay 24 . The significant decrease in oyster populations at this time also contributed to the failing health of the Bay due to decreased filtration rates of the bay water 39 and increased sedimentation rates which correlates to previous findings of decreased diatom diversity and increase in the centric:pennate diatiom ratio since the beginning of the 19 th century 4 . Therefore, it is likely that the POM and sediment of the bay were more enriched in 15 N. Consequently, we would expect a substantial shift in δ 15 N values during the 19 th century due to increased amounts of sewage and eroded soil entering the bay, which was evident in the results of our shell study.
A similar pattern was present in size changes in archaeological oysters. Archaeological size data from throughout the Chesapeake Bay indicate a significant increase in oyster size during the Historic period (18 th to 19 th centures) when compared to earlier Woodland times 34,37 . Research at sites on the Potomac and Patuxent rivers documents this trend in detail, with oysters first increasing in size during the 18 th and early 19 th centuries and then decreasing dramatically after AD 1860, perhaps from intensive eutrophication 34 . It has been suggested that the degree of organic matter loading into Chesapeake Bay has significantly increased within the last two centuries 38,43,44 , however, Cornwell et al. did not determine substantial increases in N deposition within the sediment, which could be due to increased algal blooms in the water column 38,44 and metabolic activity of estuarine organisms 43 . Our data support this hypothesis, providing the first direct evidence of N loading in this location of the bay during the 19 th century, particularly in the latter half of the 19 th century.
The potential effects of diagenetic alteration of %N and δ 15 N values must be considered. Loss of shell organic matter, and consequently N, likely occurs after burial and may be evident in the %N trends noted in our study. We argue, however, that such loss would not explain the observed δ 15 N trends. First, the observed trend in δ 15 N values would require preferential loss of 15 N, which is contradictory to the kinetics of stable isotopes. Secondly, it is difficult to imagine why diagenetic influence of N isotopes would stop in pre-seventeenth century shells, and even reverse to some extent in the earliest samples. Third, Black 28 was not able to influence shell δ 15 N values during a series of prehistoric cooking methods studies unless the shell was heated to the point at which it began to physically disintegrate. Finally, the close correspondence between our data and the sediment core-based proxies of anthropogenic N noted above indicates that the shell data trends accurately follow levels of environmental N. We argue that the trends in δ 15 N values noted in this study are primarily the result of changes in the sources contributing to the oysters' habitat. The source for the somewhat higher δ 15 N values in the Early Woodland samples is not readily apparent.
This study demonstrates the potential utility of shell δ 15 N proxies. The abundance of this and similar species in archaeological sites worldwide suggests that detailed records of anthropogenic N impacts can be created from shell midden deposits. Such baseline data are valuable not only in addressing modern pollution concerns, but also in assessing the extent of ancient human habitation, land-use, agricultural practice, and related activities. This approach may be less expensive than sediment core analysis and therefore permit wide geographic coverage to facilitate regional N reconstructions. Additionally, archaeological contexts may allow reliable age control of samples, which is sometimes difficult in cores with disturbed sediments. Sclerochronological shell sampling also permits seasonal analysis of δ 15 N values in large, fast growing species 27 , which provides even more detailed information. Further research is needed to better assess the validity of shell %N data, the impacts of diagenesis and prehistoric cooking methods, and the possible transport of shells before burial.

Methods
Oyster shell samples were collected from six different archaeological sites within the Rhode River Estuary of Chesapeake Bay (Fig. 1). Researchers with the Smithsonian Institution collected archaeological samples while excavating shell middens in the region 21,37 . Modern oyster samples were collected from an oyster reef located adjacent to one of the archeological sites and in relatively close proximity to the other archaeological sites.
Ninety archaeological shell samples ranging in age from ~100 to 3,200 years old and thirty modern shell samples were analyzed for %N and δ 15 N values. The ancient shells were all well preserved and robust (i.e., limited taphonomic effects, no noticeable alteration of color of the shell or ligament scar, and non-chalky texture). The soft tissues were removed from the modern shells and the shells were cleaned using DI water and a soft bottlebrush followed by five minutes of ultrasonic cleaning. The shells were dried overnight at room temperature. The left valve of each shell sample was bisected along the longest axis using a slow-speed diamond-wafering saw. Each shell was subsampled using a hand held micro-drill. In order to create a lifetime average of the %N and δ 15 N values, a transect was drilled in cross-section through the resilifer to a depth of approximately 2 mm (Fig. 4). No carbonate powder from the exterior surface of any valve was sampled. Gillikin et al. 27 has shown that the oxidative cleaning required for coral and foraminifera samples is not required for dense bivalve shells. The powdered samples were stored in 4.5 mL round-bottomed borosilicate vials until elemental analyzer -isotope ratio mass spectrometry (EA-IRMS) analysis.
To obtain comparable amounts of nitrogen (i.e., mass spectrometer peak area in millivolts (mV)) between the modern and archaeological shell samples, different masses were used. Between 34-35 mg of powdered samples of modern shells were transferred to 5 × 9 mm tin capsules and 100-110 mg of archeological shells were transferred to 9 × 10 mm tin capsules. The samples were analyzed in the Alabama Stable Isotope Laboratory, in the Department of Geological Sciences at the University of Alabama. Determinations for %N and δ 15 N values were made using a Costech ECS 4010 elemental analyzer (EA) coupled to a Thermo Delta V isotope ratio mass spectrometer without acid pre-treatment of the samples following Versteegh et al. 45 . The EA was fitted with a carbon trap [glass trap (Costech part no. 071121) filled with CO 2 absorbant (Costech part no. 021020)] to remove CO 2 and operated with a combustion temperature of 1020 °C (reduction column = 650 °C). The δ 15 N data are reported in parts per mil (‰) vs. AIR. Delta notation is calculated using the equation δ 15 N = [(R sample − R standard )/R stantard ] * 1000. Precision (1σ) was better than 0.09‰ based on analysis of two standards over a range of isotopic values (ammonium sulfate standard IAEA-N-2; δ 15 N = 20.3‰) and Elemental Microanalysis high organic sediment standard B2151; δ 15 N = 4.4‰). Two elemental standards, acetanilide (%N = 10.36%) and B2151 (%N = 0.5%), were used to build the calibration curve for each run and check for accuracy, respectively. Thirty-four samples were analyzed in duplicate with precision better than 0.07‰. One-way ANOVA and Tukey Post Hoc tests were completed on the shell samples based on time period to determine if there were any statistically significant differences between periods over time.