Carcinogenesis vol.18 no.5 pp.1013–1020, 1997
Carcinogenesis induced by UVA (365-nm) radiation: the dose–time
dependence of tumor formation in hairless mice
Annemarie de Laat, Jan C.van der Leun and Frank R.de
Gruijl1
Department of Dermatology, University Hospital Utrecht, Utrecht
University, PO Box 85500, 3508 GA Utrecht, The Netherlands
1To
whom correspondence should be addressed
Although ultraviolet B (UVB wavelengths 280–315 nm)
dominates the carcinogenic effect of sunlight, ultraviolet A
(UVA 315–400 nm) is estimated to contribute 10–20% to
the carcinogenic dose; a substantial background that is not
affected by a depletion of the ozone layer. Furthermore,
certain high-power modern tanning lamps emit mainly
long wave UVA (UVA1; 340–400 nm). For a proper risk
estimate of UVA exposure its carcinogenicity relative to
that of UVB exposure needs to be determined more accurately. To this end we determined the dose–time relationship
for skin tumor induction in hairless mice that were irradiated daily with custom-made Philips 365-nm sources.
Irradiation of the group exposed to the highest of the four
daily doses (430, 240, 140 and 75 kJ/m2) had to be
discontinued because severe scratching set in after 3 months
(no tumors). In the lower dose-groups the prevalence curves
for skin carcinomas (percentage of tumor-bearing mice
versus logarithm of time) ran virtually parallel, and were
similar to those found with daily UVB exposure. However,
the relationship between the daily dose (D) and the median
tumor induction time (t50) appeared to differ: with UVB
we found that t50 Dr 5 constant, with r 5 0.6, whereas
with UVA1 we found r µ 0.4. This would imply that 365nm carcinogenesis shows less of a dose-dependency than
UVB carcinogenesis, and that 365-nm radiation becomes
more carcinogenic, relative to UVB, as the daily doses
are lowered. This relative shift at low doses complicates
extrapolation of UVB to UVA risks in humans. Based on
the t50 from the lowest dose-group we found that the
carcinogenicity at 365 nm (per J/m2) is 0.9310–4 times that
at 293 nm, the wavelength of maximum carcinogenicity in
hairless mice. This result for 365-nm carcinogenicity falls
well within the margins of error of the wavelength dependency that was estimated earlier from experiments with
broadband UV sources.
Introduction
Ultraviolet A (UVA*) radiation (subdivided into UVA2: 315–
340 nm, and UVA1: 340–400 nm) is the main part of the solar
UV radiation reaching the earth’s surface. Its contribution to the
erythemogenic and melanogenic potential of natural sunlight is,
*Abbreviations: ultraviolet A, UVA 315–400 nm; ultraviolet 2, UVA2 315–
340 nm; ultraviolet A1, UVA1 340–400 nm; ultraviolet B, UVB 280–315 nm;
SCUP, Skin Cancer Utrecht Philadelphia; SCUPm, SCUP action spectrum for
mice; IR, infrared; D, daily dose; t50, median tumor induction time(s), the
time until 50% of the mice in a group bear carcinomas or precursors; SCC,
squamous cell carcinoma(s); σ, standard deviation; p, power of time.
© Oxford University Press
however, relatively small compared with that of ultraviolet B
(UVB 280–315 nm). Because UVA radiation has been supposed
to be harmless while it can tan at high dosages, certain modern
tanning devices are composed of high power UVA sources
with little emission in the UVB. Sunscreens were mainly
designed to protect against UVB and offered little protection
against UVA radiation. However, animal experiments have
proven that chronic UVA irradiation can induce skin damage,
including photo-aging (degenerative changes in dermal connective tissue) (1) and skin cancer (2–5).
The first report on skin cancer in mice by UVA alone dates
from 1982 (2). To rule out the possible involvement of small
amounts of UVB, additional experiments were carried out by
Van Weelden et al. (3) and Sterenborg and Van der Leun (4)
with more rigorously filtered UV sources (predominantly
giving UVA1 radiation). Both studies confirmed the carcinogenicity of UVA radiation. The tumor induction kinetics and
skin reactions (e.g. scratching before tumor development,
hyperkeratosis) differed from those seen after UVB irradiation.
Also a high percentage of benign papillomas was noted by
van Weelden et al. [reported by Kelfkens et al. (5)] in the
cohort of first appearing tumors. Kelfkens et al. (5) found in
an experiment with a filtered UVA source with appreciable
output in the UVA2 wavelength that, when papillomas were
excluded from analyses, tumor development with UVA exposure showed the same kinetics as with UVB exposure. Thus
the carcinogenicity of UVA and UVB radiation could be
conveniently compared.
Berg et al. (6), using the same filtered UVA2 source as
Kelfkens et al. (5), in combination with an FS40 broadband
UVB source, found that UVA and UVB exposures added up
arithmetically in the induction of skin carcinomas in hairless
mice: the radiant energies of different wavelengths have to be
weighted and added to arrive at an overall effective UV dose
(photo-addition). This is fundamental to a consistent derivation
of the wavelength dependency (action spectrum) of tumor
induction from experiments with broadband UV sources. Such
an action spectrum has been generated from experiments with
14 different UV sources, which was carried out by the Skin
and Cancer Hospital in Philadelphia and by the Department
of Dermatology of Utrecht University and is called the Skin
Cancer Utrecht–Philadelphia (SCUP) action spectrum for mice
(SCUPm) (7). The carcinogenicity peaks at 293 nm and is
about a factor of 104 lower for wavelengths .340 nm.
Assuming that the proliferative basal cells in the epidermis
are the target cells, the data can be corrected for differences
in epidermal transmission between mouse and human (8).
Thus, the UVA contribution to the carcinogenicity of sunlight
for humans is estimated to be around 10–20% (9); this is a
substantial background to the UVB-induced carcinogenesis,
and only the latter is increased by ozone depletion. The
uncertainty in the SCUP action spectra is relatively large in
the UVA1 wavelength region, due to a lack of data acquired
with narrow band irradiation, and due to a lack of information
1013
A.de Laat, J.C.van der Leun and F.R.de Gruijl
on the dose–time relationship in this wavelength region. We
performed the present experiment to fill in these gaps in
information in the long wave UVA region. A good definition
of the action spectrum in the UVA1 is important for the mouse
to human extrapolation for an assessment of risk from ozone
depletion and from tanning with high powered UVA1 sources.
Materials and methods
Animals and their maintenance
In this experiment albino hairless mice (SKH:HR1; Charles River) were used.
A total of 128 animals (64 female, 64 male) with ages ranging from 11 to 15
weeks were randomly divided into four groups of 24 animals, and two groups
of 16 animals. No significant differences were found in the responses of male
and female mice. Animals were housed separately in steel cages subdivided
into eight compartments with wire mesh tops. Standard mouse chow (Hope
Farms RMH-B) and tap water were available ad libitum. Mice were kept in
an air-conditioned room at an ambient temperature of about 24°C, which rose
to 26–28°C during UVA exposure due to heat dissipation in the electrical
circuits. The room was illuminated with yellow fluorescent tubes (Philips
TL40W/16) in a 12-h cycle (switched on at 6:00, and off at 18:00). These
lamps do not emit any measurable UV radiation. No daylight entered the
room. Permission for the experiments was obtained from an independent
ethical committee, in accordance with Dutch legislation and EU guidelines.
Irradiation set-up and dose-groups
The mice were irradiated from above, while being allowed to move freely in
the cage. A UVA source was mounted above each cage, and this source was
switched on and off automatically by an electronic clock. The outlines of the
irradiation setup are depicted in Figure 1. The twelve UVA sources were high
pressure mercury lamps (2 kW, 1 kW or 400 W) developed by Philips and
specially low- and high-cut filtered to obtain radiation from the spectral line
at 365 nm. The filter system consisted of a visible light 3-mm MUG-2 glass
filter (‘Woods glass’), a ‘330 glass’ (manufactured by Philips for lamp
envelopes) to filter out the UVB and shorter UVA radiation below 340 nm,
and a transparent plexiglass (gs 2058, Röhm Kunststoffe, Darmstadt, Germany)
container filled with water (optical path length 50 mm) to filter out the infrared
(IR) radiation. The visible light and IR were filtered out to minimize heat
effects. During the irradiation period the temperature in the cages never
exceeded 28°C. The spectrum of the filtered lamp is shown in Figure 2. Due
to limitation in the electrical power supply the four UVA dose-groups were
irradiated in three shifts at different times. The daily irradiation time was 2 h
and the mice were exposed from 7:00 till 9:00, or 11:30 till 13:30 or 16:00
till 18:00. Mice in group 1 received a daily (dorsal) dose (D) of 430 kJ/m2
(250–400 nm). This daily dose is about 60% of the dose that causes acute
effects such as edema and erythema. Mice in groups 2, 3 and 4 received 240,
140 and 75 kJ/m2 per day, respectively. Due to the geometry of the set-up,
there was some variation in the irradiances over the cage. The cages were
rotated 180° every week. At corner compartments of the cages the irradiance
dropped by ~10% (maximally 20%); the lower exposures in corner compartments was not compensated for by the rotation. We could not detect significant
differences between responses of the animals in corner positions or center
positions. One group of 16 mice served as unirradiated controls, and were
exposed only to the room light from the non-UV emitting fluorescent tubes.
Another group of 16 mice were used as positive controls for UVB exposure
and were irradiated with North American Philips F40 sunlamps (see reference
10 for lamp details); a total of 920 J/m2 (250–400 nm) was given daily from
11:30 till 13:30 to the dorsum of the animals. The F40 dose was routinely
monitored with a Robertson–Berger meter (Solar Light Company, Philadelphia,
PA) in combination with a micro-ampere meter (Keithley Inc., Cleveland,
OH) calibrated against an Optronics 742 photospectrometer (Optronic Laboratories Inc., Orlando, FL). The UVA dose was determined with an UVX meter
and an UVA sensor of Waldmann (Waldmann, Schwenningen, FRG) in
combination with the micro-ampere meter, calibrated against the Optronics
742 photospectrometer and against the Kipp E11 thermopile. The exposure
rates were measured every fortnight. The radiant output of a lamp was
controlled with an electronic dimmer circuit, which could be re-adjusted
if necessary.
Animal observations and data analysis
The UVB-exposed animals were checked on a weekly basis, and the UVA
exposed animals every fortnight. All deviations from normal skin appearance
(tumors, change of color, hyperkeratosis, scratch marks, wrinkling etc.)
were recorded. A (small) skin lesion was classified as a tumor when it
rose up from the surrounding skin surface and/or was clearly vascularized.
Cysts could be clearly distinguished. Animal mortality was also recorded;
1014
Fig. 1. Scaled schematic representation of a cross-section of the exposure
configuration, with (A) UVA lamp, (B) MUG 2 glass filter, (C) ‘330 glass’
filter and (D) plexiglass container filled with water.
Fig. 2. Relative spectral energy output of the filtered Philips UVA lamp.
after 400 days of irradiation mortality was inexplicably high in one cage
of the lowest dose-group. If a tumor was spotted on an animal for the
first time, the observation had to be confirmed in the next check-up or
otherwise the observation was disregarded. Tumors were subdivided
according to diameter into: ,1 mm, ù1 mm, ù2 mm, ù4 mm. During
the experiment we distinguished between evident papillomas (pedunculated,
protruding tumors with a ‘cauliflower-like’ surface) and other tumors by
visual inspection. Not all tumors were first scored at a sub-1-mm level
because sometimes they either grew too fast, or developed into scaly/
leathery (hyperkeratotic) skin patches, which hampered early detection.
These patches were .1 mm and did not classify as tumors. Only tumors
that arose in unscarred skin were registered for induction times (i.e. tumors
that could have been promoted by scarring were disregarded). If a tumor
UVA carcinogenesis versus UVB carcinogenesis
was first seen at t 5 ti, and t 5 ti–1 was the previous check-up time,
then the tumor induction time was defined as (ti 1 ti–1)/2. Thus we get
induction times for papillomas and non-papillomas with diameters ,1 mm,
1 mm, 2 mm and 4 mm for each mouse. One statistically significant
outlyer (P , 0.005) for non-papilloma tumor induction time (171 days)
in the 140-kJ/m2 dose-group was rejected from the data set. Mice with a
large total tumor mass (0.5 to 1 cm3), and/or with a bad condition were
physically removed from the experiment. Three thresholds for scratch
marks were tested to ascertain the interference of severe skin damage with
tumor appearance: scarring on more than 10, 25 and 50% of the dorsal
skin, and at 50% occurrence the animals were physically removed from
the experiment.
Tumor response of a group is described by the prevalence, i.e. the percentage
of tumor-bearing mice. Graphical representation of the prevalence versus time
is based on an actuarial method that corrects for censoring by deaths without
tumors, described by Kaplan and Meier (11). This procedure, adapted to
carcinogenesis by Peto et al. (12), computes the chance of tumor-free survival.
The death-corrected tumor prevalence is then given by one minus this chance.
As was established from earlier experiments, the prevalence data can be
fitted with log-normal or Weibull distributions using a maximum likelihood
method (13,14). This method, too, includes a correction for animals that died
or were removed from the experiment before contracting any tumors. The
log-normal distribution gives mainly a statistical description of lapse time till
a first tumor appears on an individual animal. The log-normal fit gives values
for the median tumor induction time (t50), the time at which 50% prevalence
is reached, and the standard deviation (σ) in the natural logarithm of the
tumor induction time around the median. The Weibull distribution has
a somewhat stronger theoretical foundation in multistage carcinogenesis
(14–17), see the Appendix.
Results
Scratching
Already after 50 days of irradiation some mice in the highest
dose-group started to scratch and severely damaged their dorsal
skin and sometimes their ears. The incidence of mice bearing
scratch lesions increased rapidly. After ~90 days, 50% of the
mice in the highest dose-group had developed scratch marks.
Because of the known promoting effect of scarification on
tumor induction (18), tumors arising on previously damaged
sites cannot be considered as purely UVA induced, and are,
therefore, excluded from the analyses. In the highest dosegroup only one mouse developed a tumor on unscarred skin
on the head. Another five mice in this dose-group developed
at least one papilloma in scar tissue. After 165 days of
irradiation all mice in this dose-group were sacrificed because
of the severity of skin damage. No tumor induction information
could therefore be extracted from this dose-group. In the other
dose-groups scratching also occurred, but it set in later in
relation to tumor occurrence, thus allowing tumor induction
times to be measured. At 240 kJ/m2 per day, 14 mice had to
be removed because of extensive scarification (.50% dorsal
skin involved) prior to the development of any tumors.
Lowering the threshold for discarding an animal from 50 to
25% of scarred dorsal skin decreased the estimated median
induction time, t50, of 1-mm tumors from 418 to 379 days.
Lowering the threshold from 25 to 10% did not affect the t50
any further, and a threshold at 25% was, therefore, deemed
appropriate. There was no discernible interference from
scratching with tumor induction at lower daily doses, i.e. in
the groups with 140 and 75 kJ/m2 per day.
The time at which the mice start to damage their skin as
well as the severity of the damage was clearly dose-dependent.
A more detailed analysis of the induction of scratch marks
[dose–time dependency and an attempt to prevent it by
suppression of mast cell activity by Ketanserin (Janssen
Research Foundation, Beerse, Belgium)] will be published
elsewhere.
Skin changes
Besides the above mentioned scratch lesions other dosedependent skin changes were observed in the UVA-irradiated
mice. Most of the time, a reddening of dorsal skin preceded
the actual damaging of the dorsal skin, which was due to
scratching. Other skin changes observed were a yellowish
coloration (loss of pink color), a scaly dryness, skin thickening,
leathery appearance, sagging and wrinkling of the irradiated
skin. Flaky skin, especially just behind neck region (most
likely due to extensive hyperkeratosis) was seen in three mice
of the 240-kJ/m2 dose-group and two mice of the 140-kJ/m2
dose-group.
Tumor types
Tumors were subdivided according to macroscopic appearance
in papillomas and non-papillomas. Post-mortem samples of
tumors were taken from mice that had to be removed from
the experiment. A total of 60 formalin-fixed samples were sent
in for histopathological examination and the results of the
classification are shown in Table I. The classification of
tumors by macroscopic appearance did not deviate from the
histological examination. All but one of the 55 macroscopically
identified non-papillomas were indeed classified as squamous
cell carcinomas (SCC) or precursor lesions (actinic keratosis
or carcinoma in situ/Bowenoid tumor) and only one of the five
macroscopically identified papillomas showed some features of
neoplastic progression, and was histologically identified as a
papillary growing SCC. With increasing tumor diameter the
SCC outnumbered actinic keratoses, which were the predominant lesions at small tumor diameters (,2 mm).
Dose–time dependence of tumor induction
In the course of the experiment (620 days) animals in the 240-,
140- and 75-kJ/m2 dose-groups contracted multiple skin tumors
on scar-free, intact skin. However, the prevalence of 1-mm
papillomas in these groups was very low, not exceeding 25%,
and no curve could be reliably fitted to this scanty data. Kaplan–
Meier plots for the prevalence of papillomas are shown in
Figure 3.
Log-normal and Weibull distributions were fitted to prevalences of non-papillomas in the UVA groups and the control
UVB group. In Table II we give the parameter t50, and the
standard deviation σ of the log-normal distribution for the
different tumor categories. In Table III we give the parameter
t1 and the power of time (p) of the Weibull distribution (see
Appendix) for the same data. In the 140-kJ/m2 dose-group 16
out of 23 mice contracted one or multiple skin tumors. The
standard errors of the parameters fitted for this group are
smaller than those for the other two dose-groups in which
only seven out of 24 mice contracted a tumor. Evidently, with
fewer data on tumor induction the estimates are likely to be
less accurate. In the 75-kJ/m2 dose-group the total of tumor
contracting mice was limited by (natural) death before the
appearance of a tumor. In the 240-kJ/m2 dose-group 14 mice
were discarded from the data analysis (at a time point where
25% of the dorsal skin was scarred), thus limiting the number
of animals on which tumors could be sighted.
The parameters σ and p for the UVB control group in this
experiment are very close to those found earlier for mice
irradiated with Westinghouse FS40 sunlamps (13). To compare
the doses between these experiments we have to account for
spectral differences between the earlier Westinghouse FS40
sunlamps and the North American Philips F40 sunlamps used
here; the latter emits relatively more in the short wave UVB
1015
A.de Laat, J.C.van der Leun and F.R.de Gruijl
Table I. Observed numbers of different tumor types in an aselective sample
of the collected tumors from UVA (365-nm) irradiated mice (39 of the 60
examined tumors came from mice of the 140-kJ/m2 dose-group)
Histopathological diagnosis of
macroscopic non-papillomas
Tumor diameter (mm)
,3 (na 5 16)
Squamous cell carcinoma
1
Carcinoma in situ/Bowenoid tumor 3
Actinic keratosis
11
Non-cancerous lesion
1
Histopathological diagnosis of
macroscopic papillomas
Papilloma
Papillar growing SCC
an
(n 5 1)
1
0
ù3 (na 5 39)
33
3
3
(n 5 4)
3
1
is sample size.
log scale we expect a linear relationship as was found earlier
for UVB carcinogenesis (13), i.e. t1 ~ D –r (where ~ stands
for direct proportionality). A least-square fit for the smallest
tumor diameter yielded an r-value of 0.34 6 0.09. As the
scratch marks interfered considerably with the detection of
tumors in the 240-kJ/m2 dose-group, we also calculated the
r-value solely from the two lowest dose-groups. In this case
an r-value of 0.44 6 0.11 was obtained.
The carcinogenic action spectrum
In order to produce a consistent measure of the difference in
carcinogenicity between two UV sources (e.g. UVA versus
UVB) the dose–response relationships should run parallel, i.e.
for all dose levels a single multiplicative factor applied to the
dose from the one source should give the dose from the other
source, which would evoke the same response. However, we do
not find a strict parallelism in the dose–response relationships
between the UVB (FS40 sunlamps) and long wave UVA (365
nm); for the former we earlier found r 5 0.6 and for the latter
we now find r µ 0.4. Although the latter has a considerable
error (about 0.1), there appears to be a tendency for 365-nm
radiation to become more efficient, relative to UVB radiation,
as the daily dose is lowered. This effect is illustrated in Figure
5 where the relative carcinogenicity at 365 nm measured in
the three lowest dose-groups from the present experiment is
plotted onto a graph of the earlier determined SCUP action
spectrum for carcinogenesis in hairless mice (the carcinogenicity is calculated from t50 in days and the corresponding
daily dose D in J/m2 as 23.8/(1.19 D)3(283/[t50 – 17])1.6 (7),
the factor 1.19 corrects for the estimated 19% difference in
dose-efficiency with the earlier experiments). It is clear that
these points fall close to the SCUPm action spectrum and are
well within the margins of uncertainty that were derived earlier.
Discussion
Fig. 3. Kaplan–Meier plot showing the prevalence for 1-mm papillomas
versus time on a logarithmic scale in the UVA-exposed (365 nm) groups;
(a) 240-kJ/m2 dose-group, (b) 140-kJ/m2 dose-group, (c) 75-kJ/m2 dosegroup.
part of the spectrum. Based on the SCUPm action spectrum
(7) and the difference in lamp spectra, the tumor induction
times (both t1 and t50) in the present experiment are ~11.5%
shorter than expected, implying that the daily dose was ~19%
more effective than expected. This difference in effectiveness
of the exposure could be due to a somewhat longer exposure
time per day (120 instead of 75 min) (19) or a slightly higher
susceptibility for this cohort of mice. For further comparisons,
the UVA data will be corrected for this 19% difference in dose
efficacy. In the negative control group, exposed to yellow
light, no tumors were spotted.
If we make a Kaplan–Meier plot of the tumor prevalence
versus time on a logarithmic scale, we can draw in the fitted
Weibull curve to match. The results for 1-mm tumors in the
three lowest UVA dose-groups are shown in Figure 4. Plotted
this way, the curves run parallel, shifted along the log–time
axes toward longer times for lower daily doses. If we plot
tumor induction times (t1 or t50) versus daily doses on a log–
1016
In addition to earlier studies, proving that wavelengths over
340 nm are carcinogenic in hairless mice (4,20), this study
proves more specifically that nearly monochromatic 365-nm
radiation is carcinogenic in hairless mice. The 365-nm radiation
induces the same type of skin tumors (mainly squamous cell
carcinomas and precursor lesions and some papillomas) as
broadband UVB/UVA sources.
In accordance with earlier findings with UVA-dominated
sources (5,6), there are relatively many papillomas among the
first appearing tumors, which subsequently become rapidly
outnumbered by newly arisen actinic keratoses that progress
into carcinomas. Only a minority of the papillomas progress
to carcinomas, most of the carcinomas develop from rather
flat, laterally expanding lesions, actinic keratoses, which are
usually well vascularized.
The visible skin changes observed after chronic UVA (365nm) irradiation in this study are similar to those described
earlier (21–24) and include skin changes also seen after chronic
UVB irradiation like thickening, yellow discoloration and
wrinkling, but also changes more specific for long wave
irradiation, like scratch marks before tumor development and
sagging. Part of the differences in skin changes seen between
UVB and UVA might be explained by the deeper penetration
of UVA wavelengths (25), but the unknown underlying mechanisms are likely to be different, too. The evaluation of
histological changes in the skin of the UVA (365-nm) irradiated
mice will be published elsewhere. In humans repetitive suberythemogenic UVA irradiations results in epidermal and
UVA carcinogenesis versus UVB carcinogenesis
Table II. The log-normal parameters t50 (in days) and σ, and their standard errors derived from the prevalence for non-papillomas for different tumor
diameters
Daily dose (kJ/m2)
UVA lamp
75
140
240
F40 (UVB)
0.92
Tumor diameter (mm)
,1
1
2
4
(N a 5 24)
t50
σ
(N 5 23)
t50
σ
(N 5 24)
t50
σ
(n b 5 7)
574 6 46
0.23 6 0.06
(n 5 16)
431 6 17
0.19 6 0.04
(n 5 7)
373 6 30
0.25 6 0.06
(n 5 7)
574 6 46
0.23 6 0.06
(n 5 16)
434 6 17
0.19 6 0.03
(n 5 7)
379 6 30
0.25 6 0.06
(n 5 7)
571 6 34
0.19 6 0.05
(n 5 15)
458 6 18
0.17 6 0.03
(n 5 4)c
–
(n 5 6)
593 6 42
0.19 6 0.06
(n 5 10)
535 6 27
0.19 6 0.04
(n 5 2)c
–
(N 5 16)
t50
σ
(n 5 16)
64 6 2
0.14 6 0.03
(n 5 16)
73 6 2
0.11 6 0.02
(n 5 15)
91 6 4
0.15 6 0.03
(n 5 5)
137 6 10
0.15 6 0.06
The results are presented for the three lowest UVA dose-groups and the UVB control group.
aN 5 total number of mice.
bn 5 number of tumor contracting mice.
cNumber too low to fit the data.
Table III. The Weibull parameters ti (in days) and p and their standard errors derived from the prevalence for non-papillomas for different tumor diameters
Daily dose (kJ/m2)
UVA lamp
75
140
240
F40 (UVB)
0.92
Tumor diameter (mm)
,1
1
2
4
(N a 5 24)
t1
p
(N 5 23)
t1
p
(N 5 24)
t1
p
(n b 5 7)
620 6 47
5.8 6 1.7
(n 5 16)
469 6 20
5.9 6 1.2
(n 5 7)
418 6 34
5.0 6 1.3
(n 5 7)
620 6 46
5.8 6 1.7
(n 5 16)
472 6 20
6.0 6 1.2
(n 5 7)
420 6 32
5.2 6 1.4
(n 5 7)
613 6 38
6.8 6 1.9
(n 5 15)
493 6 19
6.8 6 1.4
(n 5 4)c
–
(n 5 6)
633 6 43
7.1 6 2.2
(n 5 10)
555 6 21
8.8 6 2.3
(n 5 2)c
–
(N 5 16)
t1
p
(n 5 16)
69 6 2
7.7 6 1.4
(n 5 16)
78 6 3
7.9 6 1.4
(n 5 15)
98 6 4
6.8 6 1.3
(n 5 5)
137 6 10
9.6 6 4.0
The results are presented for the three lowest UVA dose-groups and the UVB control group.
aN 5 total number of mice.
bn 5 number of tumor contracting mice.
cNumber too low to fit the data.
dermal photo-damage (26) confirming that apart from being
carcinogenic, UVA strongly contributes to photo-aging.
The induction of severe scratching after months of 365-nm
irradiation in the highest dose-groups interferes with the
development and spotting of tumors. This phenomenon has
been reported earlier for UVA sources (3,4,27), but it has also
been reported to precede tumor induction at high levels of
312-nm radiation from Philips TL01 lamps (28). In the latter
case it clearly affected tumor occurrence as evidenced by a t50
that did not react to a further increase in daily dose. Tumors
arising in scar tissue are clearly suspected of being promoted
by non-specific trauma to the skin, i.e. not purely related to
the UVA radiation. In our highest dose-group (430 kJ/m2
per day) the scratching totally obliterated the undisturbed
occurrence of tumors. In the next dose-group (240 kJ/m2 per
day) scratch marks also censor tumor occurrences to a large
extent, but this interference can be eliminated, or at least
minimized, by discarding an animal from the analysis if the
scarred skin area becomes too large (.25% of the back skin
in a dorsal view, i.e. ~2.5 cm2). The dose–response for the
induction of scratch marks clearly differs from that for inducing
tumors as there is much less interference from scratch marks
in the two lower dose-groups (140 and 75 kJ/m2 per day).
The prevalences of papillomas in our experiment are too
low and the steepness of the prevalence curves too variable to
draw any conclusions. The Kaplan–Meier plots (Figure 3) do
not indicate that papilloma induction kinetics by 365-nm
irradiation differs from that of earlier reports with broadband
UVA sources (5,6): an early onset of papillomas is seen but
the prevalence rises very slowly. The prevalence for nonpapillomas in the present UVB-irradiated control group shows
a time course that is very similar to that observed earlier, i.e.
the steepness given by p 5 7.9 6 1.4 (see Appendix) for
1-mm tumors is within the range found earlier, of six to nine
1017
A.de Laat, J.C.van der Leun and F.R.de Gruijl
Fig. 6. The dose-dependency of tumor formation: the median tumor
induction times (t50 on a logarithmic y-axis) for 1-mm tumors versus the
daily UV dose (D on a logarithmic x-axis) of Westinghouse FS40
(broadband UVB) exposure (1) compared with 365-nm exposure (n). The
drawn lines depict the weighted least-square fits to the data showing that t50
is direct proportional to D–r. The r-values and their standard errors are
depicted along the lines.
Fig. 4. Prevalences for 1-mm non-papillomas (in a Kaplan–Meier plot) and
the fitted Weibull curves in the UVA-exposed (365 nm) groups; (a)
240-kJ/m2 dose-group (b), 140-kJ/m2 dose-group, (c) 75-kJ/m2 dose-group.
Fig. 5. Previously determined SCUP action spectrum for skin cancer
induction in SKH:HR1 mice (—) with upper and lower margins of
uncertainty (...) completed with present data on the relative carcinogenicities
at 365 nm measured in the different dose-groups (s). The relative
carcinogenicity shifts from 0.56310–4 in the 240-kJ/m2 dose-group to
0.77310–4 in the 140-kJ/m2 dose-group up to 0.90310–4 in the 75-kJ/m2
dose-group.
1018
(or, the spread in time σ 5 0.11 6 0.02 is close to the earlier
range of 0.12 to 0.18) (13). The time course of non-papilloma
prevalences under daily 365-nm irradiation is also similar,
albeit that the steepness p 5 5.7 6 1.4 touches the lower end
of the previous range (or the spread σ 5 0.22 6 0.05 touches
the upper end of the range). Although the steepness (p) of the
prevalence with 365-nm radiation is not significantly different
(2.2 6 2.0) from that with UVB radiation, if we also consider
the p-values of UVB radiation from earlier experiments (13),
it tends to be lower in the present experiment. The variation
in irradiation over the cage (maximally 20% lower in extreme
corners of the eight-compartment cages) in the present UVA
experiment could be somewhat higher than in UVB experiments and thus increase the spread in tumor induction times,
i.e. decreasing p. We estimate, however, that this effect does
not increase p by more than approximately 7%.
The median induction time t50 was found to be inversely
proportional to the daily dose (D) to the power r (i.e. t50 ~
D–r) with r 5 0.62 6 0.03 for the smallest tumors induced by
Westinghouse FS40 sunlamps (13,29); and in a similar smaller
study r 5 0.53 6 0.03 (30) and r 5 0.584 6 0.002 with
Philips TL01 lamps (31). In the present study we find r 5
0.35 6 0.08 (if we limit ourselves to the two lowest dosegroups, r 5 0.44 6 0.11). Although the uncertainty in r at
365 nm is rather large, the dose-dependency of tumor induction
times appears to be lower with 365-nm radiation than with
UVB radiation (see Figure 6): r is about 0.23 6 0.09 lower.
Multiplying the p-value for the power of the time dependency
by r (as derived for t1) yields an estimate for the p1-value for
the power of the dose-dependency, as described in Equation 4
in the Appendix. Taking p 5 5.7 6 1.4 and r 5 0.34 6 0.09
for 365-nm radiation, we find that p1 5 1.9 6 0.7. We reported
earlier that p1 5 4.3 6 0.5 for UVB radiation (14), which
would imply that the p1 value for 365-nm radiation is about
2.4 6 1.0 lower; i.e. about the same difference as for the pvalue (2.2 6 2.0), albeit that latter difference was not significant
owing to the larger error. Clearly, the errors are too large to
determine these differences precisely and draw a firm conclusion, but the present data indicate that the p1 and perhaps the
p values are lower with 365-nm radiation than with UVB
UVA carcinogenesis versus UVB carcinogenesis
radiation. In terms of the multi-step interpretation of the
Weibull description (see Appendix, and Reference 14) this
could mean that somehow a process that is UV-driven with
UVB irradiation is not operating under chronic 365-nm exposure. The nature of such processes is a matter of conjecture,
but it could be related to a UVB-induced mutation that provides
developmental advantage. In connection with the latter, Ziegler
et al. (32) argued that early p53-mutated cells would have
such an advantage owing to a deficient apoptotic response,
and this hypothesis was further substantiated by Berg et al.
(10). Interestingly, we found that p53-mutations occur at a
much lower frequency in UVA- than in UVB-induced tumors
(Van Kranen et al., in press.)
In sum, the dose-dependency of tumor formation appears
to be lower with UVA1 radiation than with UVB radiation (r
equals 0.35 versus 0.6, respectively). This difference in dosedependency implies that the ratio in carcinogenicity of UVA1
over UVB will increase as the daily dose is lowered. Consequently, the carcinogenicity of UVA1 relative to that of UVB
cannot be presented by a single multiplicative factor, which
implies that the carcinogenic action spectrum in the UVA1 is
not uniquely definable. Figure 5 demonstrates this point by
plotting the different relative carcinogenicities of three different
regimens of daily UVA1 (365-nm) exposure: the 365-nm
carcinogenicity is found to be 0.6–0.9310–4 times that of 293nm. This entire range is well within the margins of uncertainty
in the SCUPm action spectrum in the UVA1 region. From this
it follows that the definition of the action spectrum in the
UVA1 region cannot simply be made more accurate by
incorporating the present data. The shift in relative UVA1
carcinogenicity also complicates the extrapolation of UVA1
carcinogenicity from mice to humans. Such an extrapolation
may be guided by analysis of mutation spectra in oncogenes
or tumor suppressor genes from skin tumors from mouse and
man. These mutation spectra may provide information on the
relative contribution from direct UV(B)-induced DNA damage
and indirectly UV(A1)-induced DNA damage.
Berg et al. (6) found that carcinogenic doses from broadband
UVA1/UVA2 sources practically added up arithmetically to
carcinogenic doses from broadband UVB/UVA sources. The
carcinogenic effects from UVA sources in these experiments
were dominated by the carcinogenicity from UVA2 radiation.
Hence, it may be concluded that a single carcinogenic (SCUP)
action spectrum appears to describe the combined effects of
broadband UVB/UVA2 sources correctly. Deviations may be
expected to occur if UVA1 becomes dominant, but in the
UVA1 spectral region the error margins in the action spectrum
are so large that these deviations may not become apparent.
In conclusion, the present experiment shows a different
dose–time relationship for skin tumor induced by 365-nm
radiation when compared with UVB radiation; the UV dosedependency appears to be less steep. The actual experimental
data fall well into the anticipated range based on the earlier
derived murine action spectrum (SCUP-m) and the uncertainty
therein. The underlying differences between 365 nm and UVB
in UV-driven processes in carcinogenesis need to be elucidated
for a meaningful extension of the carcinogenic action spectrum
in the UVA1 region. Risk assessments for humans that use
UVA1 sources for tanning could be complicated by a relative
increase in carcinogenicity of UVA1 in comparison to that of
UVB with lower dose levels. This increase suggests that the
risk of UVA1 under conditions of human exposures could well
be larger than previously expected.
Acknowledgements
We are very grateful to Charles C.E.Meulemans and the late Piet Tielemans
of Philips Lighting BV for all their efforts in the development and the
manufacturing of the UVA source and appropriate filter system. We thank
Hans Sturkeboom for maintenance of the animals and the assistance with
construction of the irradiation set-up. We also thank Ilse Bihari for preparing
and staining paraffin sections and Dr Johan Toonstra for histological characterization of the excised tumors. This work was partially supported by Philips
Netherlands BV and by the Environment Programme of the European
Community, Grant ENV4-CT96-172.
References
1. Kligman,L.H. (1991) The hairless mouse and photo-aging. Photochem.
Photobiol., 54, 1109–1118.
2. Forbes,P.D., Davies,R.E. and Cole,C. (1982) Carcinogenesis by UVA
alone. 10th Annual Meeting of the American Society of Photobiology.
Pergamon Press, New York (Abstract).
3. Van Weelden,H., De Gruijl,F.R. and Van der Leun,J.C. (1986)
Carcinogenesis by UVA, with an attempt to assess the carcinogenic risks
of tanning with UVA and UVB. In Urbach,F. and Gange,R.W. (eds) The
Biological Effects of UVA Radiation. Praeger Publishers, New York, pp.
137–146.
4. Sterenborg,H.J.C.M. and Van der Leun,J.C. (1990) Tumorigenesis by a
long wavelength UV-A source. Photochem. Photobiol., 51, 325–330.
5. Kelfkens,G., De Gruijl,F.R. and Van der Leun,J.C. (1991) Tumourigenesis
by short wave ultraviolet-A: Papillomas versus squamous cell carcinomas.
Carcinogenesis, 12, 1377–1382.
6. Berg,R.J.W., De Gruijl,F.R. and Van der Leun,J.C. (1993) Interaction
between ultraviolet-A and ultraviolet-B radiations in skin cancer induction
in hairless mice. Cancer Res., 53, 4212–4217.
7. De Gruijl,F.R., Sterenborg,H.J.C.M., Forbes,P.D., Davies,R.E., Cole,C.,
Kelfkens,G., Van Weelden,H., Slaper,H. and Van der Leun,J.C. (1993)
Wavelength dependence of skin cancer induction by ultraviolet irradiation
of albino hairless mice. Cancer Res., 53, 53–60.
8. De Gruijl,F.R. and Van der Leun,J.C. (1994) Estimate of the wavelength
dependency of ultraviolet carcinogenesis in humans and its relevance
to risk assessment of stratospheric ozone depletion. Health Phys., 67,
319–325.
9. Kelfkens,G., Gruijl de,F.R. and Leun van der,J.C. (1990) Ozone depletion
and increase in annual carcinogenic ultraviolet dose. Photochem.
Photobiol., 52, 819–823.
10. Berg,R.J.W., Van Kranen,H.J., Rebèl,H.G., De Vries,A., Van Vloten,W.A.,
Van Kreijl,C.F., Van der Leun,J.C. and De Gruijl,F.R. (1996) Early p53
alterations
in
skin
carcinogenesis
by
UVB
irradiations:
immunohistochemical detection of mutant p53 protein in clusters of
preneoplastic epidermal cells. Proc. Natl Acad. Sci. USA, 93, 274–278.
11. Kaplan,E.L. and Meier,P. (1958) Non-parametric estimation from
incomplete observations. J. Am. Stat. Assoc., 53, 203–224.
12. Peto,R., Pike,M.C., Day,M.E., Gray,R.G., Lee,P.H., Parish,S., Peto,J.,
Richards,S. and Wahrendorf,J. (1980) Guidelines for simple sensitivity
test for carcinogenic effects in long-term animal experiments. In IARC
Monographs on the Evaluation of the Carcinogenic Risk of Chemicals in
Humans, Suppl. 2. IARC, Lyon, France, pp. 311–426.
13. De Gruijl,F.R., Van der Meer,J.B. and Van der Leun,J.C. (1983) Dosetime dependency of tumor formation by chronic UV exposure. Photochem.
Photobiol., 37, 53–62.
14. De Gruijl,F.R. and Van der Leun,J.C. (1991) Development of skin tumors
in hairless mice after discontinuation of ultraviolet irradiation. Cancer
Res. 51, 979–984.
15. Armitage,P. and Doll,R. (1957) A two-stage theory of carcinogenesis in
relation to the age distribution of human cancer. Br. J. Cancer, 11, 161–169.
16. Fisher,J.C. (1958) Multiple-mutation theory of carcinogenesis. Nature,
181, 651.
17. Pike,M.C. (1966) A method of analysis of a certain class of experiments
in carcinogenesis. Biometrics, 22, 142–161.
18. Pound,A.W. and Withers,H.R. (1963) The influence of some irritant
chemicals and scarification on tumour initiation by urethane in mice. Br.
J. Cancer, 17, 460–470.
19. Kelfkens,G., Van Weelden,H., De Gruijl,F.R. and Van der Leun,J.C. (1991)
The influence of dose rate on ultraviolet tumorigenesis. J. Photochem.
Photobiol. B., 10, 41–50.
20. Kligman,L.H., Crosby,M.J., Miller,S.A., Hitchins,V.M. and Beer,J.Z. (1992)
Carcinogenesis by long wavelength UVA and failure of reciprocity. In
Urbach,F. (ed.) Biological Responses to Ultraviolet A Radiation. Valdenmar
Publishing Company, Overland Park, KA, pp. 301–308.
1019
A.de Laat, J.C.van der Leun and F.R.de Gruijl
21. Bissett,D.L., Hannon,D.P. and Orr,T.V. (1987) An animal model of solaraged skin: histological, physical and visible changes in UV-irradiated
hairless mouse skin. Photochem. Photobiol., 46, 367–378.
22. Bissett,D.L., Hannon,D.P. and Orr,T.V. (1989) Wavelength dependence of
histological, physical and visible changes in chronically UV-irradiated
hairless mouse skin. Photochem. Photobiol., 50, 763–769.
23. Kligman,L.H., Kaidbey,K.H., Hitchins,V.M. and Miller,S.A. (1987) Long
wavelength (.340 nm) ultraviolet-A induced skin damage in hairless mice
is dose dependent. In Passchier,W.F. and Bosnjakovic,B.F.M. (eds) Human
Exposure to Ultraviolet Radiation: Risks and Regulations. Elsevier Science,
Oxford, pp. 77–81.
24. Fourtanier,A., Labat-Robert,J., Kern,P., Berrebi,C., Gracia,A.M. and
Boyer,B. (1992) In vivo evaluation of photoprotection against chronic
ultraviolet-A irradiation by a new sunscreen Mexoryl SX. Photochem.
Photobiol., 55, 549–560.
25. Bruls,W.A.G., Slaper,H., Van der Leun,J.C. and Berrens,L. (1984)
Transmission of human epidermis and stratum corneum as a function of
thickness in the ultraviolet and visible wavelengths. Photochem. Photobiol.,
40, 485–494.
26. Lowe,N.J.,
Meyers,D.P.,
Wieder,J.M.,
Luftman,D.,
Borget,T.,
Lehman,M.D., Johnson,A.W. and Scott,I.R. (1995) Low doses of repetitive
ultraviolet A induce morphologic changes in human skin. J. Invest.
Dermatol., 105, 739–743.
27. Kaase,H., Zidowitz,B. and Berger,H. (1984) Spectrale strahlungsmessung
an UV-bestrahlungsgeraeten und experimente zur abschaetzung der
photocarzinogenese. Lichtforschung, 1, 43–47.
28. Gibbs,N.K., Traynor,N.J., Mackie,R.M., Cambell,I., Johnson,B.E. and
Ferguson,J. (1995) The phototumorigenic potential of broad-band (270–
350 nm) and narrow-band (311–313 nm) phototherapy sources cannot be
predicted by their edematogenic potential in hairless mouse skin. J. Invest.
Dermatol., 104, 359–363.
29. De Gruijl,F.R. and Van der Leun,J.C. (1980) A dose–response model for
skin cancer induction by chronic UV exposure of a human population. J.
Theoret. Biol., 83, 487–504.
30. Slaper,H. (1987) Skin cancer and UV exposure: Investigations on the
estimation of risks. PhD Thesis, University of Utrecht, The Netherlands.
31. Sterenborg,H.J.C.M., Van Weelden,H. and Van der Leun,J.C. (1988) The
dose–response relationship for tumourigenesis by UV radiation in the
region 311–312 nm. J. Photochem. Photobiol. B, 2, 179–194.
32. Ziegler,A., Jonason,A.S., Leffel,D.J., Simon,J.A., Sharma,H.W.,
Kimmelman,J., Remington,L., Jacks,T. and Brash,D.E. (1994) Sunburn
and p53 in the onset of skin cancer. Nature, 372, 773–776.
Received on September 12, 1996; revised on December 1996; accepted on
February 20, 1997
Appendix
According to probability theory the prevalence can generally be written as
[1 – e–Y]3100%, where Y is the expected tumor yield, i.e. the average number
of tumors per mouse (Y is the cumulative hazard in probability theory). For
the Weibull distribution
Y 5 (t/t1) p
[1]
where t stands for time (days of exposure) and t1 is the time at which Y 5 1
and the prevalence reaches 63%; p is a constant, the power of the time with
which Y rises; it corresponds to the steepness of the prevalence curve. This
power p is likely to be proportional to the number of rate-limiting steps in
the process of carcinogenesis, and for 1-mm tumors induced in mice by UVB
p 5 6.9 6 0.8 (6 standard error) (14).
The following relationships between the log-normal and Weibull
parameters hold:
t50 5 t1 (ln2) 1/p
[2]
and from first derivatives of prevalences versus ln(time) around the 50% it
follows that
σ µ 1.14/p
[3]
In the UVB it was found that t1 D 5 constant (14), where D is the daily
UV dose and r 5 0.62. Substitution of this dose-time dependency in Equation
1 transforms it into
r
Y 5 (D/Do) p1 (t/to) p
[4]
where Do, p1 and to are new constants, with p1 5 r p 5 4.3 6 0.5 for 1-mm
tumors induced by UVB radiation. Equation 4 shows that the tumor yield is
1020
also simply proportional to the daily dose to a certain power p1 which is
likely to be proportional to the number of rate-limiting steps driven by UV
radiation. The fact then that the power p1 of the daily dose is smaller than
the power p of the time indicates that only a part of the rate-limiting steps in
the process are UV-driven; e.g. clonal expansion of pre-neoplastic cells could
introduce a UV-independent intermediary stage (16). This interpretation
appears to be in line with the results obtained from experiments in which
daily UV exposures were discontinued (14), and are also used here as a frame
of reference.