chapter 20 solutions molecular biology of the cell

CANCER AS A MICROEVOLUTIONARY PROCESS

DEFINITIONS

201 Carcinogenesis

202 Malignant

203 Cancer stem cells

204 Sarcoma

205 Primary tumor

206 Tumor progression

207 Genetically unstable

208 Benign

209 Replicative cell senescence

2010 Carcinoma

TRUE/FALSE

2011 False. A carcinoma consists of a variety of normal cells, along with the can-cer cells. The fibroblasts in the supporting connective tissue, the associatedinflammatory cells, and the cells of the newly formed blood vessels all arenormal cells that are present because they are in contact with the cancer cellmass or have been recruited into the tumor. These normal cells do notevolve from the cancer cell population.

2012 True. Genes can be shut off, for example, by changes in chromatin structure;specific covalent modifications of histones can attract complexes of chro-matin-binding proteins that are stably maintained following DNA replica-tion. Gene transcription can also be eliminated by methylation of CpG din-ucleotides in the promoter region. Specific histone modifications and DNAmethylation are often linked.

2013 False. There is an optimum level of genetic instability for the development ofcancer. A cell must be mutable enough to evolve rapidly, but not so mutablethat it accumulates too many harmful changes and dies.

2014 True. Many cancers appear to be maintained by a small population of stemcells. These cancer stem cells usually divide more slowly than the cells in thebulk of the tumor, and they are less sensitive to treatments aimed at rapidlydividing cells. If the stem cells are not killed, the cancer is likely to return.

In This Chapter

CANCER AS A A457MICROEVOLUTIONARYPROCESS

THE PREVENTABLE A460CAUSES OF CANCER

FINDING THE CANCER- A461CRITICAL GENES

THE MOLECULAR BASIS A464OF CANCER-CELL BEHAVIOR

CANCER TREATMENT: A469PRESENT AND FUTURE

A457

Chapter 20

20Cancer

A458 Chapter 20: Cancer

THOUGHT PROBLEMS

2015 The incidence of cancer increases dramatically with age because it takesmutations in several critical genes to disable a cells normal mechanisms forcontrolling its growth. Since growing cells continually accumulate muta-tions, which they pass on to their progeny cells, the chance that a cell willaccumulate a critical set of mutations increases with age. The steep rise incancer incidence in older women seen in Figure 201 reveals that colon can-cer increases as the sixth power of age, suggesting that it arises only aftermutations have occurred in six or so genes that regulate cell growth in thecolon.

2016 The key difference in the incidences of colon cancer and osteosarcomas isthe size of the population of cells at risk for the disease. Colon cancer arisesfrom the population of proliferating cells in the colon, which are present inroughly the same number throughout life. This population can accumulatemutations over time, giving rise to the age dependence seen in Figure 201.By contrast, the cells responsible for osteosarcomas are present in muchgreater numbers during adolescence, when their proliferation is required toincrease the size of the skeleton, than they are in young children or adults.In this case, it is the number of cells at risk that is the most important deter-minant of the frequency of cancer.

Reference: Knudson AG (2001) Two genetic hits (more or less) to cancer. Nat.Rev. Cancer 1, 157162.

2017 The time at which the death rates due to breast and cervical cancer slow cor-responds to the menopause, at which time the production of estrogendeclines. Estrogen normally promotes the proliferation of cells in the breastand uterus. Thus, the decline in estrogen would reduce the population ofproliferating cells, thereby reducing the risk of cancer in these tissues.

Reference: Armitage R & Doll R (1954) The age distribution of cancer and amulti-stage theory of carcinogenesis. Reprinted in (2004) Br. J. Cancer 91,19831989.

CALCULATIONS

2018 A tumor that arose from 50 cell doublings would contain 250 cells, which is1.1 1015 cells. If 108 cells is 1 gram, then the tumor would weigh 1.1 107 g,which is 24,000 pounds. Thus, a limit of 50 cell divisions, by itself, does notprovide much protection against cancer. Even in a 40-year-old, whosefibroblasts divide about 40 times in culture, a tumor arising from 40 cell divi-sions would weigh more than 20 pounds.

This calculation assumes that all cells survive. It is likely, however, thatmany cells die by apoptosis, especially at early stages in the evolution of can-cer. In combination with extensive cell death, a limit of 50 cell divisionscould protect against cancer.

2019A. The cell-cluster model for cancer formation predicts an exponential rela-

tionship between carcinogen concentration and cancer formation. In theequation, Nxn, the concentration of mutagen, would influence the probabil-ity of mutation, x, so that 1, 2, and 4 times the carcinogen concentrationwould give 1x, 2x, and 4x probabilities of mutation, respectively. The num-bers of cancers should then be N(1x)n, N(2x)n, and N(4x)n, respectively. Toillustrate this with numbers, let N = 109, x = 102, and n = 5, as they are in thebody of the problem. Substituting these numbers gives 0.1, 3.2, and 102.4

THE PREVENTABLE CAUSES OF CANCER A459

cancer clusters for 1, 2, and 4 times the carcinogen concentration. This pre-dicted exponential dependence on carcinogen concentration does notmatch the linear dependence observed experimentally.

B. The tumor-progression model for cancer formation predicts a linear rela-tionship between carcinogen concentration and cancer formation, so longas the frequency of mutation is low. A slowly progressing tumor presents amoving target for mutation. At an early time, say when it has one mutation,carcinogen treatment can stimulate the acquisition of a second mutation inone of its cells, allowing that cell to progress to the next stage. It is unlikely,however, to introduce two mutations into one cell in the small population ofcells that constitutes the early tumor. At a later stage, when the tumor cellshave, say, three changes, the same argument applies: the carcinogen canrealistically induce an additional mutation in only one of its cells. Inductionof one mutation at a time gives a linear dependence on carcinogen concen-tration.

References: Fisher JC & Hollomon JH (1951) A hypothesis for the origin ofcancer foci. Cancer 4, 916918.

Armitage P & Doll R (1954) The age distribution of cancer and a multi-stagetheory of carcinogenesis. Reprinted in 2004 in Br. J. Cancer 91, 19831989.

DATA HANDLING

2020 Although the numbers are small, there seems to be a reasonable correlationbetween smoking and lung cancer. Statistical analysis shows that these dif-ferences are significant (P = 0.04). Similarly, the consistent upward trend inheart attacks with increased smoking also turns out to be statistically signif-icant (P = 0.03). Follow-up studies on this cohort (the most recent after 50years) have confirmed the initial impression conveyed by the original, pre-liminary study. Compared to lifelong nonsmokers, lung cancer is 8-fold, 14-fold, and 25-fold more prevalent among smokers of 114, 1524, and morethan 25 cigarettes per day, respectively. Moreover, in these longer studies itis clear that other cancersfor example, mouth, nose, throat, stomach, liver,kidney, and bladderare also significantly elevated in smokers, althoughnot so dramatically as the incidence of lung cancer.

References: Doll R & Hill AB (1954) The mortality of doctors in relation totheir smoking habits: A preliminary report. Reprinted in 2004 in Br. Med. J.328, 15291533.

Doll R, Peto R, Boreham J & Sutherland I (2004) Mortality in relation tosmoking: 50 years observations on male British doctors. Br. Med. J. 328,15191528.

2021 Development of most cancers seems to require a gradual accumulation ofmutations in a number of different genesat least five or six. In the ongoingpresence of cigarette smoke these mutations evidently accumulate at anincreased rate (over their accumulation in the absence of cigarette smoke).By stopping smoking, an individual returns to the normal, slower rate ofmutation accumulation. Thus, whatever mutations remain to be generatedin a reformed smoker are generated at a slower rate than in a continuingsmoker. The slower rate of accumulation of mutations translates into a lowercumulative risk.

Reference: Peto R, Darby S, Deo H, Silcocks P, Whitely E & Doll R (2000)Smoking, smoking cessation, and lung cancer in the UK since 1950: combi-nation of national statistics with two case-control studies. Br. Med. J. 321,323329.


THE PREVENTABLE CAUSES OF CANCER

DEFINITIONS

2022 Tumor promoter

2023 Carcinogen

TRUE/FALSE

2024 True. The modifications introduced into dietary carcinogens by the action ofthe P-450 enzymes of the liver can convert them from relatively inert com-pounds to reactive ones that damage DNA. The normal function of theseenzymes is to convert ingested toxins to harmless compounds for easyexcretion. Unfortunately, their action on some chemicals renders themhighly mutagenic.

2025 False. Infectious agents are thought to participate in the formation ofroughly 15% of human cancers. In most cases, the specific mechanism isunknown, but there are strong associations, for example, between hepatitisviruses type B and type C and liver cancer, between H. pylori infection andstomach cancer, and between blood fluke infection and bladder cancer.

2026 False. Although it is popular to think so, there is scant evidence to supportthose ideas. However, we certainly know of specific instancessuch as 2-naphthylamine and asbestoswhere industrial products cause humancancers.

THOUGHT PROBLEMS

2027 These data are consistent with the idea that cancer is a multi-step process inwhich cancer-causing changes accumulate over time. The 25-year delaybetween exposure and cancer reflects the time it takes for lung cells to accu-mulate a sufficient number of changes to become cancerous. From otherstudies it is known that cigarette smoke contains tumor initiators and tumorpromoters, both of which contribute to the progression from normal to can-cerous cells.

Your uncles suggestion that there is a genetically predisposed fraction ofthe population that is prone to lung cancer does not match the data. If afixed fraction of the population were genetically predisposed, the incidenceof lung cancer would be relatively constant over time. It would not beexpected to track with per capita smoking.

DATA HANDLING

2028 The highly rearranged karyotypes and their similarity from tumor to tumorsuggest that the cancer cells themselves are being transmitted from devil todevil. It is extremely unlikely that an infectious agent such as a virus or amicroorganism could induce the same set of complicated rearrangementsin different animals. Most importantly, the existence of a chromosome-5inversion in one Tasmanian devil, which is not present in chromosome 5 ofits tumor cells, argues strongly that the tumors are not generated from thehost devils own cells. It appears that this cancer has arisen from a rogue lineof cancer cells, from a tumor of unknown origin, that has acquired the capa-bility for parasitic existence. This is one of just two examples of naturaltransmission of cancer by tumor cells, the other being a venereal disease indogs. A special case of such transmission occurs occasionally during organtransplantation in humans. But the requirements for organ transplanta-tionmatching tissue and immune suppressionhighlight just how

FINDING THE CANCER-CRITICAL GENES A461

unusual natural transmission is. The cancer cells responsible for facialtumors in Tasmanian devils must somehow evade the new hosts immunedefenses.

Reference: Pearse A-M & Swift K (2006) Transmission of devil facial-tumourdisease. Nature 439, 549.

2029A. These results, which were reported in 1933, were the first indication that the

tendency to form breast tumors was maternally transmitted in mice. Theresults of the experiments in Table 201 cannot be readily explained on thebasis of standard types of chromosomal mutations, be they recessive, dom-inant, or X-linked. If you assume that the high strains are homozygous fora recessive mutation and the low strains are homozygous for the wild-typeallele, all the F1 progeny would be heterozygous and unaffected. For a dom-inant mutation, the males, as well as the females, should transmit the muta-tions to their offspring. X-linked mutations would be expected to giveaffected males in every other generation. Although it was unknown at thetime these experiments were carried out, the pattern of inheritance shownin Table 201 matches the expectations for mitochondrial mutations, whichare inherited via the egg cytoplasm. (See Problem 14110 for more discus-sion of these patterns of inheritance.)

B. The key to understanding how the tendency to form breast cancers is inher-ited is the foster-mother experiment. Although presented as an accident inthis problem, it represented a critical insight at the time. Because fostermothers could pass this trait on to mice with which they shared no geneticconnection, inheritance could not be due to chromosomal (or mitochon-drial) mutations. The link between mothers and daughters was identified asthe milk. The so-called milk factor was later identified as a virus, which wenow call mouse mammary tumor virus (MMTV).

References: Little CC (1933) The existence of non-chromosomal influence inthe incidence of mammary tumors in mice. Science 78, 465466.

Bittner JJ (1936) Some possible effects of nursing on the mammary glandtumor incidence in mice. Science 84, 2172.

Paigen K (2003) One hundred years of mouse genetics: An intellectual his-tory. I. The classical period (19021980). Genetics 163, 17.

FINDING THE CANCER-CRITICAL GENES

DEFINITIONS

2030 Oncogene

2031 Transformation

2032 Proto-oncogene

2033 Retinoblastoma

2034 Cancer-critical gene

2035 Tumor suppressor gene

TRUE/FALSE

2036 False. Oncogenes, which are mutated, overactive forms of proto-oncogenes,can be detected in this way. Their addition to the genome can convert a cell


to a cancer cell. By contrast, tumor suppressor genes have their effectsbecause they are inactive; it is their absence that causes cancer. One cannotuse the same sort of transformation assay to detect something that is notthere.

2037 False. It is not that DMBA is a specific mutagen, but rather that the Ras geneis converted to its activated, cancer-causing form by a particular A-to-Talteration that leads to a very specific amino acid change. DMBA causesmutations throughout the genome, but only those at the specific site in theRas gene give rise to cells that have cancerous properties and thus are iden-tified in the assay.

2038 True. An inherited predisposition to cancer often occurs because one copyof a tumor suppressor gene is mutated. Heterozygous cells, which retain onewild-type copy of the gene, are fine, but they are at increased risk for cancerbecause a single event (rather than the two in a normal individual) can inac-tivate the remaining good copy, causing loss of heterozygosity.

THOUGHT PROBLEMS

2039 Oncogenes correspond to stuck accelerators. In their mutated, overactiveform they drive a cell to proliferate in a way that is not responsive to normalcontrols. Defective tumor suppressor genes correspond to broken brakes.They normally function to inhibit steps in signaling pathways; that is, to actas brakes. When tumor suppressor genes are defective, signaling pathwaysare unrestrained. Defective DNA maintenance genes correspond to badmechanics. These genes normally operate to maintain the genome duringits propagation and in the face of DNA damage. Defective DNA maintenancegenes lead to genetic rearrangements or increased point mutations, either ofwhich can convert a proto-oncogene to an oncogene, or eliminate a tumorsuppressor gene.

Reference: Vogelstein B & Kinzler KW (2004) Cancer genes and the pathwaysthey control. Nat. Med. 10, 789798.

2040 Antiproliferative genes such as Rb encode proteins that stop the cell cycle.During normal cell division, these proteins must be turned off. If they wereoverexpressed in all cells, it is likely that the machinery that keeps these pro-teins turned off would be overwhelmed, and cell division would stop. Thus,this cure for cancer might be successful but the patient would be dead.

2041 Cancer cells have additional changes that typically disable cell-cycle check-points and apoptotic mechanisms. In the absence of these regulatory con-trols, which are fully operational in normal cells, overexpression of Mycdrives cell growth and proliferation of cancer cells.

CALCULATIONS

2042A. If the integration events were random, then 0.00005 [fi = (100 kb/2 106 kb)]

would be expected to occur in the target sequence. The fraction of integra-tion events expected to occur outside the target would be 0.99995 (fo = 1 fi).

B. The probability of not finding (PN) a second integration at Il2rg in a surveyof 600 tumors with a retroviral integration at Lmo2, is the probability of notfinding it in one, raised to the power of 600. Thus, PN = (0.99995)600 = 0.97.This means that in 97 out of 100 times you survey a new set of 600 tumors,you would not expect to find a second integration at Il2rg.

C. The probability of finding (PY) a second integration at Il2rg in a survey of 600tumors is 1 minus PN, or 0.03. This means that 3 times out of 100, you wouldexpect to find a second integration at Il2rg.

FINDING THE CANCER-CRITICAL GENES A463

D. Given that only 2 out of 600 tumors actually had a retrovirus integrated atLmo2, the probability of finding a tumor with dual integrations at Lmo2 andIl2rg would be 0.0001 [0.03 (2/600)], or 1 chance in 10,000.

E. The calculation in part D depends on the specific assumptions. If the targetsize were 10 kb (instead of 100 kb), the calculation in part D would decreaseto 1 chance in 100,000. If there were, on average, fewer than 2 retroviral inte-grations per tumor, the calculation in part D would decrease; if there weremore than 2 integrations per tumor, the calculation would increase.

The assumption of randomness of retroviral integration is the most diffi-cult to evaluate in terms of this calculation. Retroviruses are decidedly non-random in their integration, with most varieties showing some degree ofpreference for actively transcribed genes. An entirely different approach tothis question takes such preferences into account. If 2 out 600 tumors had aretroviral integration at Lmo2, and 2 out of 600 had a retroviral integration atIl2rg, then the chance of having both is (2/600) (2/600) = 0.000011, orabout 1 in 100,000 tumors. The chance of not finding a dual integration in600 tumors would be PN = (0.999989)600 = 0.993. Thus, there would be a7/1000 chance of finding a dual integration. This differs by a factor of 100from the result calculated in part D. Yet, both suggest that dual integrationshould be a relatively rare event.

References: Dave UP, Jenkins NA & Copeland NG (2004) Gene therapy inser-tional mutagenesis insights. Science 303, 333.

Hacein-Bey-Abina S, von Kalle C, Schmidt M, Le Deist F, Wulffraat N, McIn-tyre E, Radford I, Villeval J-L, Fraser CC, Cavazzana-Calvo M & Fischer A(2003) A serious adverse event after successful gene therapy for severe com-bined immunodeficiency syndrome. N. Engl. J. Med. 348, 255256.

DATA HANDLING

2043A. Fibroblasts and tumor cells from the same patient have different patterns of

hybridization because the tumor cells have lost portions of the Rb gene. Lossof this gene is a very rare somatic event that affects less than one in a millioncells. Only in the retina does its loss cause uncontrolled growth and tumorformation. No doubt the same proportion of fibroblasts also lose the Rbgene, but its loss from fibroblasts has no known biological consequence, soits absence cannot be readily detected.

B. The fibroblasts from the patient with unilateral retinoblastoma appear to beidentical to those from normal cells, suggesting that the patient with unilat-eral retinoblastoma inherited two good Rb genes. Fibroblasts from thepatient with bilateral retinoblastoma are not normal. Three of the fourrestriction fragments are present at half the normal intensity, suggestingthat one of the Rb genes contains a deletion that encompasses those threerestriction fragments. Note that the three affected fragments are adjacent onthe map of the Rb gene (see Figure 208B).

The tumor cells from both patients are abnormal. The patient with uni-lateral retinoblastoma is missing two fragments entirely and a third is presentat half the normal intensity. This pattern indicates that each copy of the Rbgene has undergone deletion: one deletion encompasses the 9.8-kb and the6.2-kb fragments; the other encompasses these two fragments and the 5.3-kbfragment. The patient with bilateral retinoblastoma is missing three fragmentsentirely and the remaining fragment is present at only half the normal inten-sity. This pattern indicates that the one good Rb gene has been entirelydeleted, leaving only the 6.2-kb fragment from the original inherited deletion.

C. These results are exactly what is expected from the hypothesis thatretinoblastoma is due to the loss of both copies of the Rb gene. Many casesof retinoblastoma have now been examined, and they all show loss or alter-ation of the Rb gene. Thus, retinoblastoma develops in the absence of func-tional Rb.

References: Fung Y-KT, Murphree AL, TAng A, Qian J, Hinrichs SH & Bene-dict WF (1987) Structural evidence for the authenticity of the humanretinoblastoma gene. Science 236, 16571661.

Knudson AG (1971) Mutation and cancer: statistical study of retinoblas-toma. Proc. Natl Acad. Sci. U.S.A. 68, 820823.

2044A. Most of the 463,248 sequence changes that remained after the reading errors

were removed have nothing to do with cancer. The authors of this studyapplied six additional filters to eliminate sequence changes that are unlikelyto contribute to the functional differences between the normal cells and thetumor cells. See if your suggestions are included in this list.1. Filter out changes that do not alter the encoded amino acid sequence; for

example, mutations to synonymous codons. (259,957 changes wereeliminated by this criterion.)

2. Filter out changes that are also present in the DNA from the two normalindividuals that were included in the analysis. (163,006 changes wereeliminated by this criterion.)

3. Filter out changes that correspond to known sequence polymorphismsin the human population. (11,004 changes were eliminated by this crite-rion.)

4. Filter out changes that cannot be confirmed upon reamplifying and rese-quencing the sample. (Of the 29,281 sequence differences that remainedafter applying the above filters, 9295 were not confirmed and thereforeeliminated.)

5. Filter out changes that are also present in normal tissue from the sameindividual that had the tumor. (18,414 of the 19,986 sequence differencesthat remained after applying filter 4 were eliminated were eliminated bythis criterion.)

6. Filter out changes in sequences that have closely related sequences else-where in the genome. There can be problems deciding which genomiclocation is the true source of the sequence read. (265 of the remaining1572 sequence differences were eliminated by this criterion, leaving 1307potential cancer-relevant mutations.)

B. Deciding which of these 1307 mutations are likely to contribute to the can-cers is not an easy task. One approach is to look for mutant genes that arefound in multiple breast tumors or in multiple colorectal tumors. Theunderlying assumption is that similar cancers should have similar sets ofcausative mutations. The authors used this sort of analysis to identifyroughly 12 cancer-related mutations in breast tumors, and about 9 in col-orectal tumors. (The rest of the mutations are likely to be passenger muta-tions.) Because only about half the genes in the genome were analyzed(13,023/25,000), the real number of cancer-relevant mutations in thesetumors may be closer to 20.

C. The sequencing strategy used hereamplifying and sequencing exonswas designed to detect small changes in sequence: point mutations andshort deletions. Larger deletions and gene rearrangements would not bedetected because they would not give an informative PCR product.

Reference: Sjoblom T et al. (2006) The consensus coding sequences ofhuman breast and colorectal cancers. Science 314, 268274.

THE MOLECULAR BASIS OF CANCER-CELL BEHAVIOR

DEFINITIONS

2045 Papillomavirus


2046 p53

2047 Colorectal cancer

TRUE/FALSE

2048 True. That is why oncogenes in their overactive, mutant form tend to drivecell growth and proliferation, and why the loss of tumor suppressor genesremoves natural blocks in these pathways, which also promotes cell growthand proliferation.

2049 False. Activation of a single oncogene is generally not sufficient to convert anormal cell into a cancer cell. Typically, in mice engineered to express anoncogene such as Myc or Ras, cells in some tissues that express the onco-gene show enhanced proliferation, and, over time, occasional cells willundergo further changes to give rise to cancers.

2050 False. Cancer cells rarely have mutations in multiple components of thesame regulatory pathway. Mutation in any one component of a pathway isusually sufficient to inactivate the pathway and promote cancer. Inactiva-tion of more than one component in a pathway would have no benefit forthe cancers evolution.

2051 True. p53 protein normally acts to limit the harm done by DNA damage.Cells that are severely damaged are driven to commit suicide by apoptosis;mildly damaged cells are prevented from dividing until the damage isrepaired. In the absence of p53 these two safeguards are eliminated, allow-ing some cancer cells to proliferate even when exposed to damaging effectsof irradiation and many anticancer drugs.

2052 False. In general, carcinomas show either chromosome instability or defec-tive mismatch repair, but not both. This represents another example of thegeneral principle that cancers have one mutation per pathway: in this casethe pathway for genetic stability. Genetic instability can arise either byincreased rates of chromosome rearrangement or by increased rates of pointmutation. It seems that increasing genetic instability in either way serves theneeds of the evolving cancer cell. Evidently, stimulating both chromosomeinstability and point mutations does not speed up the evolution of the can-cer cell.

THOUGHT PROBLEMS

2053 Mutant B-Raf with glutamate at position 599 is thought to be active becauseit carries a negative charge (COO) in the same region of the protein atwhich a negative charge (PO4) is normally introduced by Ras-inducedphosphorylation. Evidently, the negative charge alters the conformationalequilibrium of B-Raf, shifting it toward the active form. It is common prac-tice these days to test whether a phosphate at a site activates (or inactivates)a protein by engineering a version of the protein with a glutamate at thatposition. Often (but not always) the glutamate change will mimic the effectof the phosphate.

Reference: Wang L, Cunningham JM, Winters JL, Guenther JC, French AJ,Boardman LA, Burgart LJ, McDonnell SK, Schaid DJ & Thibodeau SN (2003)BRAF mutations in colon cancer are not likely attributable to defective DNAmismatch repair. Cancer Res. 63, 52095212.

2054 These observations argue strongly that MMTV generates an oncogene uponintegration into the mouse genome. It is extremely unlikely that MMTVwould integrate so often in the same region of the genome by chance. More-over, a unique transcript is generated in the region of the integrated virus,

THE MOLECULAR BASIS OF CANCER-CELL BEHAVIOR A465

suggesting that a gene is turned on in response to the neighboring viralsequences.

These results summarize the initial characterization of the Int1 locus inmice. It was later shown to be homologous to the wingless locus inDrosophila, both of which are now referred to as Wnt genes. These genessecrete a signaling molecule that triggers a signaling pathway in other cellsthat activates expression of a set of Wnt-responsive genes, some of whichpromote cell proliferation.

Reference: Nusse R & Varmus HE (1982) Many tumors induced by the mousemammary tumor virus contain a provirus integrated in the same region ofthe host genome. Cell 31, 99109.

CALCULATIONS

2055A. Four different haploid gametes can combine to give 16 different diploid

products. Of the several different ways of representing these possibilities,the most concise is the Punnett square, a time-honored tradition withgeneticists (Table 203). However the possibilities are represented, the 16combinations of gametes form 9 distinct genotypes. Four genotypes are rep-resented once (the ones along the diagonal from upper left to lower right inTable 203). One genotype is represented four times (the one along the diag-onal from lower left to upper right). The remaining genotypes are each rep-resented twice (symmetrically arrayed about the diagonal from upper left tolower right). The ratios of the genotypes, along with their expected frequen-cies among 36 progeny, are given in Table 204).

Progeny with the genotypes p53 +/+ Mdm2 / and p53 +/ Mdm2 / areunder-represented in these experiments: none were found. More detailedexperiments have shown that these mice do not survive because they dieearly in embryonic development.

B. It is striking that p53 +/+ Mdm2 / and p53 +/ Mdm2 / do not survive,whereas p53 / Mdm2 / mice do. In the absence of Mdm2, it seems thateven a haploid amount of p53 is lethal. Since Mdm2 is a ubiquitin ligase, onereasonable explanation for this result is that Mdm2 normally keeps the cel-lular concentration of p53 very low. In the absence of Mdm2, p53 is nolonger destroyed at the proper rate and accumulates to levels that are toxic.The basis for cell toxicity in this instance is not entirely clear, but it is thoughtthat the abnormally high amounts of p53 arrest the cell cycle and promoteapoptosis.


Table 203 A Punnett square showing all possible genotypes resulting from randomassortment of p53 Mdm2 gametes (Answer 2055).

FEMALE

GAMETES p53+Mdm2+ p53+Mdm2 p53Mdm2+ p53Mdm2

p53+Mdm2+ p53+Mdm2+ p53+Mdm2+ p53+Mdm2+ p53+Mdm2+

p53+Mdm2+ p53+Mdm2 p53Mdm2+ p53Mdm2

p53+Mdm2 p53+Mdm2 p53+Mdm2 p53+Mdm2 p53+Mdm2


p53Mdm2+ p53Mdm2+ p53Mdm2+ p53Mdm2+ p53Mdm2+


p53Mdm2 p53Mdm2 p53Mdm2 p53Mdm2 p53Mdm2


MA

LE

References: de Oca Luna RM, Wagner DS & Lozano G (1995) Rescue of earlyembryonic lethality in mdm2-deficient mice by deletion of p53. Nature 378,203206.

Jones SN, Roe AE, Donehower LA & Bradley A (1995) Rescue of embryoniclethality in Mdm2-deficient mice by absence of p53. Nature 378, 206208.

DATA HANDLING

2056A. An Arf-knockout mouse would be expected to be more prone to cancer than

a wild-type mouse. In the absence of Arf, Mdm2 would be more active thanin a wild-type mouse. Overactive Mdm2 would, in turn, tend to repress p53activity more than normal. Thus, the consequence of an Arf knockout wouldbe reduced p53 activity. If this lowered activity impaired the ability of p53 toforce abnormal cells to undergo cell cycle arrest or apoptosis, more precan-cerous cells would escape detection and more tumors would form.

B. A p53 +/+ Mdm2 / mouse will not be rescued by knockout of the Arf gene. Inthe absence of Mdm2, the activity of p53 will be maximized. Because Mdm2is absent, the link between Arf and p53 is missing. Thus, no change in Arflevels can affect p53 activity, and p53+/+ Mdm2 / Arf / mice will die just likep53 +/+ Mdm2 / mice do.

C. Mice that express the Myc oncogene will overstimulate Arf activity, whichwill decrease Mdm2 activity, which will cause an increase in p53 activity.Increased p53 activity (so long as it is not increased to the point where it istoxicsee Problem 2055) will tend to increase the mouses ability to forceabnormal cells into cell cycle arrest and apoptosis. This increased activity ofp53 operates in opposition to the proliferation-promoting activity of the Myconcogene. In an Arf +/ mouse there is less Arf, hence, less of a decrease inMdm2 activity and less of an increase in p53 activity. Because the prolifera-tion-promoting activity of the Myc oncogene is opposed to a lesser extent(by the lower p53 activity), Arf +/ mice generate tumors more quickly and dieat a younger age.

References: Quelle DE, Zindy F, Ashmun RA & Sherr CJ (1995) Alternativereading frames of the INK4a tumor suppressor gene encode two unrelatedproteins capable of inducing cell cycle arrest. Cell 83, 9931000.

Zindy F, Williams RT, Baudino TA, Rehg JE, Skapek SX, Cleveland JL, RousselMF & Sherr CJ (2003) Arf tumor suppressor promoter monitors latent onco-genic signals in vivo. Proc. Natl Acad. Sci. U.S.A. 100, 1593015935.

2057A. As discussed in Problem 2043, the absence of a shoulder on any of the three

curves suggests that in all cases only a single event is needed to trigger

THE MOLECULAR BASIS OF CANCER-CELL BEHAVIOR A467

Table 204 Actual and expected genotypes of progeny mice from crosses betweendoubly heterozygous p53+/ Mdm2+/ mice (Answer 2055).

GENOTYPE PROGENY MICE PROGENY MICE PROGENY MICE (NUMBER) (EXPECTED RATIOS) (EXPECTED NUMBERS)

p53+/+ Mdm2+/+ 3 1 2 14p53+/+ Mdm2+/ 5 2 4 12p53+/+ Mdm2/ 0 1 2 14p53+/ Mdm2+/+ 7 2 4 12p53+/ Mdm2+/ 11 4 9p53+/ Mdm2/ 0 2 4 12p53/ Mdm2+/+ 1 1 2 14p53/ Mdm2+/ 7 2 4 12p53/ Mdm2/ 2 1 2 14

tumor production in mice that are already expressing one or both onco-genes.

B. Although the rate of tumor production is much higher in mice with bothoncogenes, activation of the cellular Ras gene cannot be a required event inthe production of tumors in mice that are already expressing the Myc onco-gene. Nor can activation of the cellular Myc gene be a required event in trig-gering tumor formation in mice that are already expressing the Ras onco-gene. As indicated in part A, even when mice express both Myc and Ras,some additional event is required to produce a tumor. If expression of Mycplus Ras were sufficient for tumor formation, then all mice containing bothoncogenes would develop tumors as soon as they passed through pubertyand turned on their expression.

C. The rate of tumor production in mice with both oncogenes is much higherthan expected if the effects of the individual oncogenes were additive. Thus,the two oncogenes together have a synergistic effect on the rate of tumorproduction. As argued in part B, activation of both oncogenes is not suffi-cient to generate a tumor. Thus, the two oncogenes acting together mustopen up a pathway to tumor production that can be triggered by any one ofseveral low-frequency events or that can be triggered by one very commonevent. The nature of the activating events is unclear for these transgenicmice, but analysis of transformed cells suggests that inactivation of the p53tumor suppressor pathway may be responsible.

Reference: Sinn E, Muller W, Pattengale P, Tepler I, Wallace R & Leder P(1987) Coexpression of MMTV/v-Ha-ras and MMTV/c-myc genes in trans-genic mice: synergistic action of oncogenes in vivo. Cell 49, 465475.

2058A. The formation of the three types of patches observed in speckled kernels is

shown in Figure 2017. The formation of a colorless, nonwaxy (c-Wx) patchresults from a breakage that eliminates the dominant color (C) allele (Figure2017A). In the absence of the dominant allele, the color of the patch isdetermined by the recessive colorless (c) allele on the normal chromosome(which is not shown in the figure).

The formation of a colorless, waxy (c-wx) spot in a colorless, nonwaxy (c-Wx) patch is due to a second breakage event that eliminates the dominant


C Wx

C Wx

C-Wx

c-Wx

(A)

C gene has been lost

Wx

Wx

BRIDGEBREAKAGEFUSION

extra copies of C gene

C-Wx

C-C-Wx

(C)

C Wx

C Wx

C C

C C

Wx

Wx


Wx

Wx gene has been lost

Wx

(B)

C-Wx

c-Wx

c-wx


Figure 2017 Formation of three different types of patches observed in speckled kernels (Answer 2058).(A) Formation of a colorless, nonwaxy (c-Wx) spot. (B) Formation of a colorless, waxy (c-wx) spot inside a colorless, nonwaxy (c-Wx) spot formed as in (A). (C) Formation of an intensely colored, nonwaxy (C-C-Wx) spot. Vertical arrows pointing to the dicentric chromosomes show the positions of the breaksthat lead to formation of the patches. In each case the upper half of the starting dicentric chromosomegives rise to the new dicentric chromosome.

nonwaxy (Wx) allele (Figure 2017B). In the absence of the dominant allelethe spot is waxy (wx) due to the recessive allele on the normal chromosome(not shown).

The formation of an intensely colored patch is due to a breakage eventthat leads to a dicentric chromosome with multiple copies of the dominantcolor allele (Figure 2017C). Thus, the genetic constitution of the intenselycolored patch is C-C-Wx.

B. You would never expect to see a colored spot within a colorless patchbecause, once eliminated, the dominant color (C) allele cannot be regainedby further bridgebreakagefusion cycles.

C. You would expect to see colorless spots within an intensely colored patchbecause the dominant color (C) allele could be lost by subsequentbridgebreakagefusion cycles.

The demonstration by McClintock of bridgebreakagefusion cycles inplants was one of the earliest indications that the broken ends of chromo-somes are in some way stickyentirely different from natural chromosomeends. It is clear now that cells have an active repair pathway for joining bro-ken DNA ends together as a defense against potentially lethal double-strandbreaks. So long as breaks are rare, the correct ends are joined. But when mul-tiple breaks are present, the wrong partners can be joined, leading totranslocations or other genetic rearrangements. In humans, such rearrange-ments are often associated with cancers.

Reference: McClintock B (1939) The behavior of successive nuclear divi-sions of a chromosome broken at meiosis. Proc. Natl Acad. Sci. U.S.A. 25,405416.

CANCER TREATMENT: PRESENT AND FUTURE

DEFINITIONS

2059 Gleevec

2060 Multidrug resistance

2061 Gene expression profile

TRUE/FALSE

2062 True. Useful therapies selectively target cancer cells and leave normal cellsrelatively unaffected. This selective action always depends on a key differ-ence between normal cells and cancer cells. For example, most anticancerdrugs and ionizing radiation damage DNA. These treatments preferentiallykill cancer cells because the cancer cells have a diminished capacity to sur-vive the damage.

2063 True. Tamoxifen is an estrogen receptor antagonist. By preventing the bind-ing of estrogen to the receptor, tamoxifen prevents the proliferation of breastcancer cells, which is dependent on estrogen binding.

2064 False. Although there are likely to be some cancer cells in a hypermutablepopulation that are resistant to a single drug, it is unlikely that cells will beresistant to multiple drugs. For such a combination therapy to work, thedrugs must be designed so that a single mutation cannot render cells resis-tant. For example, if the multiple drugs in a cocktail can all be pumped outof the cell by an amplified ABC transporter, then treatment of a cancer cellpopulation will select for cells that have amplified that geneand the treat-ment will fail. If, on the other hand, a single mutation cannot render a cellresistant to all of the drugs in the cocktail, the treatment has the potential foreliminating all the cancer cells and producing a cure.

CANCER TREATMENT: PRESENT AND FUTURE A469

THOUGHT PROBLEMS

2065 Technological advances in our ability to detect cancers since 1970 meansthat we can diagnose them at earlier stages in the course of the disease.Thus, even in the absence of better treatment regimes, a patient might beexpected to survive somewhat longer now than in the past because they willbe at a slightly earlier stage of the disease at the 5-year mark.

Reference: Weinberg RA (2006) The Biology of Cancer, p726. New York: Gar-land Science.

2066 The promyelocytes of APL are blocked at an intermediate stage in theirdevelopment, at a point where they still divide and increase in number. It isthis unchecked increase in number that causes problems for the cancerpatient. Normally, such precursor cells divide only a few times before theyterminally differentiate into a nondividing blood cell. By triggering the dif-ferentiation of promyelocytes into terminally differentiated neutrophils,which no longer divide, treatment with all-trans-retinoic acid eliminates theproblems caused by unchecked proliferation.

APL arises by one of a few types of translocation that fuses the retinoic acidreceptor (RAR) gene on chromosome 17 with a gene on another chromo-some to make a hybrid protein that interferes with the normal developmen-tal program. It is not yet clear how the fusion protein blocks development,although it likely does so by interfering with the function of the normal RARreceptor. In some way, treatment with all-trans-retinoic acid allows APL cellsto move through the block.

Reference: Warrell RP Jr, de The H, Wang Z-Y & Degos L (1993) Acute prom-yelocytic leukemia. N. Engl. J. Med. 329, 177189.

2067 The products of oncogenes are the only feasible targets for such smallmolecules. The product of an oncogene has a dominant, growth-promotingeffect on the cell. Thus, if the growth-promoting oncogene product wereinhibited, the cell might return to a more normal state. This is the underly-ing rationale for searching for drugs that inhibit oncoproteins.

By contrast, the products of tumor suppressor genes and DNA mainte-nance genes are not targets for anticancer drug development. These twoclasses of gene cause cancer by not making their product. Thus, there is noabnormal product to be inhibited in cancer cells that arise by mutation oftumor suppressor or DNA maintenance genes.

2068A. The 1-hour incubation allows the binding reactionstest compound to

kinase and ATP analog to kinaseto come to equilibrium. Making the mea-surements at equilibrium gives a much more reproducible assay and allowscomparisons between test compounds to be made on an equal footing.

B. At equilibrium, some of the phage-attached kinases will be bound to the testcompound and some will be bound to the ATP-analog-coated magneticbeads. The relative proportions depend on how strongly the test compoundbinds and its concentration. In the presence of a weakly binding test com-pound, most of the phage will be attached to the magnetic beads, which willyield a high count in the plaque assay. In the presence of the same concen-tration of a strongly binding test compound, most of the phage will beattached to the test compound and will be washed away at the end of theincubation. Thus, strongly binding test compounds will give a low count inthe plaque assay.

C. The assay is set up to measure competition between the ATP analog and thetest compound for the ATP-binding site on the protein kinase. Thus, it isexpected that the test compounds that score well in this assay will bind at ornear the ATP-binding cleft on the kinase. This assay can also identifymolecules that bind to other sites (allosteric sites) but alter the conforma-tion of the ATP-binding site, so that the kinase no longer binds ATP. One


might imagine other sites where a useful inhibitor might bind to a kinase toinhibit its functionat a key proteinprotein interface, for examplebutthose types of inhibitors will not be picked up in this assay system.

D. Although all protein kinases bind ATP in an evolutionarily conserved bind-ing site, the binding sites are not identical. The amino acid differences inand around the binding site provide slightly different binding surfaces thatcan be exploited to develop compounds that selectively bind to one kinasebut not others.

References: Griffin JD (2005) Interaction maps for kinase inhibitors. Nat.Biotechnol. 23, 308309.

Fabian MA et al. (2005) A small molecule-kinase interaction map for clinicalkinase inhibitors. Nat. Biotechnol. 23, 329336.

CALCULATIONS

2069A. The Kd values can be estimated directly from the graph in Figure 2015, as

the concentration of test compound that give a half-maximal response. TheKd values for BIRB-796, VX-745, and SB203580 are about 0.3 nM, 3 nM, and15 nM, respectively.

B. The phage concentration is 17 pM, well below the Kd(test) values for the com-pounds tested, the lowest of which is 0.3 nM (300 pM). Thus, the estimatesin part A are valid. Phage concentration = (1010 phage/mL) (mole/6 1023phage) (103 mL/L) (1012 pmol/mole) = 17 pmol/L = 17 pM.

Reference: Fabian MA et al. (2005) A small molecule-kinase interaction mapfor clinical kinase inhibitors. Nat. Biotechnol. 23, 329336.

DATA HANDLING

2070A. Iressa, which is being used clinically in the treatment of non-small-cell lung

carcinomas, and Gleevec have similar specificities. Iressa has fewer off-tar-get binding interactions, and binds only one of those with an affinity than iswithin a factor of 10 of the main target. BIRB-796 binds to more off-targetproteins than Gleevec or Iressa, and it binds three of them with only 10-foldlower affinity. Staurosporine, which is the least specific of all, is a potentinhibitor of many protein kinases.

B. The clustering of the binding targets on the kinome is expected. After all, thekinases that are most closely related are the closest together on the tree.Closely related kinases should have more similar binding sites, and thusmight be expected to bind an inhibitor similarly. Indeed, it is the similarityof all kinases, and especially of closely related kinases, that makes drugdevelopment for them such a challenge.

C. There is no way that you (or anyone) could predict from these data thatBIRB-796 would bind Abl(T315I). Resistant variants arise by such subtlechanges that direct measurements must be made.

D. This high-throughput screen can be adapted to look for drugs that bind resis-tant variants of protein kinases. If the particular mutant versions responsiblefor resistance are identified, they can be included in the screen to find drugsthat are active against them. It is hoped that screens such as the onedescribed here can be used to identify a second generation of drugs that willprove beneficial in the treatment of cancers resistant to the primary drug.

Reference: Fabian MA et al. (2005) A small molecule-kinase interaction mapfor clinical kinase inhibitors. Nat. Biotechnol. 23, 329336.

CANCER TREATMENT: PRESENT AND FUTURE A471