Strong correspondence between nitrogen isotope composition of foliage and chlorin across a rainfall gradient: implications for paleo-reconstruction of the nitrogen cycle

Nitrogen (N) availability influences patterns of terrestrial productivity and global carbon cycling, imparting strong but poorly resolved feedbacks on Earth’s climate system. Central questions concern the timescale of N cycle response to elevated CO2 concentration in the atmosphere and whether availability of this limiting nutrient increases or decreases with climate change. Nitrogen isotopic composition of bulk plant leaves provides information on large-scale patterns of N availability in the modern environment. Here we examine the utility of chlorins, degradation products of chlorophyll, hypothesized to persist in soil subsequent to plant decay, as proxies for reconstructing past plant δ15N. Specifically, we test the hypothesis that δ15N of plant leaves (δNleaf) is recorded in δ15N of pheophytin a (δNpheo) along the leaf–litter–soil continuum across an array of ecosystem climate conditions and plant functional types (C3, C4, legumes, and woody plants). The δ15N of live foliage and bulk soil display marked declines with increasing rainfall, consistent with past studies in Hawaii and patterns worldwide. We find measurable chlorin concentrations along soil–depth profiles at all sites, with pheophytin a present in amounts required for isotopic analysis (> 10 nmol). δNpheo in leaves, litter, and soil track δNleaf of plant leaves. We find potential for δNpheo records from soil to provide proxy information on δNleaf.


Introduction
A combination of high biological requirements, limited natural supply, and high chemical reactivity and mobility gives nitrogen (N) important controls over photosynthesis and respiration in the terrestrial biosphere (Vitousek and Howarth, 1991).
The response of the terrestrial carbon (C) cycle to climate change and increased atmospheric carbon dioxide (pCO 2 ) will depend largely on the availability of N (Hungate et al., 2003;Wang and Houlton, 2009;Ainsworth and Long, 2005;Denman, 30 Humans have modified the N cycle directly through production of reactive N (Galloway et al., 1995) and likely indirectly through several climate-induced changes receiving ongoing investigation, including changes in biological demand (McLauchlan et al., 2013), changes in organic mineralization (Durán et al., 2016) and mineral weathering rates (Houlton et al., 2018), and changes in N-fixation rates. Experimental manipulations have yielded tremendous insights into mechanisms 5 underlying complex carbon-climate-N cycle dynamics, yet limitations of geography, complexity, and time have left open questions about the overall N biogeochemical response to climate change at the landscape and global scales.
Insight into actual outcomes on real landscapes over decadal-to-millennial timescales would be helpful to predicting how feedbacks between C and N cycles will play role in climate change this century (Luo et al., 2004). If we could observe how N 10 availability to plants responded to past periods of change in climate and pCO 2 , such as during transitions from glacial to interglacial periods, we could improve our projections for future integrative responses of the N cycles to a complex, changing world (Hungate et al., 2003;Houlton et al., 2015). Knowledge of past N cycle behaviour would provide both insight into longterm dynamics and important baselines from which we can better assess anthropogenic impacts.

15
Natural abundance stable isotopes of N (δ 15 N) are useful integrators of N cycle processes in modern environmental systems (Robinson, 2001), and are natural sources of evidence for N cycle behaviour in the past. Due to high reactivity and mobility of N, point-based concentration measurements give temporally limited views of dynamic plant-soil-microbial cycling of N.
δ 15 N of plants and soils reflect a time-integrated signal of the N cycle. The gaseous loss processes (denitrification, ammonia volatilization, and anammox) which fractionate the light and heavy isotopes in terrestrial ecosystems are substrate-dependent, 20 making δ 15 N sensitive to changes in the availability of nitrogen (Houlton and Bai, 2009). Conversely, at low N availability, there is likely to be less isotopic expression of these pathways, as well as potential for greater dependence of plants on ectomycorrhizae, thereby reducing isotopic difference between plant foliage and N sources (Hobbie and Ouimette, 2009). As a result, higher δ 15 N leaf corresponds to higher N availability to plants on average. (e.g., Handley et al. (1999); (Amundson et al., 2003;Houlton et al., 2007;Craine et al., 2009;Martinelli et al., 1999)). Reconstructions of δ 15 N leaf would accordingly 25 provide information on availability of N to plants in the past.
The same reactivity of N that makes δ 15 N leaf a valuable proxy for N availability in modern landscapes makes obtaining a primary N isotope signal from dead, buried, and decomposed organic matter particularly challenging. Primary δ 15 N signatures are altered by decomposition and diagenesis (Thackeray, 1998;Meyers and Ishiwatari, 1993;Hedges and Oades, 1997), and 30 bulk interpretations are confounded by preferential preservation and accumulation of macromolecules with distinct δ 15 N values (Hobbie and Ouimette, 2009;Wedin et al., 1995). Because low-oxygen environments preserve organic matter, sediment accumulations in small lakes have been used as sources of terrestrial δ 15 N leaf (McLauchlan et al., 2013). However, lake ecosystem processes can obscure the original δ 15 N of terrestrial sources (Meyers and Ishiwatari, 1993). Land-based paleo-proxy δ 15 N leaf inferences have targeted bulk N protected structurally, in fossil faunal material , in tree wood (Hietz et al.), and pollen (Descolas-Gros and Scholzel, 2007). These approaches require controversial case-by-case defence of the primary nature of δ 15 N due to diagenetic vulnerability (Harbeck et al., 2004;Thackeray, 1998) andredistribution of N (Gerhart andMcLauchlan, 2014). In the case of faunal material, it is also necessary to consider species-specific trophic enrichment factors, age-effects, and dietary protein content (Sponheimer et al., 2003;Overman and Parrish, 2001). Many of 5 these approaches are further limited for understanding landscape-level patterns by the poor spatial distribution of samples.
Compound-specific isotope analysis (CSIA) of sedimentary material offers advantages over these methods: Isotope ratios of individual compounds can be more resistant to diagenesis than those of bulk materials, and have the further advantages of deriving from more constrainable sources. This last characteristic permits analysis of material from integrative depositional 10 environments. In marine N biogeochemistry, subaqueous sediments that collect and bury organic compounds in time series deposits are widely used for paleoenvironmental reconstruction (Junium et al., 2011;Meyers, 1997). Soils likewise accumulate organic material over time, though not perfectly analogously to subaqueous systems.
The ideal characteristic compound for CSIA reconstruction of δ 15 N leaf in soil is N-rich, produced only above-ground, and 15 retains the primary δ 15 N of plant leaves. We suggest that degradation products of the chlorophyll molecule (chlorins) meet all the above criteria. Chlorins derive from chlorophyll through processes of senescence, decomposition, diagenesis, and grazing (Keely, 2006;Treibs, 1936). They have been successfully extracted from plants, litter, organic soil layers, sediments, and coal deposits (Kennicutt et al., 1992;Sanger, 1971a;Hodgson et al., 1968;Bidigare et al., 1991;Dilcher et al., 1970), but have not previously been sought in mineral soil horizons. 20 δ 15 N of chlorophyll and chlorins derived from algal pigments deposited in sediments have been used to infer N cycling processes in aquatic systems (Higgins et al., 2010;Sachs and Repeta, 1999;Enders et al., 2008;Tyler et al., 2010), both as combined chloropigment fractions and individual pigment compounds (Higgins et al., 2010;Kusch et al., 2010). Several studies have shown that the δ 15 N of plant chlorophyll reflects bulk foliar δ 15 N with a smaller offset than that between algal 25 chlorophyll and bulk algal δ 15 N (Chikaraishi et al., 2005;Bidigare et al., 1991;Kennicutt et al., 1992), but δ 15 N has not previously been measured on chlorins extracted from decomposed plants, terrestrial organic matter, or soil. The behavior of chlorophyll a degradation products and their δ 15 N values along the leaf-soil continuum, and any environmental effects such as climate on isotope offsets, are also unknown.

30
Here we examine chlorin fractions for the presence of individual compounds in sufficient abundance for isotopic analysis.
Pheophytin a (pheo a) is a chlorin previously found in greater relative abundance than any other in organic soils and litter (Sanger, 1971a;Gorham and Sanger, 1967), and is therefore of particular interest. We next examine δ 15 N of abundant chlorins in leaves, litter, and soil and compare values to bulk foliar sources to evaluate potential use of the compound as a proxy for past terrestrial foliar δ 15 N. For a chlorin to be a useful foliar δ 15 N proxy in the terrestrial environment, we hypothesize two conditions: First, it must persist in quantities sufficient for isotopic analysis. Second, it must record the δ 15 N value of the leaves it derived from. To meet this second condition, the chlorin δ 15 N must retain foliar δ 15 N with at most a constrainable isotopic offset, independent of environmental conditions and throughout the processes of biosynthesis, senescence, and decomposition. 5 We evaluate these hypotheses on the natural climate gradient of Kohala Mountain, on the Big Island of Hawaii. These ecosystems are well suited for testing questions about relationships between climate, biogeochemistry, and the preservation or degradation of organic compounds. Relatively low plant diversity and broad environmental niches permit comparison of the bulk and compound-specific δ 15 N of leaves, litter, and soil, of similar grassland communities in sites ranging widely in climate and δ 15 N, to investigate patterns of deviation of compound-specific values from bulk values. 10

Site Description
The four sites of this study are located on Kohala Mountain, the oldest of five volcanoes that make up the Big Island of Hawaii ( Fig. 1). The sites are a subset of a well characterized climosequence running from the crest of Kohala down the leeward side (see , (Kelly et al., 1998b), (Vitousek and Chadwick, 2013), (von Sperber et al., 2017)). Prevailing 15 winds from the northwest create a dramatic gradient in precipitation, from 2500 mm falling annually at the uppermost site to 210 mm at the lowest site (Giambelluca et al., 1986). In contrast to the substantial shift in precipitation, the elevation of the mountain produces only a moderate gradient in air temperature, ranging from 17°C at the upper site to 23°C at the lowest site.
Soils have been forming since emplacement of the Hawi lava flows, approximately 150,000 years ago Spengler and Garcia, 1988;Wolfe and Morris, 1996). This gradient thereby brings into focus precipitation as the most 20 substantial systematic factor of change across sites.
Climate history for this region has been reconstructed from pollen assemblages in cores from bogs near the Kohala summit (Hotchkiss, 1998), and inferred from patterns of sea-level change  Aeolian deposits (Porter, 1997), and soil calcite deposits at the drier sites Porter, 1997). During glacial periods the summit was cooler and 25 drier, while the lower sites show less climate variability, having remained arid throughout soil development.
Vegetation across the gradient was altered by clearing of land for pasture in the last two hundred years (Cuddihy and Stone, 1990;Chadwick et al., 2007), resulting in the introduction of non-native species and grasses with the C4 carboxylation pathway. All of the sites in this study experience grazing by cows, with sites F and I grazed most heavily. Site C is lowland 30 dry scrubland and grassland (150 to 500 mm annual rainfall), dominated by buffel grass (Cenchrus ciliaria) and the leguminous tree keawe (Prosopis pallida). Sites F (790 mm annual rainfall) and I (1260 mm annual rainfall) lie in lowland dry and mesic forest, woodland, and shrubland zones, which through conversion to pasture are dominated by the grass kikuyu (Pennisetum clandestinum). Site M lies in the wet forest and woodland zone (2500 mm annual rainfall), in which ohia (Metrosideros polymorpha) is the dominant tree and the native understory was previously dominated by tree fern (Cibotium spp.) and uluhe (Dicranopteris linearis), but conversion of land to pasture has introduced kikuyu and other species. The herbs clover (Desmodium incanum) and Madagascar ragwort (Senecio madagascariensis) were additionally prevalent at sites I and M. 5

Sample collection and preparation
Leaves were sampled across sites from six plant species; from the upper 1/3 of the canopy in the case of the trees: Two grasses, P. clandestinum and C. ciliaria; two herbs, D. incanum and S. madagascariensis; and two trees, P. pallida and M. polymorpha.
Two of these species are N fixers (D. incanum and P. pallida), and two have a C4 carboxylation pathway (P. clandestinum and C. ciliaria) while the others have C3. Each species was collected wherever present within a radius of ~50 m from the soil 10 pit dug at each site, but not all species were present at all sites. For grasses and herbs, at least three individuals were collected at each site and bulked for processing as a single sample. For trees, leaves were collected from at least 3 branches per tree and bulked for processing. Soils were sampled from pits dug in open, grassy areas with minimal slope. Pits were dug to a depth of greater than 50 cm, 15 with the deepest dug to 65 cm. At these sites, litter and organic horizon layers were sampled as a single genetic horizon termed "litter," consisting of organic materials with fibrous through humic composition and which had a total thickness of less than 2 cm. Mineral horizons were sampled at regular depth intervals of ~10 cm to a depth of 50 cm. Samples were bagged in Whirlpak bags and kept dark and chilled on ice in coolers until processing.

20
Live foliage was rinsed with deionized (DI) water to remove dust or other contaminants and then freeze-dried for preservation. Soils and litter were likewise freeze-dried. Dried samples were ground using either a carbide-steel shatter box (SPEX SamplePrep, University of California, Davis) or a mortar and pestle. Samples were stored in a dark freezer.

Compound extraction and purification 25
Samples were handled under very limited light conditions to minimize photo-degradation. Pigments were extracted by sonication in triplicate using ~100 mL of acetone (300 ml total) for 1 to 80 g of subsamples. Extracts were concentrated by evaporation to near dryness under nitrogen or argon. The condensed acetone extract was partitioned between 10 ml hexane and 30 ml Milli-Q water via liquid-liquid extraction to remove more polar compounds. The hexane fraction was retrieved, and dried under gentle flow of argon gas, and this crude chlorophyll extract was dissolved in dimethylformamide, and passed 30 through a syringe filter (0.4µm) to remove particulates prior to High Pressure Liquid Chromatography (HPLC) chromatography.
Two-dimensional HPLC was used to separate and purify chlorophyll a and b and their photo-reactive degradation products (Agilent 1200 Series with a diode-array detector (DAD), and a fraction collector). Sample fractions corresponding to peaks 5 collected from the first HPLC column run were subsequently run through a second column for additional purification of individual compounds. For the first HPLC separation step using reversed phase, the sample was passed through a Zorbax Eclipse XDB C18 column (9.4 x 250 mm; 5 µm) with a liquid phase consisting of acetonitrile/pyridine (100:0.5) and ethyl acetate/pyridine (100:0.5) in variable ratio, for 35 minutes at 4.2 ml/min and 30°C, after (Kusch et al., 2010). For the first five minutes of the run, acetonitrile/pyridine was 75% of the eluent, and this percentage was linearly reduced to 50% over the 10 subsequent 30 minutes of the run. For all samples, pheophytin a (pheo a) was collected upon elution at ~19 minutes. For five plant samples, chlorophyll a and pyrochlorophyll a were collected upon elution at ~12 minutes. Spectra were checked for purity across wavelengths 200-900 nm. In the second HPLC separation step, also using reverse phase, the sample was passed through an Agilent Zorbax Eclipse PAH column (4.6 x 250 mm; 5 µm), with a liquid phase consisting of acetonitrile/pyridine (100:0.5) and ethyl acetate/pyridine (100:0.5) in variable ratio for 35 minutes at a flow rate of 1 ml/min and 15°C. For the first 15 five minutes of the run, acetonitrile/pyridine was 80% of the eluent; between 5 and 30 minutes this percentage was linearly reduced from 80% to 40%; and for the last 5 minutes the column was flushed with 100% ethyl acetate/pyridine.

Analytical methods
Nitrogen isotopic composition, total N, and total C analysis of bulk soil (including litter) and leaf samples was performed using 20 an elemental analyser interfaced to a continuous-flow isotope ratio mass spectrometer (EA-IRMS) at the University of California, Davis Stable Isotope Facility. The mean value of analytical precisions obtained for standard materials is 0.3‰ for δ 15 N. C/N ratios are reported in mass units. In the case of soil, analysis was performed on ~20 mg of material using an Elementar Vario EL Cube or Micro Cube elemental analyser (Elementar Analysensysteme GmbH, Hanau, Germany) interfaced to a PDZ Europa 20-20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK). In the case of leaves, analysis 25 was performed on ~4 mg of material using a PDZ Europa ANCA-GSL elemental analyser interfaced to a PDZ Europa 20-20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK). Samples are combusted at 1000°C in a reactor packed with copper oxide and lead chromate. Following combustion, oxides were removed in a reduction reactor (reduced copper at 650°C).
The helium carrier then flowed through a water trap (magnesium perchlorate). N 2 and CO 2 were separated using a molecular sieve adsorption trap (for soil) or Carbosieve GC column (65°C, 65 mL/min) (for leaves) before entering the IRMS. Bulk 30 isotopes of two plants (Koh F buffel/kikuyu) and one soil core (4 layers, fresh and litter1) were analysed at the Japan Agency for Marine Earth-Science and Technology (JAMSTEC). δ 15 N, total N, and total C analysis of isolated pheo a fraction was performed on a Nano-EA-IRMS at JAMSTEC. This is a modification of an EA-IRMS system to achieve ultra-sensitive analysis (Ogawa et al., 2010). All samples contained at minimum 10 nmol N, resulting in δ 15 N precision of < ±0.4‰. To investigate the possibility of N contamination, nitrogenous volatile was analysed for the pheo a fraction by gas chromatography/nitrogen-phosphorus detector (GC/NPD) with 5 trimethylsilyl derivatization (BSTFA, Agilent Technologies, Palo Alto, CA) at JAMSTEC. C/N ratios from EA/IRMS were used as the "purity indicator" of each chlorophyll compound. As is described below, if a sample showed the clean spectrum pattern of a chlorin but had significantly large C/N ratio, it is likely that it contained C-containing contaminants, such as carbon hydrates, which are not detectable by DAD.

Purity of isolated compounds
While absorption spectra in UV/visible wavelengths showed no evidence for other absorptive components in the pheo a extracts, C/N weight ratios greater than 11.8 revealed inclusion of non-pheo a carbon in the extracts. C/N ratios were higher in pheo a extracted from litter (31) than plant samples (14.5), and even higher in pheo a samples from soils (113). Lack of relationship between C/N and bulk-pheo δ 15 N offset suggests that impurities do not contribute significantly to measured δ 15 N 15 values, however (Fig. 2). This supposition was confirmed by results from the GC-NPD), which showed a lack of nitrogenous compounds (e.g. containing amino groups) in the pheo a extracts. Although we caution that polar and less-volatile compounds cannot be detected by this method, the extraction methods make amino acids and other nitrogenous contaminants unlikely.
The analytical protocol is designed to remove as many of such contaminating compounds as possible; e.g. amino acids are not soluble in acetone. 20

Isotopic notation
Nitrogen isotopic compositions are reported using conventional delta (δ) notation: where R represents the 15 N/ 14 N ratio and subscripts indicate the sample or isotopic reference. The sample isotopic composition is measured directly relative to the N 2 laboratory reference gas (δS Ref ), and the composition of the sample relative to the internationally recognized δ 15 N reference, AIR, is calculated by: Epsilon notation is used to describe biosynthetic isotopic fractionation within whole leaf (bulk) tissue. Nitrogen isotopic fractionation of pheo a relative to bulk leaf tissue ( 15 ε pheo-leaf ) and chlorophyll a relative to bulk leaf tissue ( 15 ε chl-leaf ) are defined according to Equation 3: 15 +1000 − 1] (Equation 3).

Bulk isotope and C and N concentration data
Site-averaged bulk δ 15 N of all soil samples decreased with increasing precipitation (and elevation) across all sites, from an 10 average of 12.4‰ at site C to 5.1‰ at site M (Table 1). Site-averaged δ 15 N of litter decreased between sites C and M (4.3‰ to 2.8‰), but sites F and I were higher than either of these values (9.0‰ and 7.2‰). Average foliar δ 15 N likewise decreased between sites C and M (4.2‰ to 0.5‰), but site F (9.1‰) was higher than C (Fig. 3). Site-averaged bulk soil δ 15 N values for those samples on which δ 15 N pheo was measured follow similar trends (Table 2), though on average site values are slightly lower (8.4 vs. 8.9‰), reflecting the shallower average depth of the soil samples. 15 Soil %N increased from site C (0.2%) to site M (1.2%), and %C increased across these sites from 1.9% to 17.2%. C/N increased between these sites from 11.5 to 16.3 (Table 1). Litter %C and C/N was notably higher at site M (36. 3% and 25.2) than at the other sites, although litter %N was relatively flat across all sites. Vegetation %C was likewise highest at site M (43.1%), though this value was less exceptional compared with vegetation at other sites. 20 Vegetation had the highest C/N (average of 19.9), followed by litter (17.2) and soil (12.4). At sites C, F, and M, C/N of litter is closer to that of vegetation than to soil, while at site I C/N of litter is closer to that of soil than to vegetation.

Chlorin compound detection
Chlorins were detected in leaf, litter, and soil samples, including in soil mineral horizons up to a depth of 32 cm ( Table 2). As 25 chlorophyll a has greater absorbance at 660 nm than pheo a, direct proportionality between relative HPLC peak areas and relative compound abundance in the sample should not be assumed. In plant leaves, chlorophyll a was the dominant chlorin.
Chlorophyll a was also found in smaller to trace amounts in litter and some soils. In litter, pheo a was the dominant pigment, with the exception of site M, where chlorophyll a was more abundant in the litter sample. In site C, chlorophyll a was absent from litter and soil. In soils, pheo a was the most abundant degradation product. Pheo a, targeted for isotopic analysis due to its superlative abundance, was present in sufficient concentration for isotopic analysis above ~20 cm in soils at most sites.
10 15 ε pheo-leaf was largest at site C (4.0‰) and smallest at site F (0.12‰). For a given species, average 15 ε pheo-leaf tended either to remain flat or slightly decrease into the wettest sites (Fig. 5).
δ 15 N pheo offsets across leaf-litter-soil-δ 15 N of Pheo a in litter was on average 0.3‰ higher than the δ 15 N of pheo a of live foliage at a common site. Average difference between the δ 15 N of bulk litter and foliage were slightly higher, at ~ 2.6‰ at a 15 common site (Table 4). Pheo a-specific soil δ 15 N values were on average 1.3‰ higher than pheo a litter values at a common site; bulk δ 15 N soil values are 2.6‰ higher than bulk litter. The average offset between δ 15 N pheo of soil and live foliage at a site was1.1‰; the average offset between bulk soil and bulk vegetation δ 15 N was 4.9‰.

Soil depth profiles 20
Soil δ 15 N bulk was, on average, higher than the δ 15 N bulk of overlying litter, and there was a slight trend of increasing δ 15 N bulk with increasing depth in the upper ~25 cm of soil pits (Fig. 6). δ 15 N pheo of soil also displayed higher values relative to overlying litter in sites F and I and in part of the profile at site M (Fig. 6). Soil δ 15 N bulk values returned to slightly more negative values deep in the profiles; particularly notable at site F. At site C, δ 15 N pheo along the soil profile do not deviate significantly from that of the overlying litter. At site M, δ 15 N pheo values increased slightly with depth in the upper profile, but then decreased while 25 δ 15 N bulk steadily increased with greater depth. In sum, δ 15 N pheo of soil did not follow δ 15 N bulk , nor did it constantly track δ 15 N pheo of modern plants at a common site.

Evaluation of δ 15 N fractionation within leaves
Nitrogen isotopic offsets between pheophytin a and live foliage ( 15 ε pheo-leaf ) are generally small (mean of 1.4‰) and well constrained (s.d. ± 2.3‰, Fig. 3) across our sites, marking δ 15 N pheo as a useful proxy for bulk δ 15 N leaf . The robustness of 15 ε pheoleaf across the wide range in environmental variables and values of δ 15 N leaf observed across these sites (Fig. 4) suggests that isotope effects from species or physiological changes will be relatively small sources of variation in δ 15 N pheo compared with 5 changes to primary δ 15 N leaf . For example, average community foliar δ 15 N leaf varies by about ~10‰ across terrestrial biomes in the modern environment (Craine et al., 2009), and paleo records have inferred comparable shifts in δ 15 N leaf of 10‰ between glacial-interglacial cycles (Stevens et al., 2008). The possible effect of species shifts, and particularly growth of forests, should be considered when evaluating proxy δ 15 N pheo records, however. One species, P. pallida, has 15 ε pheo-leaf of 6.5‰. The other tree studied here, M. Polymorpha, had the next largest 15 ε pheo-leaf (2.5‰), while herbaceous plant values were all considerably 10 smaller (-0.29 to 1.4‰), raising the possibility that this is a systematic isotope effect in trees, perhaps due to their greater size, longevity, and resulting N storage and redistribution requirements relative to herbaceous plants.

Evaluation of δ 15 Npheo from soil as a proxy for δ 15 Nleaf
Pheo a was present in quantities sufficient for isotopic analysis in litter and the uppermost mineral soil sample at all sites, and 15 down to 30 cm at the wettest site (Fig. 4), inviting exploration of soil δ 15 N pheo as a proxy for δ 15 N leaf . δ 15 N pheo of soil and litter match δ 15 N pheo of overlying leaves at only one studied site (site C). Two alternatives could explain the lack of coherence in leaf, litter, and soil pools across most sites: 1) there is isotopic alteration during senescence and mineralization of chlorophyll, and 2) litter and soil pheo a pools have other sources than the plants we sampled, either on the current landscape, or from previous vegetation covers. 20 To investigate the expression of fractionation on demetallation of chlorophyll in these samples, we can compare the δ 15 N pheo with chlorophyll a from the same sample. Chlorophyll a was only in sufficient abundance for isotopic measurement in live plant leaves. Chlorophyll a is about 0.05‰ depleted in 15 N relative to bulk leaves in this study. This is consistent with previous studies, which found chlorophyll in plants to be 1.2‰ depleted in 15 N relative to bulk leaf tissue (Kennicutt et al., 1992;25 Chikaraishi et al., 2005). Pheo a is accordingly about 2‰ enriched in 15 N relative to chlorophyll a in our leaf samples, which could point to fractionation either within the leaf or in the laboratory (Sachs, 1997). Because chlorophyll abundance is greatly reduced between leaves and litter, any fractionation is likely to be expressed as a difference between leaves and litter, not as a difference between litter and soil or within soil. Fractionation on demetallation from chlorophyll should therefore not account for changes in δ 15 N pheo in the soil. 30 Soil is enriched in 15 N relative to litter at our sites, consistent with observations from a wide range of environments, including temperate rainforest (Menge et al., 2011), temperate deciduous forests (Templer and Dawson, 2004), boreal forests (Hyodo and Wardle, 2009), tropical forests (Martinelli et al., 1999), temperate grasslands (Baisden et al., 2002;Brenner et al., 2001), and elsewhere (Hobbie and Ouimette, 2009). Because post-depositional processing of N involves the preferential loss of 14 N from the soil N pool without replacement, through removal of the products of mineralization and denitrification, bulk soil δ 15 N 5 is highly dynamic and will tend to increase along a decomposition gradient. In contrast, if there is no fractionation on breakdown of pheo a, δ 15 N pheo in soil should not depart from the δ 15 N pheo of litter inputs.
In soil profiles across the climosequence, δ 15 N bulk and δ 15 N pheo do not exhibit similar depth patterns (Fig. 6). This suggests that different processes govern these two records in soil. While δ 15 N bulk values drift increasingly positive across the leaf-litter-soil 10 continuum, pheo a values remain centered around site average leaf values, and sometimes show slight decreases in value from leaf to litter and soil (Table 4). This suggests that mixing of inputs may account for litter and soil pheo a δ 15 N values, while fractionating losses from the bulk soil N pool are required to explain δ 15 N bulk values.
δ 15 N pheo of litter is a window into possible isotopic effects of decomposition on δ 15 N pheo of plant leaves. The difference between 15 δ 15 N pheo of leaves and litter averaged close to zero and was substantially smaller than the difference in δ 15 N bulk (0.3‰ versus 2.3‰), supporting greater isotopic fidelity along the decay continuum in pheo a δ 15 N than bulk δ 15 N. Some of the site variability in isotopic offsets between litter and leaves/soil is likely due to differences in how well decomposed the collected litter was from site to site: at sites C, F, and M, C/N of litter is closer to that of vegetation than to soil, while at site I C/N of litter is closer to that of soil than to vegetation. 20 It is likely that litter samples do not reflect equal contributions from the plants sampled at the same site. Litter was collected from the surface of the soil pits, while vegetation was sampled from a radius of many meters, and plant taxa were not identified in litter samples to allow for source attribution. Soil algae and bacteria are considered insignificant contributors to soil pheo a, as the absorbance spectra of soil extracts showed that chlorophyll degradation products are dominated by higher plant 25 contributions.
Pheophytin has two pathways of generation: Like chlorophyll a, it is biosynthesized in plant leaves from glutamate where it serves as an electron acceptor in photosystem II (Klimov et al., 1977). Additionally, it is a product of chlorophyll degradation, whether via senescence, grazing, or decomposition across the leaf-soil continuum (Kräutler and Hörtensteiner, 2014;Sanger, 30 1971b), in which the central Mg is replaced by two H atoms. Transformation from chlorophyll to pheo a involves breaking N-Mg bonds, and so has potential for fractionation of N isotopes. However, the N-Mg bond is not a normal covalent bond, but a bond loosely connecting a ligand and metal in complex. Because the energy needed to break this bond is substantially smaller than the covalent bond, Mg loss from chlorophyll would theoretically be expected to have little or no N isotope fractionation.
Senescence is unlikely to impart significant fractionation from bulk leaves given the observation that bulk leaf δ 15 N does not change on abscission (Kolb and Evans, 2002) and the understanding that N contained in the tetrapyrrole structure of chlorophyll is not recycled by the plant (Eckhardt et al., 2004). However, fractionation on demetallation of chlorophyll to pheo a has been observed in a laboratory setting with an effect of up to 2‰ (Sachs, 1997). Mineralization of pheo a is unlikely to alter δ 15 N of the remaining pheo a pool because known products of chlorin defunctionalization retain the four atoms of tetrapyrrole N and 5 their associated bonds (Keely, 2006).
It is unlikely that environmental changes that took place between production of litter and growth of current leaves explain the observed differences between δ 15 N of plants and litter, given that litter at these sites is estimated to be no more than a few years old. Soil pools, however, could contain pheo a that is considerably older. There is reason to expect that deeper pheo a 10 will be older than overlying pheo a in soil profiles, and increasing recognition that compounds can be preserved in the soil matrix much longer than their inherent lability would predict (Mikutta et al., 2006;Torn et al., 1997;Marin-Spiotta et al., 2011;Kramer et al., 2012).
While we do not know the age of the sampled soil pheo a, radiocarbon dates of 4130 and 8030 yBP were measured on soil 15 organic carbon (SOC) deep within the soil profiles at sites A and D, respectively, along the climosequence (Chadwick et al., 2007), suggesting that our studied soils may reflect organic contributions from several thousands of years of soil development.
Bulk soil carbon isotope values suggest that C4 pasture grasses, introduced over the last two hundred years, have been incorporated in the SOC in the top 40 cm of soil on these sites (Kelly et al., 1998a).

20
Patterns of radiocarbon (Δ 14 C) depletion of SOC in profile (Baisden et al., 2002;Townsend et al., 1995;Tonneijck et al., 2006;Trumbore, 2009) indicate that SOC age increases with soil depth. Departures from this trend have been explained by the introduction of modern material by plant roots (Baisden et al., 2002), but as chlorophyll a and pheo a are photosynthetic pigments, we expect that inputs to the soil originate exclusively as litter accumulated at the soil surface, and for age of the chlorophyll derivatives therefore to increase with depth. Bioturbation and episodic leaching could be disruptive of space-for-25 time trends, but based on the pattern observed in bulk SOC Δ 14 C, are unlikely to interfere with millennial scale patterns given appropriate vertical sampling distances. In fact, in aggrading profiles, bioturbation has been shown to contribute to the increase in SOC age with depth, by transporting SOC downward over short distances and migrating upwards as soil accumulates overhead (Tonneijck and Jongmans, 2008).

30
If down-profile δ 15 N pheo values did represent prior landscape δ 15 N leaf values, these data suggest greater changes to the N cycle at sites F and I than C and M, which could reflect heavier grazing at sites F and I. Climate has been more constant throughout the Holocene at lower elevations than higher elevations on Kohala (Hotchkiss, 1998), which could further account for the relatively constant δ 15 N pheo values at site C.
If the total soil δ 15 N pool reflects fractionation on loss of N, but the soil δ 15 N pheo pool does not, we expect that the more important gaseous losses are relative to leaching losses at a site, the higher δ 15 N bulk and the greater the offset between δ 15 N pheo and δ 15 N bulk will be. Environments where denitrification is more important relative to leaching tend to be dry environments, due to much smaller isotope effects of non-gaseous losses (Houlton and Bai, 2009), and drier sites do indeed have higher δ 15 N 5 in this study. Differences between δ 15 N pheo and δ 15 N bulk do not show clear trends along the climosequence, however. This could be due to the complexity of soil moisture at these sites, reflected in the variability in down-profile soil bulk δ 15 N.
Direct testing of the hypothesis of increasing age with depth of pheo a would be enabled by improving the purification method of pheo a from soil. Removing contaminating C would make stable or radiocarbon of pheo a available tools to shed light on 10 sources of pheo a in soil profiles, age of the compounds, and with what resolution change in N cycling over the temporal window provided by the chlorins depth profile is observable (Ishikawa et al., 2015). These questions could also be averted by measuring δ 15 N pheo in soil profiles with constrained ages, such as buried horizons.

Implications of the δ 15 Npheo proxy
We suggest two key opportunities for application of a soil-based δ 15 N pheo proxy data to advance understanding of past terrestrial 15 N dynamics. First, the ability to track changes in foliar δ 15 N over time implies comparatively direct insight into factors affecting δ 15 N of plants, notably the availability of nitrogen. Central questions concerning the timescale of N cycle responses to elevated CO 2 concentration in the atmosphere, and whether availability of this limiting nutrient increases or decreases with climate change, could be explored by selecting a time series that covers changes in atmospheric pCO 2 (Goulden, 2016).

20
Second, comparison of compound-specific δ 15 N pheo with other δ 15 N proxies over the same time domain could provide insights into processes that cause them to deviate from one another. Alternative sources of δ 15 N proxies include subaqueous sediment deposits in lakes, ungulate tooth enamel, and bulk wood, black carbon, or soil. Deviations in records of δ 15 N of pheo and tooth enamel at a common site would highlight changes in factors affecting dietary isotope fractionations, such as animal growth rates. δ 15 N pheo records could provide information on terrestrial sources of δ 15 N relevant to aquatic sediment δ 15 N records, and 25 allow aquatic and terrestrial signals to be distinguished. δ 15 N pheo could validate bulk proxy records, or highlight diagenetic limitations of the record. Within soil horizons, comparing δ 15 N pheo with δ 15 N bulk could provide information both on N availability to plants and dominant pathways of N loss, hydrologic or gaseous, at a site, allowing for comparison of multiple N cycle dynamics over time.

Conclusions
δ 15 N pheo in leaves provides a molecular recorder of foliar δ 15 N, and provides a means to trace leaf nitrogen signatures in litter and soils. The compound is found in litter and upper soil horizons in abundance sufficient for isotopic analysis, where it shows greater fidelity to leaf δ 15 N than does bulk material. δ 15 N pheo of soil may reveal temporal changes in nitrogen cycle behavioure.g. the availability of nitrogen to plants, and whether nitrogen losses from the ecosystem had a dominant atmospheric or 5 hydrologic fate. Due to the well-constrained photoautotrophic sources of chlorins, their lack of confounding heterotrophic enrichment, and their minimum N isotope fractionation during burial processes, CSIA of chlorins offers several advantages over bulk isotope analysis on the study of soil C and N cycles. δ 15 N pheo of soil is a most promising proxy for δ 15 N leaf where organic matter inputs are high and profiles aggrade. The potential to investigate decadal-to-millennial scale N cycle dynamics will depend largely on conditions for preservation of soil organic matter. 10

Data availability
The data can be accessed by email request to the corresponding authors and have been submitted to the open access database PANGAEA (Data submission 2019-08-01T06:12:00Z https://issues.pangaea.de/browse/PDI-21268).

Author Contributions
SG and BH acquired funding for the project. SG, BH, and OC collected samples. NO, KF, and BH contributed laboratory 15 space and materials. SG, NO, KF, YC, NO, and HS collaborated on laboratory methodology; SG, NO, HS, and YC conducted analyses of samples. SG performed data analysis, generated figures, and wrote the initial draft; BH, NO, KF, YC, HS, and OC contributed to the manuscript. Tonneijck, F. H., and Jongmans, A. G.: The influence of bioturbation on the vertical distribution of soil organic matter in volcanic ash soils: a case study in northern Ecuador, European Journal of Soil Science, 59, 1063Science, 59, -1075Science, 59, , 10.1111Science, 59, /j.1365Science, 59, -2389Science, 59, .2008Science, 59, .01061.x, 2008 Torn, M. S., Trumbore, S. E., Chadwick, O. A., Vitousek, P. M., and Hendricks, D.  this study (C, F, I, and M) are circled in orange. Vegetation zones are highlighted in the inset map: lowland dry scrubland and grassland between 150 and 500 mm of rainfall (yellow); lowland dry and mesic forest, woodland and shrubland from 500 mm to 2000 mm (green); and wet forest and woodland above 2000 mm (blue).
Figure 2: C/N ratios for pheo a isolates higher than 11 (dashed line) reveal contamination by carbon compounds. Effect of this contamination on pheo a δ 15 N is investigated by comparing "pheo a" molar C/N with the offset between pheo a and bulk δ 15 N for litter (black), soil (red), and vegetation (green) samples. Lack of a systematic relationship between high C/N ratios and greater similarity of pheo a δ 15 N to bulk values is evidence against contamination of the pheo a samples by N from the greater bulk sample. 5 Figure 3: Patterns in isotopes of bulk material (open circles) and pheo a extracts (closed triangles) from plant leaves (Veg, green), litter (black), and soil (red). Deep soils (blue) have a depth of greater than 30 cm and only one pheo a isotope measurement was made for these samples. Lines connect site averages between bulk (solid) and pheo a (dashed) measurements.