Appendix S4
Individual genetic comparisons
In addition to the effect size comparisons provided in the main text, we used more traditional
methods to compare individual genetic diversity estimates between historical and contemporary
populations. These were general linear models (GLMs) combined with multiple comparison tests
(using the function glht in the R package MULTCOMP (Hothorn et al. 2008) and permutation tests
(number of permutations = 5000). We relied on GLMs and multiple comparisons for genetic
diversity estimates that had associated mean values (e.g. number of sequences/alleles per
individual). Permutation tests were used for data that did not have associated mean values (e.g.
total number of polymorphic loci), and allowed us to assess whether the results were biased by
unequal sample sizes across populations. GLMs and permutation tests were performed in R
v2.15 (R Core Team 2012). The results of individual genetic comparisons for each species are
presented below.
South Island saddlebacks
Severely bottlenecked South Island saddlebacks showed a substantial loss in the number of
MHCIIB nucleotide sequences and translated sequences from historical mainland to
contemporary samples (Table 1, main text). The number of nucleotide sequences and number of
translated sequences for historical mainland samples was above the 95% CI from the
permutation tests, suggesting that the difference between historical mainland and contemporary
ones was significant. Samples from Big and Ulva islands had fewer than expected nucleotide and
translated sequences, suggesting that these were also significantly lower than expected.
Historical mainland samples had a greater number of segregating sites, but fewer average
nucleotide differences than expected from permutation tests. Historical island samples had fewer
average nucleotide differences. Samples from Big and Motuara Islands had greater than expected
average nucleotide differences, suggesting that the few sequences remaining in these populations
were highly divergent.
As a measure of allelic richness, theta k reflected a pattern of loss of MHCIIB diversity in
South Island saddlebacks. Theta k was significantly greater for historical mainland samples
compared to any other group (95% CIs do not overlap; Table 1, main text; Fig. A4.1). However,
theta k was similar between historical island and contemporary island samples (95% CIs overlap;
Table 1, main text; Fig. A4.1). Our GLM indicated that there were no significant differences
between historical and contemporary samples with respect to the number of translated sequences
per bird. However, differences in APD were significant, being highest for historical mainland
samples, followed by historical island samples and then contemporary samples (GLM F =
143.30, p < 0.00; Table 1, main text; Fig. A4.2).
Fig. A4.1: Theta k among South Island saddleback, North Island saddleback, South Island robin, and
North Island robin populations. Points represent mean values, while horizontal bars indicated 95%
confidence intervals (95% CI). Non-overlapping CIs indicated significant differences in theta k between
populations.
Figure A4.2: APD for MHCIIB and microsatellite diversity among South Island saddleback, North Island
saddleback, South Island robin and North Island robin populations.
With respect to microsatellite diversity, there was a substantial loss in the number of
polymorphic loci between historical mainland South Island saddlebacks and all other populations
(Table 1, main text). The number of polymorphic loci in historical mainland samples was above
the 95% CI from our permutation tests. Allelic richness standardized to sample size (n = 4) was
significantly greater in historical mainland samples (Table 1, main text; GLM F = 10.55, p <
0.00). APD was also greater in historical mainland samples compared to all other populations
(GLM F = 199.70, p < 0.00; Table 1, main text; Fig. A4.2).
North Island saddlebacks
For North Island saddlebacks, the numbers of MHCIIB nucleotide sequences and translated
sequences for the historical island and Hen Island populations were within the 95% confidence
intervals of the permutation test. However, Tiritiri Matangi Island and Mokoia Island had fewer
than expected nucleotide sequences and translated sequences (Table 1, main text). The number
of segregating sites and the average number of nucleotide differences between sequences were
not consistently different between historical and contemporary samples.
Although theta k decreased from historical to contemporary North Island saddleback
samples, differences were not significant (95% CIs overlapped; Table 1, main text; Fig. A4.1).
Our GLMs indicated that differences in the number of translated sequences per bird were also
not significant. APD did not differ significantly between historical and contemporary samples
(Table 1, main text; Fig. A4.2). With regards to MHCIIB functional supertypes, all of the five
North Island saddleback supertypes were represented in each population, and the number of
supertypes per bird did not differ significantly among populations.
With respect to microsatellite diversity, the number of polymorphic loci did not differ
significantly among North Island saddleback populations, as all values were within the 95%
confidence intervals from the permutation tests (Table 1, main text). Allelic richness
standardized to sample size (n = 4), average observed heterozygosity, and APD were not
significantly different between historical and contemporary samples (Table 1, main text; Fig.
A4.2).
South Island robins
For South Island robins, the numbers of MHCIIB nucleotide sequences was lower in
contemporary samples than expected based on our permutation test (Table 1, main text).The total
number of translated sequences was lower in contemporary samples and in historical island
samples, than expected based on the permutation test (Table 1, main text).The number of
segregating sites was similarly lower in contemporary samples than expected (Table 1, main
text). The average number of nucleotide differences between sequences was higher than
predicted in Dunedin and Motuara Island populations, but lower in Ulva Island, suggesting
differences in sequence divergence (Table 1, main text).
There were no differences in theta k between any South Island robin populations (95% CIs
overlapped; Table 1, main text; Fig. A4.1). The number of translated sequences per bird differed
only between historical island and Dunedin samples (GLM F = 2.89, p = 0.03), being higher, on
average, in Dunedin (Table 1, main text). APD was significantly higher in historical mainland
compared to all contemporary populations, but not compared to historical island samples (GLM
F = 15.34, p < 0.00; Table 1, main text; Fig. A4.2). We identified seven functional MHCIIB
supertypes in South Island robins. Except for Ulva Island, which had five supertypes, all
populations contained six (Table A4.1). We did not find any other differences between
supertypes in historical samples and contemporary populations.
Table A4.1: Proportions of South Island robin samples representing each MHCIIB supertype per
population. Contemporary populations are Dunedin, Motuara Island and Ulva Island. Each supertype was
defined on the basis of z-descriptors for 12 positively selected amino acids.
n
Supertype
1
2
3
4
5
6
7
Range/bird
Historical
mainland
6
Historical
island
5
Dunedin
20
Motuara
Island
16
Ulva Island
19
0.33
0.17
/
1.00
0.33
1.00
0.83
1.00
0.40
/
0.80
0.20
0.40
1.00
1.00
/
0.75
1.00
0.63
0.69
1.00
0.56
0.25
/
0.94
0.56
1.00
0.94
0.33
/
0.13
1.00
/
1.00
1.00
3-5
2-5
4-6
3-6
3-5
The numbers of polymorphic microsatellite loci were lower in contemporary Dunedin and
Motuara Island than expected based on the permutation tests, while historical populations and
Ulva Island fell within the expected range based on 95% confidence intervals (Table 1, main
text). Overall, differences in microsatellite allelic richness standardized to sample size (n = 3)
among populations were not significant (GLM F = 2.34, p < 0.07), however, our multiple
comparison test indicated that the difference between historical mainland and Dunedin was
significant (p = 0.04). Microsatellite APD was significantly greater in historical mainland
samples compared to Dunedin and Motuara Island (GLM F = 114.80, p < 0.00; Table 1, main
text; Fig. A4.2).
North Island robins
For North Island robins, the total numbers of MHCIIB nucleotide sequences and translated
sequences were lower in contemporary samples than expected based on the permutation tests
(Table 1, main text). The number of segregating sites was similarly lower in contemporary
samples than expected (Table 1, main text). However, the average number of nucleotide
differences between sequences was higher than predicted in the Dunedin and Motuara Island
populations, but lower in the Ulva Island population, suggesting differences in sequence
divergence (Table 1, main text). The number of segregating sites was lower than expected for
Tiritiri Matangi and Kapiti Islands, but the average nucleotide differences between sequences
was higher than expected on Kapiti Island (Table 1, main text).
Theta k was significantly greater in North Island robin historical mainland samples,
compared to Kapiti Island, but not compared to the other contemporary populations (Table 1,
main text; Fig. A4.1). The number of translated sequences per bird was not significantly different
between historical and contemporary samples. Similarly, there were no significant differences in
APD between historical and contemporary samples (Table 1, main text; Fig. A4.2).
We identified 10 functional MHCIIB supertypes in North Island robins. Mamaku Plateau
represented all 10 supertypes; while all other populations contained 8-9 supertypes (Table A4.2),
but differences in the number of supertypes per bird were not significant.
Table A4.2: Proportions of North Island robin samples representing each MHCIIB supertype per
population. Contemporary populations are Mamaku Plateau, Tiritiri Matangi Island, Mokoia Island and
Kapiti Island. Each supertype was defined on the basis of z-descriptors for 18 positively selected amino
acids.
n
Supertyp
1e
2
3
4
5
6
7
8
9
10
Range/bir
d
Historical
mainland
4
Mamaku
Plateau
19
Tiritiri Matangi
Island
14
Mokoia
Island
13
Kapiti
Island
16
0.75
/
0.50
0.75
0.50
1.00
/
0.50
0.75
0.75
0.84
0.74
0.74
0.37
0.26
0.95
0.53
0.42
0.53
0.63
0.93
0.57
0.57
0.36
0.36
0.86
0.50
0.50
0.29
/
1.00
0.62
1.00
0.77
0.15
0.85
0.23
0.77
/
0.62
1.00
0.69
1.00
0.06
/
0.63
0.44
0.13
0.75
0.50
5-6
3-9
3-8
4-8
4-7
With respect to microsatellite variation, the permutations did not suggest that the number
of polymorphic loci differed significantly among populations, or were biased by sample sizes
(Table 1, main text). Allelic richness standardized to sample size (n = 3) did not differ
significantly among populations (Table 1, main text). Similarly, observed heterozygosity did not
differ significantly among populations (Table 1, main text). As with MHC APD, microsatellite
APD did not differ significantly between historical and contemporary North Island robin
populations (Table 1, main text; Fig. A4.2)
References
Hothorn T, Bretz F, Peter W (2008) Simultaneous Inference in General Parametric Models.
Biometrical Journal 50, 346-363.
R Core Team (2012) R: A language and environment for statistical computing. http://www.Rproject.org.