Evidence of Enriched, Hadean Mantle Reservoir from 4.2-4.0 Ga zircon xenocrysts from Paleoarchean TTGs of the Singhbhum Craton, Eastern India

Sensitive High-Resolution Ion Microprobe (SHRIMP) U-Pb analyses of zircons from Paleoarchean (~3.4 Ga) tonalite-gneiss called the Older Metamorphic Tonalitic Gneiss (OMTG) from the Champua area of the Singhbhum Craton, India, reveal 4.24-4.03 Ga xenocrystic zircons, suggesting that the OMTG records the hitherto unknown oldest precursor of Hadean age reported in India. Hf isotopic analyses of the Hadean xenocrysts yield unradiogenic 176Hf/177Hfinitial compositions (0.27995 ± 0.0009 to 0.28001 ± 0.0007; ɛHf[t] = −2.5 to −5.2) indicating that an enriched reservoir existed during Hadean eon in the Singhbhum cratonic mantle. Time integrated ɛHf[t] compositional array of the Hadean xenocrysts indicates a mafic protolith with 176Lu/177Hf ratio of ∼0.019 that was reworked during ∼4.2-4.0 Ga. This also suggests that separation of such an enriched reservoir from chondritic mantle took place at 4.5 ± 0.19 Ga. However, more radiogenic yet subchondritic compositions of ∼3.67 Ga (average 176Hf/177Hfinitial 0.28024 ± 0.00007) and ~3.4 Ga zircons (average 176Hf/177Hfinitial = 0.28053 ± 0.00003) from the same OMTG samples and two other Paleoarchean TTGs dated at ~3.4 Ga and ~3.3 Ga (average 176Hf/177Hfinitial is 0.28057 ± 0.00008 and 0.28060 ± 0.00003), respectively, corroborate that the enriched Hadean reservoir subsequently underwent mixing with mantle-derived juvenile magma during the Eo-Paleoarchean.

the Paleoarchean (~3.4 Ga) Tonalite-Trondhjemite-Granodiorite gneisses (TTG), called the Older Metamorphic Tonalitic Gneiss (OMTG), of the Paleo-Mesoarchean Singhbhum Craton of Eastern India and confirm that the OMTG holds the hitherto oldest precursor rock recorded in India. We also present Lu-Hf isotopic data from these xenocryst cores and their host zircons that add new information to the Hadean zircon isotopic data repository, augmented with interpretations about the nature of their mantle source and the history of crustal formation events in this craton. In addition, the combined U-Pb and Lu-Hf isotopic data of zircons from two other Paleoarchean TTGs (∼3.4 Ga and ∼3.3 Ga) from different locations within the same craton are also presented to further elucidate the characteristics and heterogeneity of the corresponding mantle source reservoir during the Paleoarchean.

Geology of Paleoarchean TTGs, Singhbhum Craton
The Paleoarchean Singhbhum Craton, India, consists of an Archean nucleus of voluminous TTG gneisses and intrusive granitoids of ∼3.5-3.2 Ga age, flanked by three Paleoarchean greenstone successions, which are named the Iron Ore Group (IOGs) [26][27][28] (Fig. 1). The Archean nucleus of this craton is unconformably overlain by Paleoproterozoic supracrustals [27][28][29] . The Older Metamorphic Group (OMG) comprises interlayered metabasalt (amphibolite) and metasedimentary rocks (biotite-muscovite ± sillimanite ± garnet schists, quartz ± magnetite ± cummingtonite schist; quartz-sericite schist, quartzites and calc-silicates) 26,27,30,31 . The oldest age of the OMG is constrained by a 207 Pb/ 206 Pb ion microprobe age of ∼3.5 Ga, obtained from detrital zircon from quartzites from the Champua area 30,31 . However, the presence of even older inherited cores of ∼3. 55-3.6 Ga within these zircons has led authors to suggest that older crust, with a minimum age of ∼3. 55-3.6 Ga, existed in the Singhbhum Craton 30,31 . The Sm-Nd isochron age of 3.3 Ga derived from the OMG amphibolites represents their metamorphic age 32 . The Older Metamorphic Tonalitic Gneiss (OMTG) consists of thinly compositionally layered, medium-grained tonalitic to granodioritic gneisses 26,28,33 . According to Hofmann and Mazumder 28 , the OMTG represents a suite of TTGs that formed over an extended period between 3.53-3. 45 Ga, whereas the OMG represents a supracrustal assemblage that formed as a greenstone succession. The oldest age obtained from the OMTG is a whole-rock Sm-Nd isochron age of 3775 ± 89 Ma 34 . This age was later questioned and subsequently amended by Moorbath et al. 35 to be closer to 3.4 Ga. However, other older ages recently reported from the OMTG include an age of 3664 ± 79 Ma, which was derived from a whole-rock Pb-Pb isochron 36 , and a xenocrystic zircon core age of ~3.61 Ga ( 207 Pb/ 206 Pb in situ LA-ICP MS dating), which was found within a ~3.4 Ga zircon 37 . Acharyya et al. 38 reported a discordia upper intercept U-Pb zircon age of 3527 ± 17 Ma for the OMTG. Interestingly, the largest population of 207 Pb/ 206 Pb zircon ages of the OMTG from previous studies centered around ~3.4 Ga 33,37-39 , reflecting a major felsic magmatic event. However, the ∼3.6 Ga xenocrysts from the OMTG 37 and OMG quartzites 30 indicate that felsic crustal formation was initiated in the Singhbhum Craton well before the major phase of emplacement of the OMTG.
A voluminous TTG (previously named as Singhbhum Granite-I or SG-I 26 )-granitoid suite (SG-II and III 26 ), was emplaced in two phases; the older emplacement age (3.45-3.44 Ga 37 ) broadly coincides with the emplacement age of the OMTG. The latter phase of emplacement is constrained at approximately 3.35-3.32 Ga 37 . The SG batholith is composite in nature and comprises biotite-granodiorite/granite, adamellite-granite, tonalite and trondhjemite 26,27 . The SG batholith is encircled by three distinct Archean greenstone successions, namely, the eastern, western and southern Iron Ore Group 40-42 (IOG; Fig. 1). The SHRIMP U-Pb zircon ages of 3507 ± 2 Ma of dacitic lava from the southern IOG 42 and ~3400 Ma from a tuff layer in the western IOG 41 confirm the Paleoarchean ages of these greenstone successions. Nelson et al. 33 speculated that the eastern IOG formed between 3.28-3.33 Ga, although its depositional age is still currently unknown.

Results
Description of zircons and U-Pb data. The 207 Pb/ 206 Pb age data of 85 points from 23 zircon grains were obtained from four samples of the Older Metamorphic Tonalitic Gneiss (OMTG; RM-1 & 5), Singhbhum Granite Phase-I (SG-I; sample TRBG-1) and Singhbhum Granite Phase-III (SG-III; sample TRG-2); these data are presented in Table 1 and Supplementary Table S1. Two zircon grains from RM-1 (grain #3) and RM-5 (grain #11) exhibit significantly older ages (Hadean) than the rest of the analyzed grains (Eoarchean to Paleoarchean) in this study. Grain #3 is subhedral and displays oscillatory zoning in the cathodoluminescence (CL) image ( Fig. 2A); it yields three analyses with 207 Pb-206 Pb concordant ages of 4031 ± 5, 4036 ± 15, and 4057 ± 8 Ma (Table 1;  Supplementary Table S1). A second, relatively smaller, subhedral grain (grain #11) shows a homogenous core in its CL image and yields two concordant ages of 4241 ± 4 and 4239 ± 4 Ma (Table 1), while its rim shows thin oscillatory zoning (Fig. 2D) and yields discordant ages of ~3.8-3.9 Ga (Supplementary Table S1). The differences in age and Hf isotopic compositions between these two grains and the rest of the zircon population, combined with their subhedral grain shapes, indicate that these Hadean zircons are inherited in origin 43 . The oscillatory zoning and higher Th/U ratios (0.44-0.65) of the Hadean zircons suggest an igneous origin 44 , although exceptions can occur 45 . Three analyses from the oscillatory zoned rim of an old xenocrystic core with an age of 4241 ± 4 Ga (grain #11) from sample RM-5 yield >10% discordance, thus reflecting Pb loss, which implies hat this rim is probably older than ~3.8-3.9 Ga. In sample RM-5, grain #2 exhibits an inherited core with a concordant age of 3670 ± 7 Ma, which is homogenous in its CL image (Fig. 2C) and is surrounded by an oscillatory zoned growth rim. Another xenocryst from the same sample (grain #9), which has resorbed grain boundaries and broad, faint zoning visible in CL image, yields a concordant age of 3673 ± 7 Ma (Fig. 2C,F). Another older age spot in zircon from sample RM-5 (19.1) yields a concordant age of 3595 ± 12 Ma. The U-Pb analyses of the RM-1 and RM-5 zircons yield 207 Pb/ 206 Pb age data that define linear arrays, yielding concordia intercepts at ages of 3393 ± 9 Ma (MSWD = 1.7; n = 3) and 3399 ± 6 Ma (MSWD = 1.6; n = 6), respectively (Fig. 2E,F). Most of the dated zircons exhibit regular oscillatory zoning from core to rim ( Fig. 2A-D). Some grains exhibit homogenous cores surrounded by growth-zoned rims ( Fig. 2B; grain #23; sample RM-1) but yield a consistent age of ∼3.4 Ga (spot 23.1; Table-1). The lower intercept ages of RM-1 and RM-5 are ∼900 and ∼1200 Ma, respectively which broadly coincide with a ∼1.2-1.0 Ga magmatic event related to the late phase of regional dyke swarm emplacement known as the 'Newer Dolerite Dykes' 26 . Zircons from samples RM-1 and 5 are euhedral to subhedral, and the presence of irregular boundaries in some grains can be attributed to solid-state recrystallization 46  (#11) is free of inclusions. These inclusions are not confined within cracks or fissures and are therefore likely primary 43 although exceptions occur 47 .
Zircons from the granite gneiss (TRBG-1) of the Singhbhum Granite Phase-I (SBG-I) and another granite (TRG-2), collected from the Singhbhum Granite Phase-III (SBG-III), identified from the regional geological map after Saha 26 (Table 2). Zircons from TRG-2 record a U-Pb upper intercept age of 3286 ± 6 Ma (MSWD = 0.57; n = 5; Fig. 3F). Zircon grains of sample TRG-2 contain older cores with concordant ages of 3377 ± 11 Ma (grain #1; Fig. 3C) and 3367 ± 7 Ma (grain #10), which are contemporaneous with those of the RM and TRBG samples (Tables 1 and 2). The 3377 ± 11 Ma core in grain #1 (spot 1.1) from TRG-2 is identified as a xenocryst, as it contains markedly lower U concentrations than its rim 43 (Supplementary  Table S1), and the core appears to be much brighter than the rest of the grain in the CL image (Fig. 3C).  Table 1 and presented in full in Supplementary  Table S2. The Hf isotopic analysis of one spot obtained from the oldest Hadean xenocryst (4241 ± 7 Ma; spot 5-11-1) yields a subchondritic 48 . 4), the pre-4 Ga xenocrysts of the OMTG follow an array with a slope of 0.0103, corresponding to a source 176 Lu/ 177 Hf ratio of 0.019 (calculated after Amelin et al. 19 ), which intersects the chondritic uniform reservoir (CHUR) line at 4.497 ± 0.19 Ga (Fig. 4). The source Lu/Hf ratio calculated from the Hadean zircons, although slightly lower, is consistent with the source being typical mafic crust; that ranges from 0.22 19 to 0.20 21 and is far higher than that of the average TTG crust (0.01) calculated from the oldest Jack Hill zircons 20 . The intersection age (4.497 ± 0.19 Ga) of this array with the CHUR reference line is closer to the CHUR extraction age of 4.46 ± 0.12 Ga as the source reservoir of the Jack Hill zircons 21 .

Hf isotopic compositions of OMTG and SG zircons.
The Eoarchean domains in sample RM-5 with concordant ages of 3673 ± 7 Ma and 3670 ± 7 Ma, yield εHf[t] values of −4.7 ± 1.4 and −6.6 ± 1.1, respectively. Their initial Hf compositions at 3.67 Ga are near identical, e.g., 0.28027 ± 0.00008 and 0.28022 ± 0.00006 respectively, and they are notably higher than those of the Hadean xenocrysts (0.279947-0.280045). However, the oldest Paleoarchean xenocryst, which has a concordant age of 3595 ± 12 Ma, yields an εHf[t] value that is closer to a chondritic value (−1.5 ± 1.3) and a 176 Hf/ 177 Hf initial value (0.28041 ± 0.00007) that is higher than those of the older Hadean and Eoarchean age spots. The ~3.3-3.4 Ga age group of zircons from samples RM-1 and RM-5 yields more radiogenic εHf[t] values than Hadean and Eoarchean age spots, ranging from −0.4 ± 1.2 to −3.7 ± 1.8, except for two data points that fall below −4 epsilon units (−4.9 ± 1.5 and −4.3 ± 1.3). Initial Hf ratios of these spots are identical to those of other age spots with lower εHf[t] values, implying that they were derived from the same source. Initial 176 Hf/ 177 Hf values of the ∼3.3-3.4 Ga zircons display a relatively small range of values, varying between 0.28047-0.28057, identical with the average of 0.28053 ± 0.00003. This range also includes 176 Hf/ 177 Hf initial values of four discordant age spots (10-15% discordance; Table 1), indicating that despite having undergone U-Pb resetting, their Hf isotopic ratios remain unchanged. The ~3.3-3.4 Ga zircons do not exhibit any particular trend in εHf[t]-time space, but they cluster within the field delimited by the 176 Lu/ 177 Hf = 0.019 array defined by the Hadean zircon data and the CHUR reference line (εHf = 0; Fig. 4). This indicates, unlike Zack Hill zircons, the younger zircons of OMTG are not derived from the same source as the oldest crust.

Discussion
Interestingly, the subchondritic Hf composition (ɛHf[t] < 0) of the oldest (4241 ± 4 Ma) xenocryst indicates the presence of a non-chondritic mantle reservoir as early as ∼4.2 Ga. The composition of the earliest mantle reservoirs of Earth has remained controversial. The reported initial 176 Hf/ 177 Hf ratios of the Bulk Silicate Earth (BSE), i.e., 0.279685 ± 19 49 or 0.279781 ± 18 22 , which are lower than that of the chondritic reservoir, argue against the decades-old paradigm of the chondritic Earth and are explained by the accelerated decay of 176 Lu 50,51 . However, such accelerated decay is caused by the high rate of irradiation of chondritic or eucritic meteorites by γ and/or galactic cosmic rays, a process whose effectiveness has been questioned for the BSE 52 due to the restricted penetration depth of these rays 51,53 . Alternatively, it has been assumed that the Earth was developed from chondritic material but was subsequently modified by either collisional erosion during accretion 54,55 or explosive basaltic volcanism in planetesimals 56 . Hence, we assume that the source reservoir of the Hadean OMTG xenocrysts was initially separated from chondritic material, and we interpret our zircon data considering CHUR 48 as a reference frame. The source array of Lu/Hf = 0.019 fitted through these xenocrysts is comparable to 'mafic protocrust' with 176 Lu/ 177 Hf values proposed by Kemp et al. 21 (0.020) and Amelin et al. 19 (0.022) calculated from Jack Hill detrital zircon data. The age of separation (∼4.5 Ga) of the enriched reservoir from the chondritic reservoir calculated from the Hadean zircons in this study is also very similar to the CHUR separation age of the source reservoir of the Jack Hill zircons (∼4.5 Ga 24 or 4.46 ± 0.12 Ga 21 ). The development of the ~4.49 Ga enriched reservoir recorded in the Singhbhum craton is also in near-agreement with the estimated age of ~4.5 Ga for the separation of the enriched silicate reservoir upon Earth's solidification, based on recent geodynamic modeling [57][58][59] .
Assuming that the parental magma of these zircons is likely to be felsic due to the high solubility of zirconium in mafic-ultramafic magmas 60 , the Hadean (~4.2-4 Ga) zircons of the OMTG were presumably generated from minor silicic melts produced as a consequence of the differentiation or re-melting of pre-4.2 Ga juvenile protocrust of mafic composition. The formation and reworking of juvenile crust were either contemporaneous or separated by a short period of ~100-300 My during the Hadean and Archean eons 49 . To explain the nature of the enriched mantle reservoir that parented the Hadean OMTG zircons, we envisage that such a reservoir may represent an enriched, residual mafic magma generated from a partially solidified magma ocean, analogous to KREEP beneath the lunar anorthositic crust 20,21,[61][62][63] . This mafic protocrust was presumably reworked and re-melted to generate felsic melt between ~4.2-4.0 Ga without the significant addition of juvenile material from the mantle. The mineral inclusions in the Hadean OMTG xenocrysts, including K-feldspar, titanite and apatite (Supplementary Figure SF3C), were likely generated from a differentiated melt. The existence of Hadean mafic protocrust has previously been estimated based on Hadean to Paleoarchean Jack Hill's zircons [19][20][21] , ~3.7 Ga metasediments from Isua 64  Thus, it is necessary to determine the fate of this ancient, enriched Hadean reservoir in the Singhbhum Craton. The Eoarchean (~3.6 Ga) zircon age domains in sample RM-5 record the second-oldest stage of felsic melt generation; these are slightly more radiogenic than the Hadean ones. Therefore, they were probably generated from the modification of the enriched source of Hadean zircons due to its interactions with juvenile (more radiogenic) mantle melt, as is evidenced by the fact that the ɛHf[t] and 176 Hf/ 177 Hf initial compositions of these sites (Table 1) are higher than those of the Hadean ones. This also invokes the assumption that the composition of the enriched, subchondritic mantle reservoir in the Singhbhum Craton persisted without undergoing modification until the Eoarchean (~3.7 Ga). However, the identifiable vertical excursion of εHf[t] values in the time-integrated εHf[t] plot (Fig. 4)  (SG-III) are 0.28053 ± 0.00006, 0.28057 ± 0.00008 and 0.28060 ± 0.00003, respectively. These values are closely comparable except for one spot with an age of 3404 Ma with a slightly less radiogenic 176 Hf/ 177 Hf initial ratio of 0.28041. Clearly, these Paleoarchean zircons were derived from felsic melts with near identical Hf isotopic values, while minor disparity is most likely due to incomplete mixing between the enriched reservoir with depleted juvenile magma. The results of a previous petrological modeling study 67 suggested that the protolith of the OMTG was generated by the 40% partial melting of the OMG amphibolites at garnet stability depths 32 . It is unlikely that the the remnants of the earliest Hadean mafic protocrust survived the constant reworking processes until today. Remnants of the oldest mafic protocrust may have been preserved in the amphibolite enclaves within the OMTG or these enclaves could represent a modified mafic component developed from interactions between the ancient enriched reservoir and mantle-derived, juvenile mafic magma and was preserved as melting residuum of the mafic protolith from which OMTG magma was generated. Interestingly, the zircons from the ~3.5 to ~3.3 Ga TTGs of the Singhbhum Craton, which are located near Keonjhar, exhibit suprachondritic Hf isotopic signatures with average ɛHf[t] values ranging from +2.9 to +2.2 39 , suggesting that they were derived from a depleted source reservoir. However, this implies that a separate depleted reservoir, which was probably complementary with the Scientific REPoRTS | (2018) 8:7069 | DOI:10.1038/s41598-018-25494-6 ancient, enriched reservoir hypothesized in the present study, of Paleoarchean (~3.5 Ga) or even older age, existed under the cratonic lithosphere of the Singhbhum craton and also participated in the generation of TTG magma.
Based on isotopic constraints, it has been suggested that Earth's accretion was roughly complete 30 Myr after 68 80 Figure SF3C), which suggests that this ∼3.38 Ga xenocryst may have been inherited from SG-I, as it was reworked during the emplacement of the younger SG-III. The oldest concordant age of sample TRG-2 (∼3.29 Ga; grain #20, 23) is equivalent to the 3289 ± 10 Ma age spot (spot 13.1), which indicates that probably this age of the tectonomagmatic event that led to reworking of pre-existing SG-I and emplacement of SG-III.
Before the emergence of dominantly TTG crust, mafic protocrust likely prevailed as a thin, buoyant tectonic plate 83 . It is still unclear whether such proto-plates were stagnant, as the heat production of the Earth's mantle was more than three times greater during the Archean 84 , which led to more rapid mantle convection, thus triggering faster plate movement. Rapid plate movement invokes the possibility of the quick recycling of thin Hadean protocrust, thus preventing its preservation 85 . Hence, it is possible that the Hadean mafic protocrust in the Singhbhum craton may not have survived long and was recycled and assimilated into more voluminous TTG magma that was generated from a combined process involving the reworking of older, enriched crust and the serial addition of mantle-derived melt during ~3.4-3.3 Ga. The tectonic processes involved in the partial melting of the OMG amphibolites to generate the parent magma of the OMTG are still unclear. However, the geochemical data of the ~3.5-3.3 Ga TTGs of the Singhbhum craton suggest that these TTGs lack the signatures of subduction-derived magma; they are thus considered to have been generated from the reworking of pre-existing mafic crust by the repeated underplating of plume-derived mafic-ultramafic magma 39 during the Paleoarchean 81 .

Conclusions
The combined U-Pb SHRIMP and Lu-Hf isotopic data of the ~4.24 and ~4.03 Ga xenocrystic zircons from the ~3.4 Ga TTG of the 'Older Metamorphic Tonalitic Gneiss (OMTG)' of the Archean Singhbhum Craton of Eastern India contain records of the oldest crust in India. The essentially subchondritic (ɛHf[t] < 0) isotopic signatures of these Hadean zircons indicate that they originated from the reworking of older crust prior to ~4.2 Ga. The calculated 176 Lu/ 177 Hf ratio (0.019) of their source reservoir indicates the mafic nature of the older crust that originated from an enriched reservoir that separated from the chondritic reservoir at ~4.5 Ga. However, the younger and almost contemporaneous zircons from the OMTG (3.3-3.4 Ga), Singhbhum Granite Phase-I (SG-I; ~3.4 Ga) and Singhbhum Granite-III (SG-III; ~3.3 Ga) yield more radiogenic Hf isotopic signatures, indicating that this enriched reservoir persisted but underwent mixing with juvenile mantle material during the Eo-Paleoarchean.

Methods
Zircon separation, CL imaging, inclusion analysis and SHRIMP U-Pb dating were carried out at the Beijing SHRIMP Center, Institute of Geology, Chinese Academy of Geological Sciences, using a SHRIMP II following the analytical procedures described by Williams 86 . Zircon crystals were obtained using standard crushing and grinding techniques, followed by separation using heavy liquid and magnetic techniques. The hand-picked crystals were cast in epoxy resin discs and polished. The intensity of the primary O 2− ion beam was 5 nA and the spot size was 25-30 μm; each site was rastered for 150 s prior to analysis. Five scans through the mass stations were made for each age determination. The standard used for the calibration of elemental abundances was M257, which contains U = 840 ppm 87 . TEMORA, whose 206 Pb/ 238 U age is 417 Ma 88  were analyzed using the same scanning electron microscope with OXFORD IE250. The data were processed and assessed using the Squid 1.02 89 and Isoplot 3.00 90 programs. Common Pb corrections were based on the measured 204 Pb contents. The errors given in Table 1 and the concordia intercept ages for individual analyses are quoted at the 1σ level, whereas the errors for weighted mean ages in the text are quoted at the 95% confidence level.
The in situ Lu-Hf analyses of zircons from all four TTG samples were conducted on the pits generated during U-Pb dating at the State Key Laboratory of Lithospheric Evolution, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, using a 193 nm UV ArF excimer laser ablation system attached to a Neptune multi-collector ICP MS. The instrumental conditions and analytical procedures were described by Wu et al. 91 . Each measurement included an ablation time of ∼26 s for 200 cycles, a repetition rate of 6-8 Hz, a laser power of 100 mJ/pulse and a spot size with a diameter of 44 μm. Helium was used as the carrier gas for the ablated aerosols. The average 176 Hf/ 177 Hf ratios of the Mud Tank 92 and Plešovice 93 standards obtained in this study after repetitive analyses were 0.282500 (n = 25) and 0.282484 (n = 18), respectively. All Lu-Hf isotopic results are reported with 95% confidence limits.
The Pb/ 206 Pb concordant and some discordant (10-15% discordant) age data of zircon spots with Hf isotope values consistent with concordant ones are summarized in Tables 1 and 2. All U-Pb age data and Lu-Hf isotopic data are listed in the Supplementary Material SF1 and 2. During interpretation, zircon 207 Pb/ 206 Pb age data with >10% U-Pb discordance and Th/U ratios of <0.15 were commonly disregarded. However, some discordant data, such as those with 176 Hf/ 177 Hf ratios identical to those of the concordant population, were included because although their U-Pb ratios have been modified, their original Lu-Hf isotopic ratios were preserved.
Availability of materials and data. All data generated or analysed during this study are included in this published article and its Supplementary Tables (SF1 and 2).