Supplementary information:

Classification of Dinornis:

Since the very first publications1,2 the primary, if not sole, basis for the distinction of different Dinornis species has been size. Earlier researchers recognised Dinornis specimens on each island as different species, because similar length leg bones in the South Island were generally stouter than their North Island counterparts3-5. Cracraft (1976) applied the biological species concept to the analysis of Dinornis size variation, and rationalised a plethora of taxa to just 4 species, with 3 occurring on both islands6. Worthy (1989a) synonymised the fourth species, and maintained 3 after being influenced by an apparently trimodal structure in the large Makirikiri collection 7,8. Small differences were detected in D. novaezealandiae basicranium, post-orbital cranial widths, and leg bone robustness compared to the other species. However, the middle and upper size groups were weakly differentiated and it would be possible to accommodate all specimens in a bimodal structure if the group of larger individuals was allowed a greater absolute size range than that of the smaller ones as might be expected if each group was to have similar relative variation. The possibility of combining all individuals into one species was not considered, as the resultant size variation would have been outside that of any known birds. The continued acceptance of the three-species model was likely to have been influenced by the largely separate geographic distributions of the mid-sized D. novaezealandiae and large D. giganteus.

The morphological features distinguishing the three Dinornis species are, other than size, relatively minor and may in fact be size and/or age related, e.g. both the post-orbital processes and basicranial mamillar tuberosities are ligamental attachment points and could be expected to increase in size with advanced maturity and in larger individuals. Where parallel shifts in size variation have been detected in D. giganteus and D. struthoides7,8, they can be related to either geographic or temporal variation.

Morphological measurements:

Morphological data were drawn from the most comprehensive treatment of moa size variation available9, which also summarises previous work and estimates. Adult status is easily discernible in moa leg bones, and can be diagnosed by fused patella, complete synostosis of distal tarsal bones and metatarsals, and the shape and density of femur articular surfaces9. Individual size was determined by adding the lengths of the three main leg bones, to give comparable estimates for height at back. These measurements are not thought to differ markedly when correct anatomical posture is used (similar to that of an emu or cassowary), due to the additional height of pelvis above femur, foot-pads, cartilaginous joints etc. Individual weights were calculated using femur length and well-established algorithms10 where femur length = 61.64 * mass0.359. If not exact, this rough estimate is nevertheless comparable between individuals and taxa, and is consistent with other algorithms10-12 . The various strengths and rationales of these weight and height estimates are discussed in detail in Worthy and Holdaway (2002) along with the mean, maximum, and minimum values for height and weight for 8 Dinornis struthoides, 12 Dinornis novaezealandiae and 13 Dinornis giganteus individuals using three different algorithms. The measurements based on the available palaeontological data reveal that on average females (DIGI and DINO) were around 150% heavier, and 126% taller than sympatric males (DIST). When individuals with molecular sex data (table 4) are compared, the differences range from 256% to 351% the weight, and 135 to 173% the height, although this is likely to be influenced by the smaller number of DIST specimens, especially from eastern SI sites. Further morphological analyses and molecular sexing data will be required to accurately determine the full extent of the RSD however. For example, identification problems involving individuals around the DIST and DINO size boundary, combined with regional variation, currently prevent comprehensive analyses.

Phylogenetic methods:

Metropolis-Hastings Markov chain Monte Carlo (MCMC) integration was used to jointly estimate the phylogenetic tree and parameters of the substitution model13. An HKY+G+I model of substitution was used to characterize the evolution of the concatenated control region and protein coding sequence (525bp control region, 1435bp of cytochrome oxidase I, II, tSer, tLys, and ATP8 genes). The tree presented is the maximum a posteriori (MAP) tree obtained from the MCMC analysis, because no informative priors were used this tree will correspond closely to a maximum likelihood (ML) estimate of the tree under the same model of substitution. The analysis was performed on control region sequences from 32 Dinornis specimens, and protein-coding sequences from 7 specimens (Digi660, Dist700, Dist238, Dino699, Dino237, Digi697, Dist753), using two Megalapteryx sequences as outgroups. The estimated tree topologies did not substantially differ if the protein-coding sequences were omitted. In particular the split between the North and South Island mitochondrial lineages has a posterior probability of 100% both with, and without, the protein data. It should be noted that this split represents the separation of the mitochondrial lineages, not the two populations, and the latter must necessarily be more recent.

A second analysis was undertaken, using only the 525bp of the control region, to estimate the divergence time of the North/South split in the Dinornis genealogy. The 9 protein-coding sequences were not included as they exhibited a different rate of evolution, complicating rate calibrations. In this analysis an HKY+G substitution model without a category for invariant sites was found to be sufficient to describe the evolutionary process in the control region. Assuming a standard avian control region rate estimate of approximately 20.8% per million years18, we used MCMC to obtain the posterior estimate for the time of the most recent common ancestor of Dinornis. The estimated age was 510,000 years BP with 95% highest posterior density (HPD) interval of (195,000 - 950,000). The benefit of the MCMC approach to dating lies in the ability to incorporate the uncertainty in the tree topology and branch lengths, and in doing so providing confidence intervals that accurately reflect that uncertainty.

Real-time PCR methodology:

We used real-time PCR to quantify the relative copy number of nuclear DNA (ADH) templates in the moa extracts to examine whether variation in PCR efficiency might explain the failure of the D. struthoides specimens to amplify the female-specific KW1 PCR products. The amplification of double stranded products was detected on an Applied Biosystems ABI 7000 thermocycler using SYBR Green detection chemistry. 25ul PCR reactions used a 2X SYBR master mix (Applied Biosystems) and 800nM of each ADH specific primer (RT-ADH5/6-f: TGGTATGCAACTGAAAGATACAGTCAC-3’ and RT-ADH5/6-r: CGCTGAGCAGGTTAGAGAGGG-3’), supplemented with 0.125U Hi-Fidelity Platinum Taq polymerase (Invitrogen). Thermocycling conditions followed manufacturers instructions. CT values were determined for each moa DNA extract across a dilution series to ensure the quantitative nature of the amplification. Most of the extracts contained inhibitors which prevented accurate quantitation (particularly those from swamps), and only six samples (table 1) had CT values which responded to the dilution series in a uniform manner. However, the comparative template numbers (below) show that the nuclear DNA concentration is broadly similar across a variety of putative taxa, and that the two D. struthoides specimens have some of the highest amounts of nuclear DNA. Consequently, it appears unlikely that the failure to amplify female-specific KW1 PCR products in all 10 D. struthoides is due to stochastic drop-out in PCR reactions. To further examine this possibility we repeated the 85bp ADH and KW1 amplification seven times in three specimens (Dist788, Dist744 and Digi799) in accordance with standard genotyping practice when dealing with low copy number DNA samples14. All three samples amplified the ADH product seven times, whereas only Digi799 amplified the KW1 product (data not shown). Consequently, we conclude that the D. struthoides specimens do in fact lack the KW-1 female specific locus.

Table 1. Nuclear DNA template concentration:

DNA extract
/ Copy number of nuclear ADH templates. Raw CT values and relative % copy number
Dist700 / 31.70 (100%)
Digi799 / 32.85 (45.1%)
Dist753 / 33.06 (39%)
Dino699 / 35.21 (8.8%)
Dino786 / 36.07 (4.8%)
Dino782 / 37.24 (2.1%)

The copy number of nuclear DNA templates is given as raw CT values, and as a relative percentage of the best preserved specimen, Dist 700.

Figure 1.

Emu* AAGGCAAAACgGCTTCTGAAAcAGCCgGGGCTTTT-AAAACCATTTTTAGCAgGGGgCacACGCCTCAg

Ostr. AAGGCCAAATGGGTTCTGAAATAGCCGGAGCTTT--AAAACTGTTTTTACCAAGGGGCACACCCCTCAG

Rhea AAGGCCAAATGAGTTCTGAAATAGCTGGGGCTTTT-AAAACTATTTTTACCAGAGGGGACACACCTCAG

Moa AAGGCAAAATGAGTTCTTAAAAAGCCAGGGTTTTTT-AAACAATTTTTAGCAGGGGGCGTGTGCCTCAG

MoaCP AAGGCAAAATGAGTTCTTAAAAAGCCAGGGTTTTTT-AAACAATTTTTAGCAGGGGGCGTGTGCCTCAG

Emu* AAATTAcaGACTGAcca-GcAAAAGAATTTaTAAAtCgcCCTTTAAACAAGCTgTTAAaGCAaTATag

Ostr AAATTACAGACTGACCAGGCAAAAGAATTTATAAATCACCCTTTAAACAAGCTGTTAAAGCAATATAA

Rhea AAATTACAGACTGACCAGGCAAAAGAATTTATAAATCACCCTTTAAAGAAGCTGTTAAAGCAATATAA

Moa AAATTACAGATGAACTGGGTACAAAAATTTTTAAACCAGCCTTTAAACAATCTGTTAAAGCAGTATAG

MoaCP AAATTACAGATGAACTGGGTACAAAAATTTTTAAACCAGCCTTTAAACAATCTGTTAAAGCAGTATAG

Alignment of moa and modern ratite KW1 sequences.

The sequence of the emu (*) was taken from published data15. The ostrich, rhea and moa sequences were generated from cloned products using the 112f and 267r primers. The sequence labelled Moa-CP is the independently replicated sequence generated in Copenhagen.

Table 2. Specimen data.

DNA Extract /

Species

/

Island

/ Site / Element sampled / Museum number /

Sex primers

(180bp) / Sex primers
(85bp) / ADH primers
(85bp) / Molec.
sex / Control region / Control region / Control region / Mitochondrial protein/tRNA 1.5Kb region
(N/S) / 112f-267r / 185f-260r / 230f-290r / (M/F) / HVR1 5'.1 / HVR1 5'.2 / HVR1 3'.1 / (COI, tSer,COII, tLys, ATP8)
Digi797 / D. giganteus / S / Bell Hill, Canterbury / TMT / MNZ S39964 / X / Ö / Ö / F / Ö / Ö / Ö / no data
Digi799ir,c,me,RT / D. giganteus / S / Bell Hill, Canterbury / F / MNZ S39875 / Ö / Ö / Ö / F / Ö / Ö / Ö / no data
Digi697 / D. giganteus / N / Waikaremoana / TT / MNZ S25761 / Ö / Ö / Ö / F / Ö / Ö / Ö / Ö
Digi713 / D. giganteus / N / Takapau Rd / Phal / MNZ S1015 / X / X / X / Undet / Ö / Ö / Ö / no data
Digi718 / D. giganteus / N / Makirikiri / TT / MNZ S145 / X / X / X / Undet / Ö / Ö / Ö / no data
Digi660ir,c,me / D. giganteus / S / Hodge Ck, Nelson / Phal / MNZ S34095 / Ö / Ö / Ö / F / Ö / Ö / Ö / Ö
Digi708 / D. giganteus / S / Pyramid Valley / F / CM AV23466 / X / Ö / Ö / F / Ö / Ö / Ö / no data
Digi709 / D. giganteus / S / Pyramid Valley / Phal / MNZ S34088 / X / X / X / Undet / Ö / Ö / Ö / no data
Digi710 / D. giganteus / S / Pyramid Valley / TMT / CM AV13779 / X / Ö / Ö / F / Ö / Ö / Ö / no data
Digi711 / D. giganteus / S / Pyramid Valley / F / CM AV8421 / X / Ö / Ö / F / Ö / Ö / Ö / no data
Digi712 / D. giganteus / S / Pyramid Valley / TMT / CM AV14449 / X / Ö / Ö / F / Ö / Ö / Ö / no data
Digi717 / D. giganteus / S / Pyramid Valley / TMT / CM AV8418 / X / X / X / Undet / Ö / Ö / Ö / no data
Digi719 / D. giganteus / S / Glenmark / TT / CM AV9532 / X / X / X / Undet / Ö / Ö / partial / no data
Digi746 / D. giganteus / S / McNabb Cheviot / F / CM SB47 / X / Ö / Ö / F / Ö / Ö / partial / no data
Digi783 / D. giganteus / N / Takapau / TT / MNZ 1014-No.6 / X / X / Ö / M? / Ö / Ö / partial / no data
Dino715 /
D. novaezeal
/ N / Waikaremoana / TT / MNZ S421 / X / Ö / Ö / F / no data / no data / no data / no data
Dino237 / D. novaezeal / S / Ellis Basin, Mt Owen / Rib / MNZ S32667 / Ö / Ö / Ö / F / Ö / Ö / Ö / Ö
Dino714 / D. novaezeal / S / Mt Owen / TT / MNZ S23342 / X / Ö / Ö / F / Ö / Ö / Ö / no data
Dino786RT / D. novaezeal / S / Takaka Fossil cave / TT (juv) / MNZ S38988 / X / Ö / Ö / F / Ö / Ö / Ö / no data
Dino782RT / D. novaezeal / N / Takapau / TT / MNZ24365 / X / X / X / Undet / Ö / Ö / Ö / no data
Dino716 / D. novaezeal / S / Takaka / TT / MNZ S211 / X / Ö / Ö / F / Ö / Ö / Ö / no data
Dino699RT / D. novaezeal / N / Waikaremoana / ? / MNZ S299 / X / Ö / Ö / F / Ö / Ö / Ö / Ö
Dist753me,RT / D. struthoides / N / Waikaremoana / TT / Brian Reeve / X / X / Ö / M / Ö / Ö / no data / Ö
Dist700RT / D. struthoides / N / Gabrielles cave / TMT (juv) / MNZ S37875 / X / X / Ö / M / Ö / Ö / Ö / Ö
Dist723 / D. struthoides / N / Te Aute / TMT / CM AV8846 / X / X / Ö / M / no data / no data / no data / no data
Dist744 / D. struthoides / N / Gabrielles cave / TT / MNZ S37874 / X / X / Ö / M / Ö / Ö / Ö / no data
Dist238 / D. struthoides / S / Maximus Cave / RIb / MNZ S28255 / X / X / Ö / M / Ö / Ö / Ö / Ö
Dist720 / D. struthoides / S / Kapua swamp / TT / CM AV8763 / X / X / Ö / M / Ö / Ö / Ö / no data
Dist721 / D. struthoides / S / Glenmark swamp / TT / CM AV9436 / X / X / X / Undet / Ö / Ö / Ö / no data
Dist724 / D. struthoides / S / Mt Owen / TT / MNZ S23570 / X / X / Ö / M / Ö / Ö / Ö / no data
Dist742 / D. struthoides / S / Glenmark swamp / TT / CM AV9440 / X / X / X / Undet / no data / Ö / Ö / no data
Dist788 / D. struthoides / S / Takaka Fossil cave / TT / MNZ S39003 / X / X / Ö / M / Ö / Ö / Ö / no data
Dist805ir,c, me / D. struthoides / S / Takaka Fossil Cave / ? / MNZ S39004 / X / X / Ö / M / Ö / Ö / Ö / no data
Dist749 / D. struthoides / S / Kapua swamp / TT / CM AV8766 / X / X / Ö / M / Ö / Ö / Ö / no data
Medi606 / M.didinus / S / Mt Owen / TT / MNZ S23808 / no data / Ö / Ö / Ö
Medi701 / M.didinus / S / Howdat Cave / TT / MNZ S28206 / Ö / Ö / Ö / Ö

Details of the moa specimens used in the study. Museums: Canterbury (CM), Museum of New Zealand (MNZ), and private collections (Brian Reeve). TMT = tarsometatarsus, TT = tibiotarsus, F = femur, Phal = phalange. Specimens that were extracted multiple times (me), cloned (c), independently replicated (ir), or quantified by real-time PCR (RT), are indicated. Morphological taxonomic identifications are given - D. giganteus (DIGI), D. novaezealandiae (DINO), and D. struthoides (DIST) with extract numbers. PCR products are indicated by ticks (success) or crosses (failure). Sequences have been deposited in Genbank (see main text for accession numbers).