Study 1
The first set of data was derived from a cohort of 70 newborns born at two hospitals in Denver, Colorado [
21].
HPRT M
f
were determined on cord blood samples using the clonogenic assay. Smoking status was initially separated into three categories: active smokers, active smokers who quit after confirmation of pregnancy (quitters), and non-smokers. In a secondary analysis, non-smokers were then broken out into those likely to have ongoing exposure to environmental tobacco smoke and those actively avoiding passive or secondary smoking exposure. These categories were based on the paradigm that the genotoxic effects of secondary smoke should be intermediate between those of non-smokers with no secondary exposure and active smokers, and their exposure levels and therefore induced mutation frequencies were expected to be closer to those of the non-exposed population than those of the active smokers. Thus, the
HPRT M
f
of the total non-smoking population (with and without evidence of passive exposure, although clearly skewed towards the passively exposed population, which contributed 20 of the 28 subjects in this category) was used as the basis of comparison for the smoking and quitting groups, and neither was found to have a significant induction of
HPRT mutants. In the present investigation, we have combined these primary and secondary analyses, using only the non-smokers with no evidence of secondary tobacco smoke exposure as the basis of comparison (this was the only category in the original study reported as having a significantly lower M
f
). These data are summarized in Table
1, panel
a.
Table 1
HPRT M
f
in newborns with and without exposure to tobacco smoke metabolites in utero
| |
HPRT M
f
(× 10-6) | | |
maternal exposure | N | mean ± SD | median | range | P1,2
| P2,3
|
a) data from Manchester et al [21] |
unexposed | 18 | 0.76 ± 0.50 | 0.61 | 0.14 – 1.9 | | |
passive only | 20 | 1.60 ± 1.43 | 1.35 | 0.30 – 5.3 | 0.021 | |
quit during pregnancy4
| 4 | 1.85 ± 1.16 | 1.60 | 0.35 – 3.2 | 0.004 | |
smoked throughout | 27 | 1.36 ± 0.99 | 0.98 | 0.28 – 3.5 | 0.019 | 0.012 |
b) data from Finette et al [22,23] |
unexposed | 26 | 0.72 ± 0.53 | 0.52 | 0.05 – 1.9 | | |
passive only5
| 22 | 1.18 ± 1.28 | 0.67 | 0.10 – 5.1 | 0.14 | |
quit during pregnancy | 8 | 0.79 ± 0.46 | 0.69 | 0.18 – 1.8 | 0.51 | |
smoked throughout | 12 | 0.71 ± 0.51 | 0.56 | 0.14 – 1.8 | 0.73 | 0.47 |
c) pooled data
|
unexposed | 44 | 0.73 ± 0.51 | 0.60 | 0.05 – 1.9 | | |
passive only5
| 42 | 1.38 ± 1.36 | 0.87 | 0.10 – 5.3 | 0.006 | |
quit during pregnancy4
| 12 | 1.27 ± 0.93 | 0.91 | 0.18 – 3.2 | 0.014 | |
smoked throughout | 39 | 1.16 ± 0.91 | 0.87 | 0.14 – 3.5 | 0.007 | 0.007 |
One outlier was identified in this data set, defined as an individual with an HPRT M
f
greater than 3 standard deviations higher than the mean for the population. This individual was born to a women who quit smoking during her pregnancy, and had an M
f
of 14.7 × 10-6, 10-fold higher than the mean of the entire population, 15-fold higher than the median value.
By breaking the "non-smoking" population into those with and without evidence of environmental tobacco smoke exposure, and using those with no evidence of such passive exposure as baseline, we now show both significant effects of tobacco smoke exposure overall in this population, as well as significant inductions in all three exposed categories. Moreover, the HPRT M
f
in the three exposed populations were not significantly different from one another (pairwise P values ranged from 0.09 to 0.75), unless the outlier was included in the analysis, in which case the "quitters" were significantly higher than all three of the other groups.
Study 2
The second set of data is derived from two related publications [
22,
23], that were designed as follow-ups to those of McGinniss
et al [
11] and Manchester
et al [
21]. In the former study, newborns in Burlington, Vermont, demonstrated no detectable effect of maternal active smoking on cord blood
HPRT M
f
, although passive exposure was not considered, and therefore might have been a confounding factor. Subjects for the follow-up studies were recruited from the same university-affiliated hospital in Vermont, and had similar
HPRT M
f
. In this study, passive exposure was assessed by interview and ongoing tobacco smoke exposure was estimated by measurement of cotinine levels in the cord blood. In general, these cotinine measurements confirmed the smoking exposure assignments based on the interviews. These data are summarized in Table
1, panel
b.
This population also contained an outlier, this one in the passively exposed group, with an M
f
of 45.3 × 10-6, 30-fold higher than the average of the population and 70-fold higher than the median value.
Finette
et al [
22,
23] reported on two different but overlapping subsets of these data, and found no evidence of any type of tobacco smoke exposure affecting
HPRT M
f
. Analysis of the entire data set, as summarized in Table
1, panel
b, confirms these results. Indeed, only if the extreme outlier is included in the analysis is any comparison even close to significant (unexposed vs. passively exposed,
P = 0.063).
Pooled data
These two studies examined similarly sized populations, and both failed initially to demonstrate an influence of tobacco smoke exposure on newborn
HPRT M
f
. These two sets of subjects are geographically distinct, and may differ in other ways, but this cannot be assessed from the published data. No other factor was reported to have significantly affected newborn
HPRT M
f
in either study, however. The M
f
of the unexposed populations from the two studies are not significantly different from one another (
P = 0.48), but the combined exposed population from the Colorado population is 1.5-fold higher than the equivalent population from the Vermont studies, which is significant (P < 0.001). This difference has been attributed to maternal environmental and socioeconomic factors, but nothing has been proven. The distribution of samples between the four smoking exposure categories differs significantly between the two studies (
P = 0.044), with the major disparity being the proportion of active smokers (39% in the Manchester
et al study [
21] vs. 17% in the Finette
et al studies [
22,
23],
P = 0.006). It is tempting to invoke this difference in population distribution to explain the higher overall M
f
of the Colorado population (mean 1.32 × 10
-6, median 0.96 × 10
-6, range 0.14–5.3 × 10
-6) than the population from Vermont (mean 0.89 × 10
-6, median 0.64 × 10
-6, range 0.05–5.1 × 10
-6) (
P = 0.006) when the outliers are not included in the analysis. However, the proportion of all tobacco-exposed individuals (including active smokers, quitters and passively exposed mothers) is not significantly different between the two populations (74% vs. 62%,
P = 0.66). The
HPRT M
f
for the pooled data set are given in Table
1, panel
c.
Analysis of the pooled data from these two studies essentially reiterates the results of the reanalysis of the data from Manchester
et al [
21] discussed above: all three groups of tobacco exposed newborns have
HPRT M
f
significantly higher than the unexposed group, and there is no significant difference between the levels of induced mutation amongst the three exposed populations. These data indicate that tobacco smoke exposure
in utero does induce detectable
HPRT mutants in the fetus, and that passive maternal exposure has a similar teratogenic effect as active maternal smoking, a finding that is not unprecedented [
30].
HPRTmolecular spectra
Despite the lack of evidence for a mutagenic effect of tobacco smoke in their newborn cord bloods, Finette
et al [
23] nevertheless examined the molecular spectrum of
HPRT mutants in two of their subpopulations, those without evidence of any maternal tobacco smoke exposure and those with passive exposure only. The mutations were classified as a) small, intragenic changes, b) gene rearrangements or deletions, or c) exon 2/3 deletions characteristic of illegitimate VDJ recombination (especially in newborn populations [
23,
31,
32]). These data, summarized in Table
2, panel
a, suggest a shift in the spectrum of the exposed population to significantly higher proportions of both small mutations and deletions attributable to VDJ recombination. Since there was no overall increase in
HPRT M
f
in this population, however, the exposed population also had a compensatory significantly lower proportion of non-VDJ mediated deletions and rearrangements, suggesting a protective effect of tobacco smoke exposure on these types of mutagenic events. We have found that the need to invoke such a protective effect is reduced if these data are put in perspective of the related studies mentioned above [
21,
23] and if mutation frequencies are used to normalize the distributions.
Table 2
HPRT mutational spectra in newborns with and without exposure to tobacco smoke metabolites in utero
a) distribution of mutant clones
|
maternal exposure | study | total independent mutants | small mutations (%) | deletions, rearrangements (%) | VDJ recombinant deletions (%) | P1
|
unexposed | Finette et al [23] | 30 | 10 (33) | 14 (47) | 6 (20) | |
mixed | McGinniss et al [11] | 41 | 7 (17) | 14 (34) | 20 (49) | 0.039 |
mixed | Manchester et al [21] | 38 | 13 (34) | 16 (42) | 9 (24) | 0.91 |
passively exposed | Finette et al [23] | 35 | 17 (49) | 6 (17) | 12 (34) | 0.036 |
b) mutation frequencies for three classes of mutants based on individual studies
|
maternal exposure | study | overall mean M
f
± SD (× 10-6) | small mutations M
f
(× 10-6) | deletions, rearrangements M
f
(× 10-6) | VDJ recombinant deletions M
f
(× 10-6) | P2
|
unexposed | Finette et al [22,23] | 0.72 ± 0.53 | 0.24 | 0.34 | 0.14 | |
mixed | McGinniss et al [11] | 0.64 ± 0.40 | 0.11 | 0.22 | 0.31 | 0.003 |
mixed | Manchester et al [21] | 1.32 ± 1.093
| 0.45 | 0.56 | 0.31 | < 0.001 |
passively exposed | Finette et al [22,23] | 1.18 ± 1.284
| 0.57 | 0.20 | 0.40 | 0.002 |
c) mutation frequencies for three classes of mutants based on pooled data
5
|
unexposed | Finette et al [22,23] | 0.73 ± 0.51 | 0.24 | 0.34 | 0.15 | |
mixed | McGinniss et al [11] | 0.99 ± 0.95 | 0.17 | 0.34 | 0.48 | 0.008 |
mixed | Manchester et al [21] | 0.99 ± 0.953
| 0.34 | 0.42 | 0.24 | 0.037 |
passively exposed | Finette et al [22,23] | 1.38 ± 1.364
| 0.67 | 0.24 | 0.47 | < 0.001 |
Summaries of the
HPRT mutational spectra generated from the earlier analysis of a newborn population from Vermont [
29] and the Colorado UHD population [
21] are also presented in Table
2
a. These spectra were generated from mutants without regard for their potential tobacco smoke exposure, so are classified as "mixed". The population of McGinniss
et al [
11,
29] contained only 20% active smokers, however, while the incidence of passive exposure of the remaining 80% of the population was not estimated. 45% of the population reported in Manchester
et al [
21] actively smoked throughout pregnancy, and another 33% reported ongoing exposure to secondary tobacco smoke; only 13% could be considered unexposed. These data might therefore be expected to begin to show the effects of both active cigarette smoking and passive secondary exposure on cord blood
HPRT mutagenesis, although the power would not be as great as if they were derived only from defined exposed groups.
In adults, active tobacco smoke exposure has been found to increase the frequency and proportion of small base changes at the
HPRT gene [
33], consistent with the known mechanisms of tobacco smoke mutagens and the types of mutations found in oncogenes in smoking-associated cancers [
34,
35]. Illegitimate VDJ recombination is a mechanism of mutagenesis unique to T- and B-lymphocytes, and is implicated in many of the molecular events associated with leukemia and lymphoma [
36,
37]. The human
HPRT gene contains cryptic sites for this DNA splicing event, resulting in the deletion of exons 2 and 3 [
31,
38], and the occurrence of this type of
HPRT mutation seems to be associated with the incidence of acute lymphocytic leukemia in children [
39]. Elevated levels of illegitimate VDJ recombination have been found in workers occupationally exposed to pesticides and herbicides [
40], especially 2,4-dichlorophenoxyacetic acid [
41,
42] and in cancer patients undergoing chemotherapy [
43], particularly with the DNA topoisomerase inhibiting agent etoposide [
44,
45].
Overall, the two newborn populations from Vermont had indistinguishable
HPRT M
f
(
P = 0.50), and the total population data from McGinniss
et al [
11] was also consistent with the unexposed group reported by Finette
et al [
22,
23] (
P = 0.51), but the passively exposed group had a significantly higher level of mutation (
P = 0.013). The distribution of mutants among the three mechanistic classes differed significantly in both cases, however, with the mutants from McGinniss
et al [
29] exhibiting less small mutation and more VDJ recombination-mediated deletion than either group from Finette
et al [
22,
23]. The Colorado population had a significantly higher M
f
than the unexposed subset of the second Vermont population (
P = 0.008), but a very similar distribution of mutants (
P = 0.91). On the other hand, the Colorado population had a similar mutation frequency as the passively exposed subpopulation from this study (
P = 0.55), but a somewhat different mutant distribution (
P = 0.068). We believe that these mutational comparisons are of little use unless both frequency and distribution are taken into account at the same time. In Table
2, panels
b and
c, the overall
HPRT M
f
from these individual studies and subpopulations (panel
b), or the M
f
generated from our pooled analysis (panel
c), are used to calculate the frequency of each type of mutant in each population, as was done in Manchester
et al [
21] and Finette
et al [
23].
Expressing the mutational classes as frequencies makes it easier to see the general trends in these studies and their inconsistencies. The frequency of VDJ recombination-mediated deletions is now increased in all exposed populations, and the results from the mixed tobacco smoke populations are consistent with an intermediate level of exposure (remember that even though these populations should contain maternal active smoking exposures, and quitters, the meta-analysis indicated that these should have induced M
f
similar to the passively exposed population). The differences in the frequencies of non-VDJ recombination-mediated deletions and rearrangements are diminished under these circumstances. The increase in frequency of small mutations observed in the passively exposed population of Finette
et al [
22,
23] is difficult to rationalize with the low levels found in the McGinniss
et al [
29] study, however, the induction in the Manchester
et al [
21] population is again intermediate between those of the two subpopulations from Finette
et al [
22,
23]. Significantly, none of the decreases observed in the frequencies of mutational subclasses from the unexposed population of Finette
et al [
23] were themselves statistically significant.