SEARCH

SEARCH BY CITATION

This summer, one of our coauthors disclosed that there were errors in sequence alignment and editing in Barton et al. (2010). We have subsequently resubmitted our clean sequences to GenBank. We have re-analysed the data and found that while all of the absolute values of the results changed, which also changed the study design, the main inferences of the original manuscript have not changed, with one exception.

Changes in sequence editing and alignment

  1. Top of page
  2. Changes in sequence editing and alignment
  3. Synopsis of changes in interpretation
  4. Comparison of original analytical results to the re-analysis
  5. Antigenic site II
  6. Patristic distance
  7. TMRCA
  8. Numbers of infections (skyline plots)
  9. Striped skunk amova
  10. Rabies amova
  11. Reference
  • 1
     Multiple sequences (both N gene and G gene data sets) had indels removed or added to bring alignments back into reading frame (24 sequences for N gene and five sequences for G gene).
  • 2
     Four sequences were removed completely from the study (3 N gene and 1 G gene).
  • 3
     32 N gene sequences needed at least one nucleotide change, and 150 G gene sequences needed at least one nucleotide change (based on reviewing the chromatograms again).
  • 4
     Two mislabeled sequences were fixed, and as a result the zone of sympatry disappeared from the study.

This last change in sequence analysis and interpretation has resulted in a less robust study design because comparisons across the same landscape are no longer possible. Nonetheless, patterns of viral demography and landscape genetics remain the same for each rabies strain.

Synopsis of changes in interpretation

  1. Top of page
  2. Changes in sequence editing and alignment
  3. Synopsis of changes in interpretation
  4. Comparison of original analytical results to the re-analysis
  5. Antigenic site II
  6. Patristic distance
  7. TMRCA
  8. Numbers of infections (skyline plots)
  9. Striped skunk amova
  10. Rabies amova
  11. Reference

We found no changes in skunk population genetics after groups of skunks were reorganized to account for the lack of a zone of sympatry. We still found strong evidence for a selective sweep in the SCSK rabies variant. Estimates of diversity and Bayesian Skyline analysis all point to the spread of an efficiently transmitted virus that spreads on the landscape faster than NCSK.

We no longer found evidence for greater diversity in the G gene in SCSK. Previously we interpreted greater diversity as an explanatory mechanism for how SCSK spread more rapidly than NCSK.

The pattern of differential landscape barriers to virus spreading among strains remains. NCSK isolate diversity and gene flow was not affected by landscape geographic features, but we found rivers as barriers to gene flow in SCSK. Previously, an amova supported this inference, but now it does not. Instead of partitioning by rivers, amova of SCSK strain diversity reflects more structuring of the virus by the county in which it was located. We do not think this detracts from our finding that rivers are barriers to gene flow in SCSK using our landscape genetic analysis.

Comparison of original analytical results to the re-analysis

  1. Top of page
  2. Changes in sequence editing and alignment
  3. Synopsis of changes in interpretation
  4. Comparison of original analytical results to the re-analysis
  5. Antigenic site II
  6. Patristic distance
  7. TMRCA
  8. Numbers of infections (skyline plots)
  9. Striped skunk amova
  10. Rabies amova
  11. Reference

After realignment and re-editing of the sequences, the following results changed between the original analyses and the re-analysis (results pertaining to the original zone of sympatry are no longer appropriate and have been removed):

Original Table 1 Diversity and neutrality statistics for NCSK and SCSK using complete N gene datasets containing both allopatric and sympatric samples for each strain suggested the presence of similar levels of diversity in both strains, but slightly higher infectivity and more purifying selection in SCSK. The same analyses using only samples from the sympatric zone for each strain revealed marginal differences in diversity between the two strains, but more intense purifying selection and twice as much infectivity in SCSK when compared to NCSK

 Sympatric and allopatricSympatric only
NCSK (N = 98)SCSK (N = 160)NCSK (N = 51)SCSK (N = 60)
Nucleotide Diversity0.03 ± 0.020.02 ± 0.010.05 ± 0.020.02 ± 0.01
Pairwise Differences26.26 ± 11.6111.93 ± 5.4237.00 ± 16.3610.86 ± 5.01
Fu & Li’s D*−7.11**−4.17**−5.56*−3.39*
Fu & Li’s F*−6.22**−3.99**−5.35*−3.61*
Tajima’s D−2.56***−2.27***−2.52***−2.36***
Fu’s Fs−23.95**−24.12**−23.79***−24.52***
dN/dS0.570.370.660.30
No. of skunks/isolate1.21.81.22.4

*P < 0.05, **P < 0.02, ***P < 0.001.

New Table 1 Diversity and neutrality statistics for NCSK and SCSK using the N gene dataset for each strain suggested the presence of similar levels of diversity in both strains, but higher infectivity and more purifying selection in SCSK

 NCSK (N = 98)SCSK (N = 152)
Nucleotide diversity0.01 ± 0.0070.01 ± 0.006
Pairwise differences8.99 ± 4.187.64 ± 3.58
Fu & Li’s D*−3.88*−2.29
Fu & Li’s F*−3.83*−2.58*
Tajima’s D−2.17***−1.89***
Fu’s Fs−24.60***−24.65***
dN/dS0.0270.005
No. of skunks/isolate1.52.5

*P < 0.05, **P < 0.02, ***P < 0.001.

Interpretation: The overall diversity is lower in the re-analyzed data but the overall trend remains the same, particularly the dN/dS and number of skunks infected with a particular isolate. We used these parameters to make our interpretation about selective sweep in the SCSK.

Original Table 2 Diversity and neutrality statistics for NCSK and SCSK using complete G gene datasets containing both allopatric and sympatric samples for each strain suggested the presence of similar levels of diversity in and infectivity in both strains, but more purifying selection in NCSK. The same analyses using only samples from the sympatric zone for each strain revealed more intense purifying selection in the NCSK when compared to SCSK

 Sympatric and allopatricSympatric only
NCSK (N = 103)SCSK (N = 166)NCSK (N = 51)SCSK (N = 60)
Nucleotide diversity0.02 ± 0.010.04 ± 0.020.02 ± 0.010.03 ± 0.01
Pairwise differences17.68 ± 7.9235.49 ± 15.5316.66 ± 7.5421.09 ± 9.44
Fu & Li’s D*−4.76**−7.08**−3.61**−4.90**
Fu & Li’s F*−4.49**−5.91**−3.64**−4.79**
Tajima’s D−2.34***−2.32***−2.02**−2.07**
Fu’s Fs−24.01***−23.80***−24.22***−24.11***
dN/dS0.0880.220.0770.56
No. of skunks/isolate0.860.941.131.07

*P < 0.05, **P < 0.02, ***P < 0.001.

New Table 2 Diversity and neutrality statistics for NCSK and SCSK using the G gene dataset for each strain suggested the presence of similar levels of diversity and purifying selection in both strains, but slightly lower diversity in SCSK

 NCSK (N = 102)SCSK (N = 166)
Nucleotide Diversity0.01 ± 0.0070.01 ± 0.006
Pairwise Differences11.18 ± 5.129.53 ± 4.39
Fu & Li’s D*−1.46−2.18
Fu & Li’s F*−2.05−2.58*
Tajima’s D−1.98**−2.10***
Fu’s Fs−24.35***−24.33**
dN/dS0.030.05
No. of skunks/isolate1.21.5

*P < 0.05, **P < 0.02, ***P < 0.001.

Antigenic site II

  1. Top of page
  2. Changes in sequence editing and alignment
  3. Synopsis of changes in interpretation
  4. Comparison of original analytical results to the re-analysis
  5. Antigenic site II
  6. Patristic distance
  7. TMRCA
  8. Numbers of infections (skyline plots)
  9. Striped skunk amova
  10. Rabies amova
  11. Reference

Original Numbers: SCSK dN/dS = 1.97; NCSK dN/dS = 1.74.

New Numbers: SCSK dN/dS = 0.00; NCSK dN/dS = 0.42.

Interpretation: The overall diversity of the G gene is reduced. Here, our interpretation differs from our original publication. We found higher diversity in G gene in the NCSK strain. This invalidates the speculation that increased variation in G gene has allowed SCSK strain to spread across the landscape more quickly.

Patristic distance

  1. Top of page
  2. Changes in sequence editing and alignment
  3. Synopsis of changes in interpretation
  4. Comparison of original analytical results to the re-analysis
  5. Antigenic site II
  6. Patristic distance
  7. TMRCA
  8. Numbers of infections (skyline plots)
  9. Striped skunk amova
  10. Rabies amova
  11. Reference

Original Numbers: NCSK: Allopatric = 5.9 mutations, Sympatric = 16.36 mutations; SCSK: Allopatric = 2.05 mutations, Sympatric = 3.4 mutations.

New Numbers: NCSK: 2.31 mutations; SCSK: 1.03 mutations.

TMRCA

  1. Top of page
  2. Changes in sequence editing and alignment
  3. Synopsis of changes in interpretation
  4. Comparison of original analytical results to the re-analysis
  5. Antigenic site II
  6. Patristic distance
  7. TMRCA
  8. Numbers of infections (skyline plots)
  9. Striped skunk amova
  10. Rabies amova
  11. Reference

Original Numbers: TMRCA for NCSK including sample SDSK030554 as ranging from 450 to 2150 years ago, and NCSK without SDSK030554 as ranging from 50 to 550 years ago. The TMRCA for SCSK ranged from 75 to 205 years ago.

New Numbers: TMRCA for NCSK as ranging from 33 to 213 years ago. The TMRCA for SCSK ranged from 44 to 136 years ago.

Numbers of infections (skyline plots)

  1. Top of page
  2. Changes in sequence editing and alignment
  3. Synopsis of changes in interpretation
  4. Comparison of original analytical results to the re-analysis
  5. Antigenic site II
  6. Patristic distance
  7. TMRCA
  8. Numbers of infections (skyline plots)
  9. Striped skunk amova
  10. Rabies amova
  11. Reference

Original Numbers: While the skyline plots for NCSK and SCSK showed a steady number of infections over time (300 and 100 infections respectively) (Overlapping CIs).

New Numbers: NCSK 200 infections; SCSK initially showed 20 infections, but was followed by a rapid increase in the effective number of infections between 2005 and 2008 (100 infections) before rapidly dropping off again to the initial number of effective infections (Overlapping CIs).

Interpretation: The Bayesian Skyline analysis supports the observation of increased spread of SCSK based on case data.

Original Table 3 Population genetic analyses of striped skunk populations divided by rabies strain revealed slightly higher genetic diversity among skunks in the south than in the north. Analysis of sympatric and allopatric populations revealed similar effective population sizes for all three populations. Analysis of the populations combined into a global population showed the presence of a population bottleneck. Seven of eight loci were out of Hardy–Weinberg equilibrium in the global population

 ARHOHEθNeBottleneck (P-value)HWE (P-value)
2 Strains
 NCSK12.57 ± 0.890.767 ± 0.010.877 ± 0.010.01
 SCSK14.50 ± 0.800.789 ± 0.010.896 ± 0.010.02
3 Groups
 Allopatric North13.25 ± 0.770.772 ± 0.040.893 ± 0.015.4942.70.03
 Sympatric11.65 ± 0.790.754 ± 0.040.875 ± 0.017.2768.40.006
 Allopatric South12.13 ± 0.760.808 ± 0.020.883 ± 0.017.5376.20.01
Global
 Pooled data15.63 ± 1.050.778 ± 0.030.889 ± 0.010.01
 22–7017.000.8490.9020.01
 22–6720.000.7330.916≪0.001
 22–1415.000.7910.8930.009
 42–2619.000.8370.916≪0.001
 42–1514.000.7440.8500.006
 42–2516.000.7210.9010.001
 22–1912.000.6740.856≪0.001
 42–7312.000.8720.8770.4

New Table 3 Population genetic analyses of striped skunk populations divided by rabies strain revealed similar levels of genetic diversity among skunks in the south and in the north. Analysis of populations divided by rivers revealed similar effective population sizes for all three populations. Analysis of the populations combined into a global population showed the presence of a population bottleneck. Seven of eight loci were out of Hardy–Weinberg equilibrium in the global population

 ARHOHEθBottleneck (P-value)HWE (P-value)
2 Strains
 NCSK13 ± 0.420.767 ± 0.020.88 ± 0.014.840.04
 SCSK14.25 ± 0.330.789 ± 0.020.895 ± 0.019.330.04
3 Groups
 North of Missouri River12.5 ± 0.410.763 ± 0.020.879 ± 0.012.670.03
 Missouri to Platte10.63 ± 0.530.824 ± 0.030.885 ± 0.0214.300.04
 South of Platte River13.38 ± 0.410.772 ± 0.030.893 ± 0.016.030.04
Global
 Pooled data15.63 ± 1.050.778 ± 0.030.889 ± 0.010.01
 22–7017.000.8490.9020.01
 22–6720.000.7330.916≪0.001
 22–1415.000.7910.8930.009
 42–2619.000.8370.916≪0.001
 42–1514.000.7440.8500.006
 42–2516.000.7210.9010.001
 22–1912.000.6740.856≪0.001
 42–7312.000.8720.8770.4

Striped skunk amova

  1. Top of page
  2. Changes in sequence editing and alignment
  3. Synopsis of changes in interpretation
  4. Comparison of original analytical results to the re-analysis
  5. Antigenic site II
  6. Patristic distance
  7. TMRCA
  8. Numbers of infections (skyline plots)
  9. Striped skunk amova
  10. Rabies amova
  11. Reference

Pairwise FST for groups on either side of Missouri was the same: 0.005, but P-value changed from 0.29 to 0.33.

Justification for redoing skunk genetic analysis: We had to reassign skunks to different populations because the zone of sympatry disappeared. The slightly altered data do not present a different interpretation.

Original Table 4 Results of AIC model selection for all landscape permeability models evaluated for skunk rabies across the Central Great Plains

Modelr-squareK−2LN(P)ΔAICWiCC-P
North
 $IBD*0.032409900.250.50.71
 IBD × RB × EM0.019411212.5<0.010.50.13
 EM*0.04340941.90.100.50.08
 RB*0.16341062.20.080.50.40
 SIBR*0.12240613.10.050.50.01
 CM*0.13340572.30.080.50.34
 IBD × RB0.18640347.1<0.010.60.09
South
 IBD*0.4521320.90.380.60.01
 IBD × RB × EM0.15913813.2<0.010.60.01
 EM0.0631379.0<0.010.50.20
 RB*0.2831345.70.030.6<0.01
 SIBR0.19213811.9<0.010.50.19
 CM0.63312912.0<0.010.60.04
 $IBD × RB*0.6361290.00.0.590.60.09

*Indicates potentially good models with ΔAICc values ≤6.0. $Indicates the AICc model with the highest weight. All models were run independently for NCSK and SCSK as molecular data suggest that the two rabies strains are distinct. Variables are as follows: r-square is the r2 value from regression analysis, K = the number of parameters included in the model determined by the number of different land cover and distance values included in the GIS land cover ecological resistance model +1 for the regression intercept parameter, −2Ln(K) is the negative log likelihood for the model with K parameters, ΔAIC is the standardized difference in AIC values for the given model, Wi is the AIC weight for the given model and can be thought of as the degree of certainty that it is the best model given the data, C is the C value from Hosemer–Lemeshow goodness of fit tests for the regression model, and the C-P value is the associated P-value for the goodness of fit test. Model variables are defined as follows: IBD, isolation by distance; CM, combined model enhanced edge effects and river barriers; EM, edge model; RB, rivers barrier model; SIBR, simple isolation by resistance; IBD × RB, interaction between IBD and RB model; IBD × RB × EM, interaction between IBD, RB, and EM models.

New Table 4 Results of AIC model selection for all landscape permeability models evaluated for skunk rabies across the Central Great Plains

ModelK−2LN(K)ΔAICCwiR2PH–L (H)H–L (P)
NCSK
 IBR (SIBR)2−51.974800.6332270.1760.035  
 RB3−52.28162.1647870.2145270.1310.029  
 EM3−53.57812.8506720.1522450.3110.038  
 IBD × RB × EM*9−52.281614.16479 0.1310.0352.840.093
 IBD × RB6−323.54959.20058 0.3950.032  
 IBD*2−595.12368.26429 0.440.00110.065
SCSK
 RB3−117.129010.1310.029  
 IBD × RB × EM*9−117.12912 0.1310.0350.14560.01
 IBR (SIBR)2−228.0921.9928 0.1760.035  
 EM3−329.71437.2582 0.3110.038  
 IBD × RB6−15996183.0056 0.3950.032  
 IBD*2−30821198.6163 0.440.0010.10.05

*Global models for which Goodness of Fit was assessed, all other models represent nested sub-models of the global models and so do not require independent GOF tests. Hosmer–Lemeshow (H) = the GOF test, test statistic and Hosmer–Lemeshow (P) = the P-value for the GOF test indicating significant lack of fit. All models with (H) ≥ 1.0 have a lack of fit. K = number of parameters in model, −2LN(K) = the maximum likelihood estimate from Mantel tests, ΔAICC = Akaike Information Criterion adjusted for small sample size bias, in our analysis only models with ΔAICC < 6 were considered competitive models, wi = the AICC model weight, and R2 = the Mantel test r-square value.

Interpretation: We found no landscape or geographic feature that predicted patterns of isolate diversity in NCSK, however, for the SCSK strain of rabies, rivers were barriers to gene flow. This interpretation remains the same as the original paper.

Rabies amova

  1. Top of page
  2. Changes in sequence editing and alignment
  3. Synopsis of changes in interpretation
  4. Comparison of original analytical results to the re-analysis
  5. Antigenic site II
  6. Patristic distance
  7. TMRCA
  8. Numbers of infections (skyline plots)
  9. Striped skunk amova
  10. Rabies amova
  11. Reference

Original Numbers: NCSK revealed no significant partitioning of viral isolates by rivers (ΦGROUP-RIVER; ΦCT = −0.007, P = 1.00); SCSK revealed that rivers were a significant barrier and accounted for ∼26% of the variation in that strain (ΦGROUP-RIVER; ΦCT = 0.26, P = 0.08).

New Numbers: (in a table form to reflect that more of the variation is between populations within groups rather than among groups).

New Table. Results of amova for NCSK and SCSK.

 NCSKSCSK
Among groups (FCT)0.05, P = 0.3210.08, P = 0.203
Among populations within groups (FSC)0.14, P = 0.0030.30, P ≪ 0.001
Within populations (FST)0.18, P ≪ 0.0010.36, P ≪ 0.001

Interpretation: We did not find that viral strain similarity clustered by groups (i.e. that rivers did not structure the viral population for either strain). This finding was different than our original interpretation in which we found that SCSK was structured by rivers. The strong signal of population structure among populations (counties) within each strain, may obscure the relationship with rivers.

Figure 1 incorrectly depicts a zone of sympatry of the two viral strains.

Figure 2 is no longer valid.