Effectiveness of Vitamin D Analogues in Treating Large Tumors and DuringProlonged Use in Murine Retinoblastoma Models

Objective  To investigate the effectiveness of the vitamin D analogues 1,25-(OH)₂-16-ene-23-yne vitamin D₃ (16,23-D₃) and 1α-hydroxyvitaminD₂ (1α-OH-D₂) in inhibiting retinoblastoma growthin large tumors in a xenograft model and with prolonged use in a transgenicmodel.

Methods  For the large-tumor study, the xenograft athymic mouse/human retinoblastomacell (Y-79) model was used. Subcutaneous tumors were allowed to grow to anaverage volume of 1600 mm³. Systemic treatment with 1 of the vitaminD analogues or with vehicle (control groups) was carried out for 5 weeks.For the long-term study, transgenic β–luteinizing hormone–largeT antigen (LHβ-Tag) mice were systemically treated with 1 of the 2 compoundsor vehicle (control groups) for up to 15 weeks. Tumor size and signs of toxicitywere assessed.

Results  In the large-tumor study, tumor volume ratios for the 1α-OH-D₂ and 16,23-D₃ groups were significantly lower than thosefor controls (P<.002). No significant differencesin tumor volume were seen between the 1α-OH-D₂ and 16,23-D₃ groups (P = .15). In the long-term study,the 1α-OH-D₂ group showed significantly smaller tumor sizecompared with its control (P<.001). No significantdifference was seen between the 16,23-D₃ group and its control.Some toxic effects related to hypercalcemia were seen in both studies.

Conclusions  In athymic mice in the large-tumor study, both 1α-OH-D₂ and16,23-D₃ were effective in inhibiting tumor growth compared withcontrols. In the long-term study, 1α-OH-D₂ inhibited tumorgrowth but 16,23-D₃ did not. Effective doses of both compoundscaused hypercalcemia and a significant increase in mortality.

Clinical Relevance  Use of 1α-OH-D₂ inhibited tumor growth in large tumorsand with long-term treatment compared with controls. Because of hypercalcemia-relatedtoxic effects seen in the present experiments, in clinical trials, serum calciumlevels should be carefully monitored. This analogue may require use with drugsthat lower serum calcium levels or use of relatively lower doses or skippeddoses. The ideal alternative solution would be to identify vitamin D analoguesthat retain the antineoplastic action without the calcemic activity.

Vitamin D compounds have been recognized for many years as having potentialusefulness in the treatment of a variety of cancers, including retinoblastoma.¹ A major obstacle to their use in human studies hasbeen drug-induced hypercalcemia.¹ Vitamin Danalogues have previously been shown in preclinical studies of retinoblastomausing xenograft and transgenic models to cause apoptosis² andto inhibit angiogenesis.³ Previous preclinicalstudies^4,5 demonstrated that 2commonly used forms of vitamin D, calcitriol (1,25-dihydroxyvitamin-D₃) and ergocalciferol (vitamin D₂), inhibited tumor growthbut with marked hypercalcemia-related toxic effects in mouse xenograft andtransgenic models. Also demonstrated was a similar inhibitory effect withthe newer synthetic vitamin D analogues 1α-hydroxyvitamin D₂ (1α-OH-D₂)^6-9 and1,25-(OH)₂-16-ene-23-yne vitamin D₃ (16,23-D₃),^10,11 with reduced calcemic effects. Allof these studies were carried out using small tumors treated for a relativelyshort time (5 weeks). The present studies were carried out to determine whethervitamin D analogues are effective in the treatment of retinoblastoma in largertumors and remain effective with more prolonged administration.

All research using mouse models of retinoblastoma conformed to the guidelinesset by the Research Animal Resources Center of the University of Wisconsinand the Association for Research in Vision and Ophthalmology Statement forthe Use of Animals in Ophthalmic and Vision Research. These guidelines permitlarge-tumor studies to be carried out in athymic mice but not in β–luteinizinghormone–large T antigen (LHβ-Tag) transgenic mice because of therelatively small size of the tumors occurring in the eyes and the morbidityassociated with orbital extension. Likewise, long-term studies were suitablefor the transgenic but not athymic mice.

Pure crystalline 1α-OH-D₂ was provided by Bone CareInternational and was prepared for administration with drug concentrationsconfirmed in the manner previously described elsewhere.¹² Asolution of 1α-OH-D₂ was diluted in coconut oil to a concentrationof 0.2 µg/0.1 mL for use in this study. Each treated mouse received0.2 µg of 1α-OH-D₂ (approximately 10 µg/kg) pertreatment. This dose has previously been demonstrated in toxicity and dose-responsestudies^1,12,13 tobe the effective dose with the least toxicity. Pure crystalline 16,23-D₃ (provided by ILEX Oncology Inc, San Antonio, Tex) was prepared forinjection as previously described elsewhere.^1,14,15 Thisdrug was diluted in mineral oil to a concentration of 0.5 µg/0.1 mL.Each mouse in the treatment group received 0.5 µg of 16,23-D₃ (approximately25 µg/kg) per treatment. This dose was found in previous toxicity anddose-response studies^1,14,15 tobe the effective dose with the least toxicity.

A total of 119 athymic "nude" mice (4-6 weeks old) were given dorsalsubcutaneous injections of 2 × 10⁷ Y-79 human retinoblastomacells suspended in 0.5 mL of Iscove culture medium supplemented with 20% fetalbovine serum and basement membrane matrix suspensions. Details of culturemethods have been previously described elsewhere.^1,14 Aftertumor cells were injected, all mice were maintained on a vitamin D–and calcium-restricted diet (PD; Purina Mills Inc, St Louis, Mo). This dietwas used to minimize hypercalcemia and to control for any effect of vitaminD in the diet. Mice were randomized to 1 of 4 treatment groups: (1) 1α-OH-D₂ (0.2 µg per mouse; 40 animals), (2) gavage (GV) control (0.1mL of coconut oil; 20 animals), (3) 16,23-D₃ (0.5 µg permouse; 40 animals), or (4) intraperitoneal (IP) control (0.1 mL of mineraloil; 19 animals). The method of administration conformed with instructionsprovided by the respective pharmaceutical companies supplying the drugs toobtain maximum blood levels. Tumors were allowed to grow for 19 days to anaverage volume of 1600 mm³ before initiation of treatment witheither oral GV (1α-OH-D₂ and GV control groups) or IP (16,23-D₃ and IP control groups). Treatment was given 5 times per week for 5weeks. Baseline tumor volume and animal body weight measurements were determinedbefore the first treatment. Each animal was weighed, and its tumor was measuredusing calipers, twice per week during treatment and just before euthanizationon the last treatment day as previously described elsewhere.^1,13,14 Investigatorswere masked to final tumor measurements and histopathologic examination findingsbut not to drug delivery method.

The LHβ-Tag mice are a well-characterized transgenic model of retinoblastoma.¹⁶ A total of 170 LHβ-Tag transgene-positive micewere randomized by sex and litter into 2 treatment groups and 2 control groupscorresponding to the 4 treatment groups described for the large-tumor athymicstudy in the previous subsection. The presence of the transgene was confirmedby polymerase chain reaction.¹⁶ These micewere maintained on the same vitamin D– and calcium-restricted diet asnoted in the previous subsection. Each mouse in the 1α-OH-D₂ treatmentgroup received 0.2 µg/d. Each mouse in the 16,23-D₃ treatmentgroup received 0.5 µg/d. Treatment was given 5 times per week. One thirdof the mice in each group were treated for 5 weeks, one third for 10 weeks,and one third for 15 weeks. Baseline body weights were recorded before thefirst treatment, and animals were reweighed twice per week during treatmentand before euthanization on the last treatment day. Details of the oral GVand IP methods of treatment with vitamin D analogues have been previouslydescribed elsewhere.^12,14 Doseswere skipped for up to 2 consecutive days in mice that either developed weightloss (>20% of baseline weight) or severe lethargy, and treatment was resumedwhen the affected animals regained the lost weight or resumed normal activity.Investigators were masked to histopathologic examination findings and tumormeasurement but not to drug delivery method.

In athymic mice, tumor size was measured daily 5 times a week in 3 dimensions(length, width, and height) by means of calipers measuring to the nearestmillimeter, and the volume was approximated by multiplying the 3 measurements.After euthanization, each tumor was excised, measured using calipers in 3dimensions, and weighed. The tumors were fixed in 10% neutral buffered formalinfor standard histopathologic processing. Five-micrometer sections were cutand stained with hematoxylin-eosin. In addition, von Kossa–stained preparationswere made. These were examined microscopically, and the histologic featureswere recorded as previously described elsewhere.^1,13,14

In transgenic mice, after euthanization the eyes were enucleated andplaced in 10% neutral buffered formalin for standard histopathologic sectioning.Four serially sectioned 5-µm-thick sections were cut from each of thesuperior, middle, and inferior areas of the globe and were stained with hematoxylin-eosin.The 4 sections from each globe area were examined under a microscope, andthe section with the largest tumor from each of the 3 areas was used for measurement.The outline of the tumor in each section was traced from a microscopicallydigitized image, and the area was measured using image analysis software (Optimasversion 6.5; Media Cybernetics, Silver Spring, Md). The 3 tumor areas fromeach representative portion of the globe were averaged together to obtainthe mean tumor measurement of each eye (expressed in square micrometers).The measurements from both eyes were then averaged to provide the mean tumorarea per mouse. Other histopathologic features were also evaluated.^1,12,15

Serum samples from representative mice in each group were obtained justbefore euthanization from the axillary vessels and were analyzed for calciumlevels by Marshfield Laboratories, Marshfield, Wis.

Kidneys, lungs, and liver were harvested from representative mice ineach group in both arms of the study and were processed histologically. Thenumber of organs harvested varied among the groups. Kidneys were stained withvon Kossa stain in addition to hematoxylin-eosin stain to ascertain the severityof renal calcification. Two sections of each kidney were examined by maskedreviewers (D.M.A. and A.K.), and the number of calcium deposits was determinedand averaged for each kidney. Each kidney was graded according to the followingscale: grade 0, no calcifications; grade I, 1 to 7 foci of calcification;grade II, 8 to 15 foci of calcification; and grade III, greater than 15 fociof calcification.

Toxic effects were assessed by using the following variables: survival,changes in body weight, serum calcium levels, and degree of kidney calcification.Animals that died before completion of the treatment protocol were not examinedwith regard to the latter 3 categories. Further details regarding methodsinvolved in toxic effect assessment are described in detail elsewhere.^12-15

Tumor weight, tumor volume or area, animal weight, serum calcium level,and kidney calcification were analyzed using 1-way analysis of variance toassess statistical differences among the groups. Pairwise comparisons werethen performed to detect statistical differences between particular dose groups.The data from various measurements were transformed to the log scale beforeanalysis to stabilize the variance. The change in animal weight (from firstto last measurement) was restricted to animals that survived until the lastmeasurement. The effect of dose of vitamin D analogues on mortality was assessedusing a generalized linear model assuming binomial variability.

The effect of "layer" (ie, animal batch) on response was accounted forin all analyses by including a blocking term in the model. All significantglobal tests for effect of dose were followed by pairwise analyses to assessdifferences between specific dose groups.

In the long-term arm of study, comparisons of each drug's efficacy overeach time point (5, 10, or 15 weeks) were also performed.

In the athymic mice with large tumors, mean change in tumor volume wasanalyzed as a proportional change during treatment (volume at treatment enddivided by volume at treatment start) (Table 1). The tumor volume ratio for the 1α-OH-D₂ groupwas significantly lower than that for the GV control group, and the ratiofor the 16,23-D₃ group was significantly lower than that for theIP control group (P<.002 for each). No statisticallysignificant differences in tumor volume were seen between the 1α-OH-D₂ and 16,23-D₃ groups (P = .15).

In the long-term study with LHβ-Tag mice, the tumor sizes after5, 10, and 15 weeks of treatment are given in Table 2. The 1α-OH-D₂ group showed significantlysmaller tumor areas compared with the GV control group at each time point(P<.001). However, no significant difference wasseen between the 16,23-D₃ group and the IP control group at anyof the 3 times.

The survival data for the study are presented in Table 1 and Table 2.There were significantly fewer survivors among athymic mice in the 1α-OH-D₂ group in the large-tumor study than in the GV control group (65% vs90%; P = .03) (Table 1). There were also fewer survivors among the LHβ-Tagmice in the 1α-OH-D₂ group in the long-term study than inthe GV control group (83% vs 100%; P = .048) (Table 2).

Serum calcium levels for both study arms are given in Table 1 and Table 2 and Table 3. All treatment groups had significantlyhigher serum calcium levels compared with their respective control groups(P≤.02), except both of the 15-week treatmentgroups, which did not show any difference (P>.25).No difference in serum calcium levels was found between the 2 treatment modalities.The LHβ-Tag mice developed well-differentiated retinoblastoma, with rosettes,some calcification, and areas of necrosis. Treated animals had smaller tumors,but the degree of differentiation and the proportional amounts of calcificationand necrosis were similar in treated and control animals. The athymic/Y-79mice had poorly differentiated tumors, with an extensive and diffuse componentof dead cells. The diffuse nature of the necrosis is similar to that seenin large intraocular retinoblastoma in enucleated human eyes. Because of thediffuse distribution of dead cells, the precise percentage of the tumor theyrepresent cannot be accurately determined using the present techniques. Itis our impression, however, that dead cells generally represented one thirdto two thirds of the cells present and were more numerous in treated eyes.The gradations of the kidney calcifications are given in Table 4. The treatment groups in both arms of the study showed agreater degree of calcification compared with controls, with the 1α-OH-D₂ groups exhibiting greater severity of calcification compared withthe 16,23-D₃ groups. Gross and histopathologic examination of lungand liver tissues from randomly selected mice in each treatment group showedno evidence of metastatic lesions or other abnormalities.

Retinoblastoma is the most common intraocular malignancy of childhood,occurring once in every 20 000 live births worldwide.¹⁷ Currentmethods of treatment include enucleation, external beam radiotherapy, scleralplaque brachytherapy, cryotherapy, photocoagulation, and chemotherapy.¹⁷ Although current treatment methods have achievedsurvival of 90%, there remains a need for improved treatment alternativesto provide better visual results and to decrease the risk of secondary nonocularcancers in hereditary retinoblastoma.¹⁸

In recent years, there has been strong interest in systemic chemotherapyand, particularly, "chemoreduction" in the treatment of retinoblastoma.^19-21 There are 4 generalcircumstances in the treatment of retinoblastoma in which chemotherapy isused^20,21: (1) to shrink tumorsin eyes with visual potential that are too large to treat with focal methodsto a size at which photocoagulation, cryotherapy, thermotherapy, or radioactiveplaques can be administered; (2) in patients younger than 1 year who haveadvanced bilateral tumors that require external beam radiotherapy for cure;(3) as a single modality (which rarely obtains a permanent response)²⁰; and (4) for the treatment of extraocular spread.²¹ Numerous serious short-term sequelae and less severecomplications can occur.²⁰ These drugs aremutagenic, and the development of secondary nonocular cancers is a well-documentedrisk.^20,22-28 Newretinoblastomas have been reported to develop in the eyes of patients undergoingtreatment with systemic chemotherapy.^20,29

Verhoeff,³⁰ in 1966, hypothesized thatcalcification induced spontaneous regression in retinoblastoma and suggestedthat treatment with vitamin D might prove effective. Vitamin D has since beenshown to inhibit the growth of retinoblastoma in vitro³¹ andin vivo in the athymic mouse model⁴ and inthe transgenic mouse model.⁵

The effect, however, is unrelated to either high serum calcium levelsor calcium deposition in the tumor, and, in fact, the clinical usefulnessof vitamin D is limited by the toxic effects associated with hypercalcemia.^1,32 The antineoplastic mechanism of actionof vitamin D compounds used in models of human retinoblastoma has been demonstratedto be apoptosis due to increased expression of p53 and p21² andalso inhibition of angiogenesis.³ Vitamin Dreceptor messenger RNAs, which are necessary for this antineoplastic effect,have been demonstrated in Y-79 retinoblastoma cells, WERI-1 cells, LHβ-Tagtumors, and 23 consecutive freshly removed human retinoblastoma specimensusing reverse transcriptase polymerase chain reactions.¹

Our initial studies were with calcitriol (the active form of vitaminD₃) and ergocalciferol (vitamin D₂).^4,5,33 Thesecompounds consistently inhibited tumor growth by more than 50% compared withcontrols. In the case of transgenic mice, animals were observed in the treatmentgroups with total regression of tumors. This was proven by serially sectioningthe globes. All the negative animals were rechecked to make sure that theycarried the transgene. In this study,³ 6 of10 animals in the high-dose calcitriol group and 4 of 12 animals in the low-dosecalcitriol group showed no tumors, whereas 16 control animals had tumors.The difference was statistically significant as assessed by a χ² test for independence. However, only 25% to 50% of animals survived5 weeks of treatment in the calcitriol and ergocalciferol treatment groupsgiven doses adequate to cause tumor inhibition or regression.

In the athymic/Y-79 xenograft treated with calcitriol, 81% inhibitionof tumor growth was seen compared with controls, and vitamin D₂ inhibitionranged from 58% to 66%, but no regression was seen in these tumors. The real-timepolymerase chain reaction expression of vitamin D receptors is barely perceptiblein the Y-79 retinoblastoma line. It is far lower than the 23 retinoblastomasamples taken from enucleated eyes from different patients.¹ Incontrast, the LHβ-Tag tumors have a greater degree of vitamin D receptorexpression. This difference seems to be related to the degree of drug response.We concluded that calcitriol and ergocalciferol showed impressive antineoplasticactivity but caused hypercalcemic activity that was too excessive for themto be used in human clinical trials.

We then turned our attention to 2 synthetic vitamin D analogues, 16,23-D₃ and 1α-OH-D₂, which have antineoplastic effects similarto calcitriol and ergocalciferol but with reduced hypercalcemic activity.In short-term experiments (5 weeks) in small tumors, 16,23-D₃ therapycaused 55% inhibition in LHβ-Tag transgenic mice³⁴ and58% inhibition in athymic mice.¹⁴ The administrationof effective doses of 16,23-D₃ resulted in 11% mortality in transgenicmice³⁴ and 25% mortality in athymic mice¹⁴ compared with controls. Use of 1α-OH-D₂ resulted in 81% tumor growth inhibition in transgenic mice¹² and 62% tumor growth inhibition in athymic mice¹³ compared with controls. Use of 1α-OH-D₂ resulted in 12% mortality in transgenic mice¹² and38% mortality¹³ in athymic mice. Thus, althoughthe mortality rate was lower than with use of calcitriol and ergocalciferol,the toxic effects of hypercalcemia were still observed. These earlier experimentswith small tumors in short-term therapy simulated the conditions in whichchemoreduction is most commonly used.

The experiments described in the present article were intended to determinethe effectiveness and the toxic effects of treatment with 16,23-D₃ and1α-OH-D₂ in large tumors and in tumors receiving long-termtherapy. The size of tumors that could be treated and the duration of treatmentwere limited by the guidelines of the University of Wisconsin Medical SchoolAnimal Care and Use Committee regarding suffering and stress caused to animals.Specifically, we wanted to determine whether drug-resistant cell lines developwith prolonged use of these drugs or with large tumors, as well as the drugs'comparative effectiveness in these situations. These experiments simulatethe circumstances of treatment for extraocular spread. In the large-tumorstudies, 16,23-D₃ therapy achieved 62% growth inhibition comparedwith controls, and 1α-OH-D₂ therapy achieved 51% inhibitionof tumor growth compared with controls. These determinations are based ontumor size alone. However, as determined by histopathologic analysis, therewas a diffuse distribution of dead cell in the tumors that constituted onethird to two thirds of the cells present, and we observed that the dead cellswere more numerous in treated animals. We are adapting techniques to accuratelyquantify numbers of viable, common, necrotic, and apoptotic cells.

In the long-term treatment studies using transgenic mice only, 1α-OH-D₂ therapy seemed to be effective. In these experiments, there was 92%regression at 10 weeks and 70% regression at 15 weeks compared with controls.Some treated eyes seemed to have totally regressed tumors on the basis ofthe slides examined, but because serial sections were not examined, this cannotbe stated with certainty. All of the groups that were effectively treatedshowed statistically significant hypercalcemia and associated increased mortalitycompared with controls (Table 1, Table 2, and Table 3). Although these toxic effects and mortality rates are considerablyless than those seen with calcitriol and vitamin D₂ treatment,it remains a potential deterrent to proceeding to clinical trials.

An Investigational New Drug application for 1α-OH-D₂ asa cancer treatment was submitted to the Food and Drug Administration in 1996,and an application for 16,23-D₃ was submitted in 1999. In phase1 trials for treatment of prostate cancer, 1α-OH-D₂ exhibitedtumor growth suppression for stabilization with low but reversible toxic effects.³⁵ This drug is currently being used in phase 2 humantrials of prostate cancer (George Wilding, MD, oral communication, October7, 2003).

Because of their mechanism of action, which differs from that of existingretinoblastoma chemotherapeutic agents, and the fact that they are nonmutagenicagents, we believe that these compounds have potential use in eye-preservingand, in cases of extraocular spread, life-preserving treatment as a componentof multidrug therapy or possibly as a potential single-modality treatment.In clinical trials of 1α-OH-D₂ and 16,23-D₃, hypercalcemia-relatedtoxic effects would need to be carefully monitored, controlling the dose givenand possibly skipping doses in response to elevated serum calcium levels.Drugs that lower serum calcium levels may need to be used. The ideal alternativesolution would be to identify vitamin D analogues that retain the antineoplasticaction but have no calcemic activity.

Correspondence: Daniel M. Albert, MD, MS, Department of Ophthalmologyand Visual Sciences, University of Wisconsin Medical School, F4/344 ClinicalScience Center, 600 Highland Ave, Madison, WI 53792-3284 (dalbert@wisc.edu).

Submitted for publication November 13, 2003; final revision receivedJanuary 23, 2004; accepted January 26, 2004.

This study was supported by grant EY01917 from the National Eye Institute,Bethesda, Md, and an unrestricted grant from Research to Prevent BlindnessInc, New York, NY.

We thank Daniel Dawson, MD, and Joel Gleiser, MD, for treating animalsand measuring tumor size and Kirsten Hope for contributing to preparationof the article.