SEARCH

SEARCH BY CITATION

Keywords:

  • Abiotrophy;
  • Cerebellum;
  • Degenerative disease;
  • Purkinje cell loss

Abstract

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Footnotes
  8. Acknowledgments
  9. References

Background: Cerebellar cortical degeneration exists in American Staffordshire Terriers. Magnetic resonance imaging (MRI) can be suggestive, but a definitive diagnosis requires histopathology.

Hypothesis: Computer-assisted MRI morphometry can be used to distinguish between American Staffordshire Terriers with or without cerebellar cortical degeneration.

Animals: Normal American Staffordshire Terriers (n = 17) and those with clinical signs of cerebellar cortical degeneration (n = 14).

Methods: This was a partly retrospective and partly prospective study. Causes of cerebellar disease were ruled out with brain MRI, cerebrospinal fluid (CSF) analysis, CBC, blood biochemistry, and clinical follow-up. On T2-weighted midsagittal MR images, the following parameters were calculated: size of the cerebellum relative to the entire brain, size of the CSF space surrounding the cerebellum relative to the cerebellum, and 2 threshold-dependent cerebellar CSF indices (with and without surrounding CSF).

Results: Statistical analyses indicated a significantly lower relative cerebellar size (P < .001) and a larger relative cerebellar CSF space (P < .001) in dogs with cerebellar cortical degeneration. The measurement of relative cerebellar size could distinguish between affected and nonaffected dogs with a sensitivity and a specificity of 93 and 94%, respectively, using a cut-off of 13.3%. Using a cut-off of 12.8%, the measurement of relative CSF space could distinguish between both groups with a sensitivity of 93% and a specificity of 100%. There was a significant difference in 1 of the 2 CSF indices between affected and normal dogs.

Conclusions and Clinical Importance: Relative cerebellar size and relative CSF space calculated from MRI are effective in American Staffordshire Terriers to differentiate between normal animals and those with cerebellar cortical degeneration.

Cerebellar cortical degeneration in dogs has been reported in at least 27 different pure breeds and 1 mixed-breed dog.1–40 The onset of clinical signs differs among breeds but clinical signs usually start at <6 months of age.1,8,16–37,39 Familial late-onset cerebellar abiotrophies have been described in Gordon Setters,2–4 Old English Sheepdogs,5 Brittany Spaniels,6,7 a Schnauzer-Beagle dog,9 Pit Bull Terriers,15 Espagneul Bretons,38 a Bernese Mountain dog,40 and American Staffordshire Terriers.10–15 Initially, clinical signs are very subtle, eg, an occasional sway of the body when making a sudden movement.13 Later, the neurologic signs consist of hypermetria, ataxia, nystagmus, and sometimes loss of menace response.13 Signs progress until affected dogs are unable to walk without falling.13 Gross pathologic examination typically reveals diffuse atrophy of the cerebellum.4,8,12,13,19,22,31,33,39,40 Histopathologic findings such as marked loss of Purkinje neurons, depletion of granular cell bodies, and shrinkage of the granular and molecular cell layer have been reported alone or in combination.2–40 In breeds with familial late-onset cerebellar abiotrophies, the mode of inheritance has been shown to be autosomal recessive in Old English Sheepdogs5 and Gordon Setters2,3 and it is suspected to be the same in American Staffordshire Terriers.13 The underlying cause has not been determined in any breed.

Aside from biopsy, no method to confirm the diagnosis of cerebellar cortical degeneration in live animals has been reported. A presumptive diagnosis is made by ruling out other cerebellar or vestibular diseases with cerebrospinal fluid (CSF) analysis, CBC, blood biochemistry, thyroid testing, urinalysis, brainstem auditory-evoked response (BAER), and computed tomographic (CT) scanning or magnetic resonance imaging (MRI) of the brain.13

The amount of CSF between the cerebellar folia has been reported to appear to be increased in several MRI studies, but no objective methods for comparing the volume of CSF in affected versus unaffected dogs have been reported.13,22,31,33,38,40 In this study, computer-assisted MRI morphometry was used to compare the cerebella of American Staffordshire Terriers with or without presumptive cerebellar cortical degeneration to determine whether it was possible to distinguish between affected and nonaffected dogs.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Footnotes
  8. Acknowledgments
  9. References

Fourteen American Staffordshire Terriers with a history and clinical signs compatible with cerebellar cortical degeneration were presented to the Department of Small Animal Medicine of the University of Leipzig between August 2000 and December 2006. Neurologic condition was graded from 0 to 13, with 1–3 points being mildly, 4–6 points being moderately, and >6 points being severely affected. The clinical signs used for scoring were as follows: ability to ambulate, grade of ataxia, deterioration of clinical signs after dorsoflexion of the head and neck and presence of nystagmus, body swaying, intention tremor, and menace response (see Table 1).

Table 1.   Neurologic condition was graded from 0 to 13 points using this score.
Symptom/Points0123
AmbulatoryYes No 
AtaxiaNoMildModerateSevere
NystagmusNoPositionalSpontaneous 
Body swayingNoYes  
Intention tremorNoMildSevere 
Menace responseNormalReducedAbsent 
Dorsoflexion of the head and neckNo changesSigns got worse  

CBC, serum biochemistry profiles, free l-thyroxine (T4) concentrations (measured by equilibrium dialysis), and thyroid stimulating hormone (TSH) concentrations were measured and urinalyses were performed on initial admission. CSF was collected from the cerebellomedullary cistern. Pandy test, total protein concentration, specific gravity, and cytologic analysis (total and differential cell count, including microscopic examination after sedimentation in a sedimentation chamber) were performed.

BAER was obtained with a commercial electrodiagnostic analytical system (Nicoleta).

MRI was performed with the dogs in sternal recumbency with a 0.5 Tesla superconducting magnet (Gyroscanb) and a human knee coil. Sequences performed included T2-weighted images in transverse and sagittal planes (TR: 2,250 ms, TE: 100 ms, slice thickness 3 mm, interslice gap 0.5 mm, field of view 190–220), fluid-attenuated inversion recovery (FLAIR) in the transverse plane (TR: 5,000 ms, TE: 100 ms, TI: 1,900 ms, slice thickness 3 mm, interslice gap 0.3 mm), turbo inversion recovery (TIR) in the sagittal plane (TR: 3,000 ms, TE: 20 ms, TI: 400 ms, slice thickness 3 mm, interslice gap 0.3 mm), T1 3D images in the sagittal plane (TR: 18 ms, TE: 5.5 ms, slice thickness 0.47 mm), and T1-weighted images pre and post-IV contrast (0.1 mmol/kg gadopentetate dimegluminec) in the transverse plane (TR: 550 ms, TE: 10 ms, slice thickness 3 mm, interslice gap 0.5 mm). Measurements were made on midsagittal T2-weighted images with matrices 256 × 256 and 512 × 512.

Midsagittal T2-weighted images were used to subjectively assess the amount of CSF surrounding the cerebellum. It was defined as moderate when the amount of CSF in the 4th ventricle appeared to be increased, and as severe when there was additionally a clear CSF space on the dorsal part of the cerebellum and increased CSF in between cerebellar foliae.

Using T2-weighted sagittal midline images, 3 different areas (brain, cerebellum, and cerebellum plus CSF; see Fig 1) were manually circumscribed. The area was calculated by by computer software (Scion Image Beta 4.03d) to count the number of voxels in the circumscribed fields. Each area was circumscribed 3 times and the mean value was used for subsequent statistical analysis. In 3 affected and 3 unaffected dogs, the 3 areas were determined 10 times each by 2 different observers, and the intra- and interobserver variability (repeatability) was described with coefficient of variation (COV).

image

Figure 1.  T2-weighted midsagittal magnetic resonance image of the brain of an American Staffordshire Terrier, showing 3 different manually depicted regions of interest (ROI): the entire brain (continuous line along following structures: tip of tentorium osseum cerebelli-cerebellum close to the bony margin-cranial margin of foramen magnum going ventrally through the medulla oblongata at a right angle to the base of the skull-base of the skull, including the subarachnoidal space-cranial margin of pons-ventral surface of midbrain-including corpus mamillare, excluding hypophysis and chiasma opticum, including bulbus olfactorius-cerebrum close to bony margin-tentorium osseum cerebelli), the cerebellum (dotted line following directly the surface of cerebellum, including the dorsal recessus of the 4th ventricle) and the cerebellum plus cerebrospinal fluid (dashed line following the cerebellum close to the bony margin-along the floor of the 4th ventricle going dorsally through the aquaeductus mesencephali-caudal margin of the colliculus rostralis-tentorium osseum cerebelli).

Download figure to PowerPoint

The brain : cerebellum ratio (hereafter referred to as relative cerebellar size; see Fig 2) and the CSF space surrounding the cerebellum in relation to the cerebellum (hereafter referred to as relative CSF space; see Fig 3) were calculated.

image

Figure 2.  T2-weighted midsagittal magnetic resonance image of the brain of an American Staffordshire Terrier. Relative cerebellar size (striped area) was calculated as follows: Relative cerebellar size = number of voxels within the cerebellum (dotted line) × 100/number of voxels within the entire brain (continuous line).

Download figure to PowerPoint

image

Figure 3.  T2-weighted midsagittal magnetic resonance image of the brain of an American Staffordshire Terrier. Relative cerebrospinal fluid (CSF) space (striped area) was calculated as follows: relative CSF space = (number of voxels within the cerebellum plus CSF − number of voxels within the cerebellum) × 100/number of voxels within the cerebellum plus CSF.

Download figure to PowerPoint

To describe the amount of CSF associated with the cerebellum, a “CSF index” was determined as follows: a computer program (Scion Image Beta 4.03d) was used to create histograms of the shades of gray in the circumscribed areas of the cerebellum with or without surrounding CSF space. The original histograms contained only about 40 categories (190–230 of a possible 256; see Fig 4A). To improve the sensitivity and normalize each individual grayscale distribution, the first and last values of the histograms were redefined as 0 and 256, respectively. Grayscale values from 1 to 15 in the new histograms were considered to represent CSF (see Fig 4B). The fraction of voxels in this interval was calculated and the result was defined as the CSF index.

image

Figure 4.  (A) Histogram of voxels within circumscribed area of the cerebellum. All shades of gray were automatically placed in about 40 of possible 256 categories. Arrows denote the first and the last acquired values. (B) The converted histogram. The first and last acquired values were redefined as 0 and 256, respectively. Grayscale values from 1 to 15 in the new histogram were considered to represent cerebrospinal fluid (CSF) (“CSF voxels”). CSF indices were calculated as follows: CSF index of the cerebellum = number of CSF voxels × 100/number of voxels within cerebellum or CSF index of the cerebellum plus CSF = number of CSF voxels × 100/number of voxels within cerebellum plus CSF.

Download figure to PowerPoint

Control dogs (n = 17) were American Staffordshire Terriers presented to the hospital with other diseases. Dogs were of comparable age (not older than 10 years), weight, and sex without neurologic deficits. Comparability was investigated by an Aspin-Welch unequal-variance test (age), an equal-variance t-test (weight), and Fisher's exact test (sex). Structural brain lesions were ruled out with brain MRI scan in those dogs. BAER was performed in all control dogs. Bloodwork, urinalysis, and CSF analysis were not performed in control dogs.

The CSF indices, the relative cerebellar size, and the calculated relative CSF space were compared between the control group (n = 17) and affected dogs (n = 14) by a Mann-Whitney U-test. Values of P < .05 were considered to be significant. In addition, the optimal cut-off values and related diagnostic sensitivities and specificities of the relative cerebellar size and the calculated relative CSF space were determined by receiver operator characteristic (ROC) analysis.

Correlations between clinical signs and relative cerebellar size or relative CSF space were investigated by one-way ANOVA.

Results

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Footnotes
  8. Acknowledgments
  9. References

Patient Summary

Fourteen American Staffordshire Terriers with presumptive cerebellar cortical degeneration were examined at the Department of Small Animal Medicine at the University of Leipzig.

The first presentation to the hospital was at a mean age of 67 months (range, 52–80 months; standard deviation, 8.3 months). The mean age of the 17 control dogs was 70 months (range, 7–123 months; standard deviation, 32.1 months). There was no statistical difference in age between groups (P= .687).

At the time of writing, 12 affected dogs were still alive. Two dogs were euthanized 24 and 32 months after first clinical signs.

The mean weight of affected dogs was 32 kg (range, 25–45 kg; standard deviation, 5.5 kg) and 28 kg (range, 20–34 kg; standard deviation, 3.6 kg) in normal dogs. There was a significant difference in weight between both groups (P= .014).

In the group of affected dogs, there were 3 females and 11 males, and the control group included 8 female and 9 male dogs. However, there was no statistical difference in sex between both groups (P= .258).

Of the 14 affected dogs, 7 were mildly (1–3 points), 4 dogs were moderately (4–6 points), and 3 dogs were severely (> 6 points) affected. CBC and serum biochemistry profiles were within normal limits in all 14 dogs. Free T4 and TSH concentrations were within normal limits in 13 dogs (1 dog was not tested); on CSF analysis, cell counts and differentials were normal in all 14 samples. Total protein concentration was mildly increased in 8/14 samples (range, 26–39 mg/dL; reference < 25 mg/dL). Urinalyses were performed in 7 dogs. All results were normal, except for 1 dog in which cystine crystals were found. BAER analysis was normal in all dogs.

BAER analysis and brain MRI were normal in all control dogs.

MRI Findings and Measurement Results

A subjective increase in CSF surrounding cerebellum in 10/14 dogs appeared to be moderate in 7/14 or marked in 3/14 of affected dogs. Aside from these changes, all of the MR images appeared normal.

In affected dogs, the relative cerebellar size was significantly lower than in normal dogs (P < .001) (see Fig 5). The median relative cerebellar size for affected dogs was 12.4% (95% confidence interval [CI]: 10.4–13.1%). In normal dogs, the median relative cerebellar size was 14.9% (95% CI: 14.6–15.6%). Using a cut-off value of 13.3%, the measurement of relative cerebellar size could distinguish between affected and nonaffected dogs with a sensitivity of 93% and a specificity of 94%.

image

Figure 5.  Box plots: In affected dogs relative cerebellar size was significantly lower than in normal dogs (P < .001). The box represents the 25th, 50th, and 75th percentiles of the distribution; the whiskers approximate the 5–95th percentile range.

Download figure to PowerPoint

The relative CSF space in affected dogs was significantly larger than in normal dogs (P < .001) (see Fig 6). In affected dogs, the median relative CSF space was 21.7% (95% CI: 14.4–26.9%). In normal dogs, the median relative CSF space was 9.2% (95% CI: 7.1–11.0%). Using a cut-off value of 12.8%, the measurement of relative CSF space could distinguish between affected and nonaffected dogs with a sensitivity of 93% and a specificity of 100%.

image

Figure 6.  Box plots: The relative cerebrospinal fluid space in affected dogs was significantly larger than in normal dogs (P < .001). The box represents the 25th, 50th, and 75th percentiles of the distribution; the whiskers approximate the 5–95th percentile range.

Download figure to PowerPoint

Both cut-off values were generated by the ROC method.

The CSF index of the cerebellum without surrounding CSF in affected dogs was significantly higher than in normal dogs (P= .015). The median CSF index of the cerebellum for affected dogs was 0.5% (95% CI: 0.2–0.9%). In normal dogs, the median CSF index of the cerebellum was 0.3% (95% CI: 0.2–0.4%).

There was no statistically significant difference between the CSF index of the cerebellum plus CSF of affected and normal dogs (P= .149). The median CSF index of the cerebellum plus CSF for affected dogs was 0.4% (95% CI: 0.1–0.8%) and 0.3% in normal dogs (95% CI: 0.2–0.4%).

Dogs with smaller relative cerebellar sizes tended to have a more severe clinical score: the mean relative cerebellar size of severely affected dogs was 11.4% (range, 10.2–12.5%); dogs with an intermediate clinical score had a relative cerebellar size of 11.6% (range, 10.4–12.8%) and mildly affected dogs had a relative cerebellar size of 12.3% (range, 10.0–13.7%). However, according to one-way ANOVA, these differences were not statistically significant. Severely affected dogs had a mean relative CSF space of 26.7% (range, 20.4–32.6%), moderately affected dogs had a mean CSF space of 23.8% (range, 20.1–26.9%), and mildly affected dogs had a mean relative CSF space of 20.5% (range, 14.0–30.5%). The one-way ANOVA showed a lack of significance (P < .1) when severely and mildly affected dogs were compared.

The intra- and interobserver variabilities were tested. The mean COV within the observers (intraobserver variability) was low and not different between the observers (0.8 and 1.0% for the entire brain; 2.1 and 2.0% for the cerebellum plus CSF; 2.0 and 2.0% for the cerebellum). There was no difference in the mean COV on comparing measurements in normal and affected dogs. The mean COV within both observers (interobserver variability) for the 3 regions was low (0.6% for the entire brain; 1.4% for the cerebellum plus CSF; 1.3% for the cerebellum).

Discussion

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Footnotes
  8. Acknowledgments
  9. References

Cerebellar cortical degeneration is the most common neurodegenerative disorder in dogs. The clinical signs, progression, and histopathologic findings of cerebellar cortical degeneration in American Staffordshire Terriers have been described in several previous reports.10–15 MRI has shown subjective evidence of changes in cerebellar size, including increased fluid signal between the folia of the cerebellum and an enlarged 4th ventricle.13

The same subjective changes in cerebellar size were found in this study, although it was often impossible to clearly differentiate between affected and unaffected dogs (see Fig 7).

image

Figure 7.  (A) T2-weighted midsagittal image of a normal brain of an American Staffordshire Terrier. (B) T2-weighted midsagittal magnetic resonance (MR) image of the brain of an American Staffordshire Terrier severely affected by cerebellar cortical degeneration. There appears to be increased cerebrospinal fluid around the cerebellum, especially in the region of the 4th ventricle and between the folia. (C) T2-weighted midsagittal MR image of the brain of an American Staffordshire Terrier mildly affected by cerebellar cortical degeneration. It is difficult to decide whether the size of cerebellum is normal or abnormal.

Download figure to PowerPoint

MRI morphometry as performed in our study is a way to distinguish between American Staffordshire Terriers with or without cerebellar cortical degeneration.

Both 512 × 512 and 256 × 256 matrices were used initially for measurements. The 512 × 512 matrix was easier to use and appeared to yield more precise results; however, it was not always available in the control group. Because results from the 256 × 256 and 512 × 512 matrices cannot be compared directly, the 256 × 256 matrices were the only ones used for this study.

The fluid around the cerebellum and between the folia on MR images was measured by means of the cerebellar CSF index. Despite the fact that these areas seemed to be subjectively enlarged in 10/14 of the affected dogs, there was no statistically significant difference between the CSF index of the cerebellum plus CSF in affected and normal dogs. However, there was a statistically significant difference in the CSF index of the cerebellum without surrounding CSF on comparing both groups. One possible explanation for this observation is that the CSF index was defined incorrectly, because the threshold of the first 15 gray shades was chosen empirically. Some of the measured voxels may have contained brain parenchyma as well as CSF. These “mixed gray voxels” might have made the method less accurate. Therefore, in view of these uncertainties, we recommend the use of relative cerebellar size or relative CSF space to distinguish between normal and abnormal cerebella.

Previous studies have reported a reduction in cerebellar mass in affected dogs on gross pathologic examination: the cerebellum of normal dogs was 12% of the total weight of the brain versus 5–7% in affected dogs.12,13,15 In our study, the ratio of the cerebellum to the entire brain was calculated based on MR images. In dogs with presumptive cerebellar cortical degeneration, the brain : cerebellar ratio was significantly lower in affected dogs (median, 12.4%) than in normal dogs (median, 14.9%).

The calculated relative cerebellar size may be larger than the results obtained by gross pathologic examination owing to the fact that only 1 representative slice of brain was measured. In addition, volume may not directly correlate with mass. Measurements using 3D reconstruction may result in a more accurate estimate of the volume of the brain or parts of the brain. However, in our opinion, the methodical error involved in manual calculation of brain volumes would be the same or even larger. The method of measuring 1 representative slice as described here is reasonable, because degenerative changes of the cerebellum are nearly evenly distributed in cerebellar cortical degeneration. Inaccuracies in measurement may occur if the correct midline sagittal image is not used or if there are errors in manually circumscribing brain areas. In particular, it can be difficult to find the correct margin in the region of the optic chiasm and pituitary gland, and this region may have been one of the greatest sources of error. However, low intra- and interobserver variability indicates the reliability of results measuring affected and normal dogs.

There was a statistically significant difference in weight between normal and affected dogs. We believe that this difference had no diagnostically relevant effect on cerebellar size measurements because affected dogs were on average larger than the controls. There was a large discrepancy in sex in affected dogs, with a ratio of 1 female to 3.7 males, although there was no statistical difference in sex from control dogs. This result is most likely associated with the small case numbers. In a previous study of 63 dogs, both sexes were affected, with a ratio of 1 female to 1.3 males.13

The possibility that control dogs were subclinically affected cannot be ruled out. Cerebellar measurements of 1 unaffected dog were in the abnormal range. At the time of writing (20 months later), this dog has not shown any clinical signs.

Because all but 2 dogs of this study were still alive and necropsy was not permitted in the 2 euthanized animals, histopathologic examination of the CNS was not performed and case definition relied on clinical and imaging findings.

Because tumors, infarcts, traumatic lesions, and malformations could be ruled out on MRI examination, inflammatory disease should be considered as a remaining alternative to degeneration as a cause of the observed cerebellar signs. A recent report described neosporosis as a cause of slowly developing cerebellar signs.41 In MRI, the dogs with neosporosis showed moderate to marked reduction in the size of the cerebellum and mild diffuse contrast enhancement of the thickened overlying meninges.41 Because such changes were not found in our MRI images, neosporosis was considered very unlikely as a cause of the cerebellar disease in the dogs of this study. Moreover, there was no other evidence of an inflammatory brain disease. The cell counts of the CSF were normal, but the total protein concentration was slightly increased in 8/14 affected dogs. This albumino-cytologic dissociation is consistent with previous reports of American Staffordshire Terriers with cerebellar cortical degeneration.12,13

One of the affected dogs in this study had cystinuria. Cases of cystinuria with clinical signs of cerebellar atrophy have been described in humans, and a relationship between cystinuria and cerebellar disease has been discussed.42,43 Cystinuria has been reported in many breeds of dogs. Neurologic signs have not been observed in any of those.44,45

This disease can only be definitively diagnosed with histopathology, which was not available. However, breed, age of onset, clinical signs, progression of the disease, and exclusion of differential diagnoses in our affected dogs are highly consistent with cerebellar cortical degeneration. Therefore, it is reasonable to assume that our case material was representative of cerebellar abiotrophy in American Staffordshire Terriers.

Conclusion

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Footnotes
  8. Acknowledgments
  9. References

An autosomal recessive mode of inheritance has been suggested but not yet proven in the cerebellar cortical degeneration of American Staffordshire Terriers.13 A genetic test that could identify carrier and affected dogs is needed.13 Until such a test is available, computer-assisted MRI analysis may be helpful in the antemortem presumptive diagnosis of cerebellar abiotrophy not only in this particular breed but also in many others in which this disease occurs. We have shown in the present study that relative cerebellar size and relative CSF space based on computer-assisted MRI morphometry can be used to differentiate between normal American Staffordshire Terriers and those with cerebellar cortical degeneration. These measurements constitute a reliable test for the antemortem diagnosis of cerebellar cortical degeneration in American Staffordshire Terriers. It remains to be determined as to whether this method can be used in other breeds to detect dogs with the same pathology.

Footnotes

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Footnotes
  8. Acknowledgments
  9. References

aNicolet Biomedical, Division of VIASYS Healthcare, Höchberg, Germany

bNT Compact Plus; Philips Medical Systems, Amsterdam, The Netherlands

cSchering, Berlin, Germany

dScion Corporation, Frederick, MD

Acknowledgments

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Footnotes
  8. Acknowledgments
  9. References

We acknowledge Dr Maren März and Ines Merseburger (University of Leipzig) for their technical support. We also express special thanks to Dr Sophie Petersen (UC Davis) for her linguistic support.

References

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Footnotes
  8. Acknowledgments
  9. References