6
`
`
`
`Human Tumor Xenograft Models
`in NCI Drug Development
`
`jacqueline Plowman, PhD,
`Donald]. Dykes, BS,
`Melinda Hollingshead, PhD,
`Linda Simpson-Herren, BS,
`and Michael C. Alley, PhD
`
`CONTENTS
`
`INTRODUCTION
`
`HISTORICAL DEVELOPMENT OF NCI SCREENS
`
`HUMAN TUMOR XENOGRAFT MODELS IN CURRENT USE
`
`HOLLOW FIBER ASSAYS: A NEW APPROACH
`TO IN VIVO DRUG TESTING
`
`SUMMARY
`
`1 . INTRODUCTION
`
`The preclinical discovery and development of anticancer drugs by the NCI consist
`of a series of test procedures, data review, and decision steps that have been sum-
`marized recently (1). Test procedures are designed to provide comparative quantita-
`tive data, which in turn, pemiit selection of the best candidate agents from a given
`chemical or biological class. Periodic, comprehensive reviews by various NCI com-
`mittees serve not only to identify and expedite the development of active lead com-
`pounds that may provide more efficacious treatments for human malignancy, but also
`to eliminate agents that are inactive and/or highly toxic from further consideration.
`Various components in NCI’s drug discovery and development process have evolved
`in response to a combination of factors—scientific, clinical, technological, and fiscal.
`A series of review articles have charted the evolution of the drug screening program
`and have described specific elements of the process, e.g., acquisition, screening, ana-
`log development and testing, pharmacology, and toxicology (2—12). The present
`chapter provides: a brief history of the in vivo screens used by NCI; a description of
`the human tumor xenograft systems, which are currently employed in preclinical drug
`development; a discussion of how these xenograft models are employed for both ini-
`tial efficacy testing as well as detailed drug evaluations; and a description of a new
`model that may facilitate preclinical drug development.
`
`From: Anticancer Drug Development Guide: Preclinical Screening, Clinical Trials, and Approval
`Edited by: B. Teicher Humana Press Inc., Totowa, NJ
`
`101
`
`Genentech 2106
`
`Hospira v. Genentech
`|PR2017-00737
`
`Genentech 2106
`Hospira v. Genentech
`IPR2017-00737
`
`

`

`102
`
`Part II / In Vivo Methods
`
`2. HISTORICAL DEVELOPMENT OF NCI SCREENS
`
`Analyses of various screening methods available prior to 1955 indicated that (1)
`nontumor systems were incapable of replacing tumor systems as screens, and (2) no
`single tumor system was capable of detecting all active antitumor compounds (13).
`Since that time, the preclinical discovery and development of potentially useful anti-
`cancer agents by the NCI have utilized a variety of animal and human tumor models
`not only for initial screening, but also for subsequent studies designed to optimize
`antitumor activity of a lead compound or class of compounds. Although the various
`preclinical data review steps and criteria have remained essentially the same through-
`out the years, the modes and rationale of in vivo testing employed by NCI have evolved
`significantly.
`
`2.1. Murine Tumor Screens, 1955—1975
`
`In 1955, NCI initiated a large-scale in vivo anticancer drug screening program
`utilizing three murine tumor models: sarcoma 180, L1210 leukemia, and carcinoma
`755. By 1960, in vivo drug screening was performed in L1210 and in two additional
`rodent models selected from a battery of 21 possible models. In 1965, screening was
`limited to the use of two rodent systems, L1210 and Walker 256 carcinosarcoma. In
`1968, synthetic agents were screened in L1210 alone, whereas natural product testing
`was conducted in both L1210 and P388 leukemias. A special testing step was added to
`the screen in 1972 to evaluate active compounds against B16 melanoma and Lewis
`lung carcinoma. It is noteworthy that this first 20 years of in vivo screening relied
`heavily on testing conducted in the L1210 model.
`
`2.2. Prescreen and Tumor Panel, 1976—1986
`
`In late 1975, NCI initiated a new approach to drug discovery that involved pre-
`screening of compounds in the ip-implanted murine P388 leukemia model, followed
`by evaluation of selected compounds in a panel of transplantable tumors (14). The
`tumors in the panel were chosen as representative of the major histologic types of
`cancer in the US and, for the first time in NC] history, included human solid tumors.
`The latter was made possible through the development of immunodeficient athymic
`(nu/nu) mice and transplantable human tumor xenografts1n the early 1970s (15,16)
`Beginningin 1976, the tumor panel consisted of paired murine and human tumors of
`breast (CD8F. and MX-1), colon (colon 38 and CX—l [the same as HT29]) and lung
`(Lewis and LX-l), together with the 816 melanoma and L1210 leukemia used in
`previous screens.
`The majority of the early NCI testing conducted with the human tumors used
`small fragments growing under the renal capsule of athymic mice. The subrenal cap-
`sule (src) technique and assay were developed by Bogden and associates (1 7). Although
`labor-intensive, the src assay provided a rapid means of evaluating new agents against
`human tumor xenografts at a time when the testing of large numbers of compounds
`against sc xenografts seemed untenable. As experience was gained with the husbandry
`of athymic mice, longer-duration sc assays became manageable.
`A detailed evaluation of the sensitivities of individual tumor systems employed
`from 1976-1982 revealed a wide range in sensitivity profiles as well as “yield” of
`active compounds (14). The data clearly indicated that rodent models may not be
`capable of detecting all compounds with potential activity against human malignan-
`
`

`

`
`
`Chapter 6 / Human Tumor Xenograft Models 103M
`
`cies, and indicated that the best strategy for testing is to employ a combination of
`tumor systems to minimize loss of potentially useful compounds. These findings
`prompted the NCI in 1982 to develop a strategy for testing compounds that involved
`a sequential process of “progressive selection”: NCI continued to use the P388 leu-
`kemia as a prescreen, but subsequent evaluation of selected agents was conducted in a
`modified tumor panel composed of “high-yield” models from the original panel (i.e.,
`src-implanted MX-l mammary carcinoma, and ip-implanted Blé melanoma and
`L1210 leukemia) and a new model, the ip-implanted M5076 sarcoma. Thereafter,
`evaluation of selected compounds would be “compound-oriented” and use protocols
`and models, selected on the basis of prior testing results and known properties of each
`compound, that would present the compound with increased biological and pharma-
`cological challenge.
`Alternate approaches to in vivo drug evaluation have been prompted by investiga-
`tions on the metastatic heterogeneity of tumor cell populations. During the 19805,
`several investigators associated with NCI conducted studies to assess the metastatic
`potential of selected murine and human tumor cell lines (B16, A-375, LOX-IMVI
`melanomas, and P03 prostate adenocarcinoma) and their suitability for experimen-
`tal drug evaluation (e.g., 18—21). A series of investigations by Fidler and associates
`demonstrated that metastasis is not random, but selective and that metastasis consists
`of a progression of sequential steps, the pattern of which is dependent on injection
`site (22,23). Such findings support the importance of establishing in vivo models
`derived from the implantation of tumor material in host tissues that are anatomically
`corre —“seed” and “soil” compatibility. Such “orthotopic” models also have been
`developed and utilized to study lung cancer (e.g., 24), breast cancer (e.g., 25), and
`prostate cancer (26). Although it may not be possible for NCI to employ these models
`in the initial steps of in vivo drug evaluations, such models may be well suited for
`subsequent, more detailed evaluation of compounds that exhibit activity against
`specific tumor types. Metastases and orthotopic models are discussed in greater detail
`in Chapters 7 and 8 of this volume.
`
`2.3. Human Tumor Colony Formation Assay, 1981—1985
`
`Based on initial reports by Salmon and colleagues (27,28), various clinical invest-
`igators working with fresh human tumor samples from patients and/or with early
`passage human tumor xenograft materials utilized various culture techniques to iden-
`tify chemotherapeutic agents active against human malignancies (e.g., 29,30). The
`NCI sponsored a pilot drug screening project utilizing a human tumor colony-form-
`ing assay (HTCFA) at multiple clinical cancer centers. Although it was possible to
`identify unique antitumor drug “leads” using such a technique, the HTCFA could be
`employed only for a limited number of tumor types and was not found suitable for
`large-scale drug screening (31).
`
`2.4. Human Tumor Cell Line Screen, 1985-Present
`
`In 1985, the NCI initiated a new project to assess the feasibility of employing
`human tumor cell lines for large-scale drug screening (12; also see Chapter 2 of this
`volume). Cell lines derived from seven cancer types (brain, colon, leukemia, lung, mel-
`anoma, ovarian, and renal) were acquired from a wide range of sources, cryopre-
`served, and subjected to a battery of in vitro and in vivo characterizations, including
`
`

`

`
`
`104 Part II / In Vivo Methods
`
`testing in drug sensitivity assays. The approach was deemed suitable for large-scale
`drug screening in 1990 (I). With the implementation of a 60-member cell line in vitro
`screen, in vivo testing procedures were substantially altered as discussed below.
`
`3. HUMAN TUMOR XENOGRAFT MODELS IN CURRENT USE
`
`The new in vitro human tumor cell line screen shifted the NCI screening strategy
`from “compound-oriented” to “disease-oriented” drug discovery (12). Compounds
`of interest identified by the screen (e.g., those demonstrating disease-specific .dif-
`ferential cytotoxicity) were to be considered “leads,” requiring further preclinical
`evaluation to determine their therapeutic potential. As part of this followup testing,
`the antitumor efficacy of the compounds was to be evaluated in in vivo tumor models
`derived from the in vitro tumor lines used in the screen. Although only a subset of cell
`lines, selected on the basis of in vitro sensitivity, would be used for each agent, it was
`anticipated that for any selected compound, any cell line might be required as a xeno-
`graft model. In order to accomplish such an objective, a concerted developmental
`effort was required to establish a battery of human tumor xenograft models. As
`discussed below and elsewhere (32), tumorigenicity was demonstrated for the majority
`of the tumor lines utilized in the in vitro screen that became fully operational in April
`1990 (1). Then, in 1993, composition of the cell line screen was modified: cell lines
`with variable grth characteristics and those providing redundant information were
`replaced by groups of prostate and breast tumor lines. As a consequence, additional
`xenograft model development was initiated for prostate and breast cancers.
`
`3.1. Development of Human Tumor Xenografts
`Efforts focused on the establishment of sc xenografts from human tumor cell-
`culture lines obtained from the NCI tumor repository at Frederick, MD. The approach
`is outlined in Fig. l. The cryopreserved cell lines were thawed, cultured in RPMI 1640
`medium supplemented with 10% heat-inactivated fetal bovine serum (HyClone), and
`expanded until the population was sufficient to yield 210' cells. Cells were har-
`vested and then implanted sc into the axillary region of 10 athymic NCr nu/nu mice
`(1.0 x 107 cells/0.5 mL/mouse) obtained from the NCI animal program, Frederick,
`MD. Mice were housed in sterile, polycarbonate, filter-capped MicroisolatorTM cages
`(Lab. Products, Inc.), maintained in a barrier facility on 12-h light/dark cycles, and
`provided with sterilized food and water ad libitum. The implanted animals were
`observed twice weekly for tumor appearance. Growth of the solid tumors was moni-
`tored using in situ caliper measurements to determine tumor mass. Weights (mg) were
`calculated from measurements (mm) of two perpendicular dimensions (length and
`width) using the formula for a prolate ellipsoid and assuming a specific gravity of
`1.0 g/cm’— (33). Fragments of these tumors were subjected to histological, cytochemi-
`cal, and ultrastructural examination to monitor the characteristics of the in vivo
`material and to compare them with those of the in vitro lines and, where possible,
`with those reported for initial patient tumors (34). Both in vitro and in vivo tumor
`materials exhibited characteristics consistent with tissue type and tumor of origin.
`However, not unexpectedly, differences in the degree of differentiation were noted
`between some of the cultured cell lines and corresponding xenograft materials.
`The initial solid tumors established in mice were maintained by serial passage of
`30—40 mg tumor fragments implanted sc near the axilla. There was an apparent cell
`
`

`

`Chapter 6 / Human Tumor Xenograft Models
`
`105
`
`Cell lines obtained from NCI Tumor Repository
`10‘ cells/vial
`
`Cell lines started in vitro and expanded to 10" cells
`
`107 Cells implanted sc per mouse
`
`
`
`No growth;
`Try minimum
`of 3 times
`
`
`
`Poor growth;
`
`
`Low take rate;
`Continue to pass
`
`
`
`Good growth;
`Take rate, etc.
`
`generation
`
`Carry in passage
`at least 3
`
`generations
`before using in
`drug studies
`
`
`
`
`Start replacement
`from frozen stock
`
`
`circa 10th
`
`
`
`
`Fig. 1. Schematic of the development of in vivo models for drug evaluation.
`
`population selection occurring in some of the tumors as they adapted to grth in
`animals during early in vivo passage, with growth rates increasing appreciably in
`sequential passages (32). Thus, xenografts were not utilized for drug evaluation until
`the volume-doubling time stabilized, usually around the fourth or fifth passage. The
`doubling time of xenografts derived from tumor cell lines constituting both the initial
`(1990) and the modified (1993) human tumor cell line screens, plus three additional
`breast tumors, are presented in Table 1. Also provided in the table is information on
`the take-rate of the tumors, and the experience of the NCI in the use of the tumors as
`early stage sc models. The doubling times were determined from vehicle-treated con-
`trol mice used in drug evaluation experiments (only data for passage numbers 4—20
`have been included). For each experiment, the doubling time is the median of the time
`interval for individual tumors to increase in size from 200—400 mg (usually a period of
`exponential growth). Both ranges and mean values are provided to demonstrate the
`inherent variability of growth for some of the xenograft materials even after a period
`of stabilization. Mean doubling times range from < 2 d for five tumors (SF-295 glio-
`blastoma, MOLT 4 leukemia, DMS 273 small-cell lung tumor, and LOX-IMVI and
`SK-MEL-28 melanomas) to > 10 d for the MALME-3M‘and M19-MEL melanomas.
`Difficulty was experienced in establishing and/or using some of the sc models. For
`example, even though HOP-62 nonsmall—cell lung tumors exhibited good growth
`rates, poor take-rates of 70, 50, 64, and 30% attained in the second through fifth pas—
`sages, respectively, precluded their use for experimental drug testing. Although serial
`passage of OVCAR-3 ovarian tumors from sc-implanted fragments was difficult,
`
`

`

`106
`
`Part II / In Vivo Methods
`
`Growth Characteristics of sc-Implanted Human Tumor Xenografts
`
`Table 1
`
`Tumor
`
`Origin
`
`Line
`
`In vitro panel
`fiatus
`1990
`1993
`
`Mean volume
`
`doubling time
`(range) in days'
`
`Take
`Rate”
`
`Opinion for use
`as early-stage
`sc model
`
`Colon
`
`CNS
`
`SW-620
`KM 12
`HCT-l 16
`HCT- l 5
`HCC-2998
`DLD-l
`KM20L2
`COLO 205
`HT29
`
`SF-295
`SNB-75
`U251
`XF 498
`SNB-l9
`SF-539
`SF-268
`SNB-78
`
`Leukemia
`
`MOLT-4
`
`Lung:
`non-sm all
`cell
`
`HL-60(TB)‘
`CCRF-CEM
`SR
`RPMI-8226
`K-562
`
`NCI-H460
`NCI-H522
`HOP-62
`NCI-H23
`NCI-H322M
`EKVX
`HOP-92
`A549/ATCC
`HOP-1 8
`NCI-H266
`
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`
`Yes
`Yes
`Yes
`Yes
`Yes
`No
`No
`Yes
`Yes
`
`Yes
`Yes
`Yes
`No
`Yes
`Yes
`Yes
`No
`
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`No
`Yes
`
`2.4(l.7-3.9)
`2.4(1.9-3.3)
`2.6(1.8-3.4)
`3.4(l.8-5.0)
`3.5(2.4-7.7)
`3.8(3.l-5.5)
`3.9(2.5-5.4)
`4.3(2.4-8.9)
`5.1(2.4-7.6)
`
`Good
`Good
`Good
`Good
`Good
`Good
`Good
`Good
`Good
`
`Good
`Good
`Good
`Good
`Acceptable
`Acceptable
`Acceptable
`Acceptable
`Acceptable
`
`l.4(1.Q-2.0)
`3.1(2.0-4.6)
`4.3(2.4-8.9)
`4.4(2.6-8.3)
`6.9(3.l-4.4)
`8.4(one only)
`NA°
`NA
`
`Good
`Good
`Good
`Good
`Good
`Good
`Not Acceptable
`60-70%
`Not Acceptable
`60-70%
`Not Acceptable
`70%
`Minimal growth
`NA
`No Growth
`NA
`
`1.2(2.0-5.6)
`3.3(2.1-4.9,ip)
`4.6(4.3-4.6)
`5.1(one only)
`NA
`NA
`
`Acceptable
`80—100%
`Good(ip)
`85-100%(ip)
`Acceptable
`60-80%
`Not Acceptable
`80%
`Minimal growth
`NA
`Minimal growth
`NA
`
`2.1(1.3-3.0)
`2.3(1.0-3.4)
`3.6(3.3-3.8)
`3.7(2.0-6.4)
`4.0(2.7—5 .9)
`5.5(3.5-7.9)
`6.0(5.l-8.4)
`8.4(5.8-10.9)
`NA
`NA
`
`Good
`Good
`Good
`Good
`Not Acceptable
`30-65%
`Good
`Good
`Acceptable
`Good
`Acceptable
`Good
`Accepable
`Good
`Not Acceptable
`70-80%
`Minimal growth
`NA
`Minimal growth
`NA
`
`tumors grew more readily from so implants of brei derived from ip-passaged material.
`The HL-60 (TB) promyelocytic leukemia did not grow well sc, but an ascitic ip line
`with a good take-rate was established successfully (Table 1). Growth characteristics of
`sc-implanted RXF 393 renal tumors are perhaps better suited for evaluation of a
`survival end point than for measurements of tumor size. Although demonstrating
`
`

`

`107
`Chapter 6 / Human Tumor Xenograft Models
`
`Table 1 (Continued)
`
`
`Opinion for use
`Mean volume
`In vitro panel
`as early-stage
`Take
`doubling time
`status
`Tumor
`Origin
`Line
`1990
`1993
`(range) in days'
`Rateb
`sc model
`
`
`Lung:
`small cell
`
`DMSZ73
`DMSl 14
`
`Yes
`Yes
`
`Mammary
`
`No
`ZR-75-1
`No
`MX-l
`UISO-BCA-l No
`MDA MB-
`No
`23l/ATCC
`
`No
`MCF7°
`MCF7/ADR- No
`RES
`
`No
`No
`
`No
`No
`No
`Yes
`
`Yes
`Yes
`
`l.7(1.6-2.l)
`4.8(2.8-7.5)
`
`Good
`75-90%
`
`Good
`Acceptable
`
`l.8(l.5-l.9)
`2.7(2.2-3.0)
`4.1(2.8—4.8)
`4.4(2.7-7.7)
`
`4.5(2.2-8.0)
`6.1(4.2-7.9)
`
`Good
`Good
`Good
`Good
`
`Good
`Good
`
`Good
`Good
`Acceptable
`Acceptable
`
`Acceptable
`Acceptable
`
`No
`
`Yes
`
`6.6(2.8-13.6)
`
`Good
`
`Acceptable
`
`Yes
`Yes
`Yes
`Yes
`
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`No
`
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`
`7.9(4.5-10.2)
`NA
`NA
`NA
`
`Good
`Minimal growth
`No growth
`No growth
`
`Acceptable
`NA
`NA
`NA
`
`l.5(l.l-2.l)
`l.9(l.l~2.5)
`2.8(l.8-4.2)
`5.4(3.8-7.7)
`5.7(4.8-6.6)
`6.7(2.8-12.7)
`7.3(5.l-8.2)
`ll.2(7.l-16.9)
`12.3(8.7-l6.8)
`
`Good
`Good
`70-80%
`Good
`80-90%
`Good
`Good
`80-90%
`60-90%
`
`Good
`Good
`Not Acceptable
`Acceptable
`Not Acceptable
`Acceptable
`Acceptable
`Not Acceptable
`Not Acceptable
`
`3.3(2.2-4.3)
`3.4(2.6-4.9)
`5.5(5.0-5.9)
`6.2(one only)
`6.4(5.3-8.6)
`12.2(1 1.2-13.0)
`
`Good
`Good
`Good
`70-100%
`Good
`70%
`
`Good
`Good
`Acceptable
`Acceptable
`Acceptable
`Not Acceptable
`
`(continued )
`
`MDA-MB-
`435
`
`MDA-N
`HSS78T
`BT-549
`T47D
`
`No
`No
`No
`No
`
`Melanoma
`
`Ovarian
`
`Yes
`LOX-IMVI
`Yes
`SK-MEL-28
`Yes
`UACC-62
`Yes
`UACC-257
`Yes
`SK-MEL-2
`Yes
`M14
`Yes
`SK-MEL-S
`MALME-BM Yes
`Ml9-MEL
`Yes
`
`Yes
`OVCAR—S
`Yes
`SK-OV-3
`OVCAR-Bf Yes
`0VCAR-4
`Yes
`IGROVl
`Yes
`OVCAR-8
`Yes
`
`good initial growth, the RXF 393 tumors cause death in mice with low tumor burden,
`probably owing to paraneoplastic mechanisms. Other cell lines failed to become func-
`tional in vivo tumors, including two CNS, two nonsmall-cell lung, three breast, three
`renal tumor lines, and two leukemias, although minimal in vivo growth was observed
`with 9 of these 12 cultured lines (Table 1). With more extensive studies, it might be
`possible to attain improved tumor take-rates and growth by implanting tumors in
`
`

`

`108
`
`Part II / In Vivo Methods
`
`Table 1 (Continued)
`
`Tumor
`Origin
`
`Line
`
`In vitro panel
`status
`1990
`1993
`
`Mean volume
`doubling time
`(range) in days'
`
`Prostate
`
`Renal
`
`PC-3
`DU-l45
`
`CAKI-l
`RXF 631
`A498
`RXF 393
`SN12C
`786-0
`ACHN
`UO-31
`TK-lO
`
`No
`No
`
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`
`Yes
`Yes
`
`Yes
`No
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`Yes
`
`2.4(1.5-3.9)
`4.4(2.0-7.9)
`
`2.1(1.3-2.5)
`3.3(l.5-6.8)
`3.4(2.2-4.3)
`3.4(2.3-5.7)
`5 .6(3 .2-1 1.4)
`6.7(one only)
`NA
`NA
`NA‘
`
`Take
`Rate”
`
`Good
`Good
`
`Opinion for use
`as early-stage
`sc model
`
`Good
`Acceptable
`
`Good
`Good
`Acceptable
`Good
`Acceptable
`Good
`Good
`Good
`Acceptable
`Good
`Not Acceptable
`80%
`Minimal growth
`NA
`Minimal growth
`NA
`No growth
`NA
`
`“ Time for tumors to increase in size from 200-400 mg. Data are compiled from experiments using
`passage numbers 4-20. Tumors are listed in order of increasing mean doubling time/histologic type.
`”Good: reproducible take-rate of z 90%.
`‘ NA: not applicable.
`4 Based on ip implant of 1.0 x 10’ cells.
`eMCF7 growth in athymic NCr nu/nu mice requires l7B-estradiol supplementation.
`fLirnited sc data obtained from implant of 0.5 mL 25% brei derived from ip-passaged tumor:
`poor growth is attained with serial passage of fragments from sc tumors.
`
`severe combined immunodeficient (SCID) mice (scid/scid) (35). As discussed below,
`tumor take for some human lymphoma lines was markedly superior in SCID mice
`compared to athymic (nu/nu) (36) or triple-deficient BNX (bg/nu/xid) mice (37).
`Establishment of breast tumor xenografts in vivo raised issues concerning hor-
`monal requirements for growth of these tumors. For example, the importance of
`hormones in the growth of MCF7 breast carcinoma cells as solid tumors in athymic
`mice has been described (38). Our experience with this tumor also has shown the
`importance of 17 B-estradiol supplementation for the growth of the sc-irnplanted
`MCF7; 60-d release, 17 B-estradiol pellets (Innovative Research of America) are
`implanted sc in athymic mice 24 h prior to implanting MCF7 fragments in all NCI
`studies with this tumor. Growth of the remaining breast tumor xenografts appeared
`to be independent of estradiol supplements. For example, growth curves of individual
`early passage ZR-75-l tumors implanted into athymic mice receiving either estradiol
`or no estradiol supplements completely overlapped (Fig. 2A), even though in this
`early passaged material, there was a large variation in the time postimplant for growth
`of individual tumors to be observed. The independence of ZR-75-l tumor growth
`from estradiol supplements is further illustrated in Fig. ZB. Early stage vehicle-treated
`control tumors in mice receiving no estradiol supplementation demonstrate rapid
`growth (median doubling time 1.9 d) soon after implantation. The estrogen receptor
`(ER) status of the in vivo passaged ZR-75-1 tumors has not been determined, but the
`
`

`

`
`
`Chapter 6 / Human Tumor Xenograft Models 109
`
` '3
`
`10
`
`10
`
`10
`
`102
`
`10‘
`
`£103
`“EU'
`'3
`
`3 E
`
`3
`|_
`
`o
`
`10
`
`20
`
`:50
`
`0
`
`10
`
`20
`
`:50
`
`Days Poatlmplant
`
`Days Poatimplant
`
`Fig. 2 Growth comparison of individual ZR-75-l human breast tumor xenografts with and without
`estradiol supplementation. A. Passage number 5, -- with 1.7 mg sc l7B-estradiol pellet implants;
`—without estradiol supplementation. B. Passage number 14, without estradiol supplementation.
`
`growth characteristics of the tumors suggest an ER — status even though early evalua-
`tions of the in vitro cell line indicated an ER+ status (39).
`The in vivo grth characteristics of the xenografts determine their suitability for
`use in the evaluation of test agent antitumor activity, particularly when the xenografts
`are utilized as early stage so models. For the purposes of the current discussion, the
`latter model is defined as one in which tumors are staged to 63—200 mg prior to the
`initiation of treatment. Our experience with the suitability of the xenografts as early
`stage models is listed in Table 1. Growth characteristics considered in rating tumors
`include take-rate, time to reach 200 mg, doubling time, and susceptibility to spontan-
`eous regression. As can be noted, the faster-growing tumors tend to receive the higher
`ratings.
`Since non-Hodgkin’s lymphoma is one of the two principal malignancies occurring
`in the growing population of HIV-infected persons (40), the Developmental Thera-
`peutics Program (DTP) has also established a group of human lymphoma xenografts
`for evaluating potential chemotherapeutic agents (41 ). This includes AS283, an Epstein-
`Barr virus (EBV)-positive, HIV-negative Burkitt’s lymphoma derived from an AIDS
`patient (42); KD488, an EBV-negative, pediatric Burkitt’s lymphoma (43—46); and
`RL, a diffuse, small noncleaved B-cell lymphoma (47). These lines grow sc with take-
`rates in excess of 90% in SCID mice, whereas much lower take-rates occur in athymic
`or triple-deficient BNX mice. Our finding of greater take-rates for the human leu-
`kemias/lymphomas in SCID mice compared to athymic mice is consistent with the
`known capacity of SCID mice to support xenografts of normal human hematopoietic
`cells (48).
`
`3.2. Advanced-Stage sc Xenograft Models
`
`Advanced-stage sc-implanted tumor xenograft models were established originally
`for use in evaluating the antitumor activity of test agents, so that clinically relevant
`
`

`

`1 10
`
`Part II / In Vivo Methods
`
`parameters of activity could be determined, i.e., partial and complete regressions,
`durations of remission (49-51). Tumor grth is monitored, and test agent treatment
`is initiated when tumors reach a weight range of 100-400 mg (staging day, median
`weights approx 200 mg), although depending on the xenograft, tumors may be staged
`at larger sizes. Tumor size and body weights are obtained approximately 2 times/wk
`and entered into DTP’s DEC 10,000 Model 720 AXP computer. Through software
`programs developed by staff of the Information Technology Branch of DTP, in par-
`ticular by David Segal and Penny Svetlik, data are stored, various parameters of ef-
`fect are calculated, and data are presented in both graphic and tabular formats.
`Parameters of toxicity and antitumor activity are defined as follows:
`
`1. Parameters of toxicity: Both drug-related deaths (DRDs) and maximum percent relative
`mean net body weight losses are determined. A treated animal’s death is presumed to be
`treatment-related if the animal dies within 15 d of the last treatment, and either its
`tumor weight is less than the lethal burden in the control mice, or its net body weight
`loss at death is 20% greater than the mean net weight change of the controls at death or
`sacrifice. A DRD also may be designated by the investigator. The mean net body weight
`of each group of mice on each observation day is compared to the mean net body weight
`on staging day. Any weight loss that occurs is calculated as a percent of the staging day
`weight. These calculations also are made for the control mice, since tumor growth of
`some xenografts has an adverse effect on the weight of the mice.
`Optimal "lo T/C: Changes in tumor weight (A weights) for each treated (T) and control
`(C) group are calculated for each day tumors are measured by subtracting the median
`tumor weight on the day of first treatment (staging day) from the median tumor weight
`on the specified observation day. These values are used to calculate a percent T/C as
`follows:
`
`"/0 T/C = (AT/AC) x lOOwhereAT > Oor
`= (A T/ T1)X100 whereAT < 0
`
`(l)
`
`and T1 is the median tumor weight at the start of treatment. The optimum (minimum)
`value obtained after the end of the first course of treatment is used to quantitate anti-
`tumor activity.
`Tumor growth delay: This is expressed as a percentage by which the treated group
`weight is delayed in attaining a specified number of doublings (from its staging day
`weight) compared to controls using the formula:
`
`[(T — C)/C] x 100
`
`(2)
`
`where T and C are the median times in days for treated and control groups, respec-
`tively, to attain the specified size (excluding tumor-free mice and DRDs). The growth
`delay is expressed as percentage of control to take into account the growth rate of the
`tumor since a growth delay based on T — C alone varies in significance with differences
`in tumor growth rates.
`Net log cell kill: An estimate of the number of log“, units of cells killed at the end of
`treatment is calculated as:
`
`{[(T — C) — duration of treatment] x 0.301 / median doubling time}
`
`(3)
`
`where the doubling time is the time required for tumors to increase in size from 200—400
`mg, 0.301 is the log“, of 2, and T and C are the median times in days for treated and
`control tumors to achieve the specified number of doublings. If the duration of treat-
`
`

`

`
`
`Chapter 6 / Human Tumor Xenograft Models 1 l l
`
`ment is 0, then it can be seen from the formulae for net log cell kill and percent growth
`delay that log cell kill is proportional to percent grth delay. A log cell kill of 0 in-
`dicates that the cell population at the end of treatment is the same as it was at the start
`of treatment. A log cell kill + 6 indicates a 99.9999070 reduction in the cell population.
`5. Tumor regression: The importance of tumor regression in animal models as an end
`point of clinical relevance has been propounded by several investigators (49—51).
`Regressions are defined as partial if the tumor weight decreases to 50% or less of the
`tumor weight at the start of treatment without dropping below 63 mg (5 X 5 mm tumor).
`Both complete regressions (CR5) and tumor—free survivors are defined by instances in
`which the tumor burden falls below measurable limits (< 63 mg) during the experimental
`period. The two parameters differ by the observation of either tumor regrowth (CR) or
`no regrth (tumor-free) prior to the final observation day. Although one can measure
`smaller tumors, the accuracy of measuring a sc tumor smaller than 4 x 4 or 5 x 5 mm
`(32 and 63 mg, respectively) is questionable. Also, once a relatively large tumor has re-
`gressed to 63 mg,
`the composition of the remaining mass may be only fibrous
`material/scar tissue. Measurement of tumor regrowth following cessation of treatment
`provides a more reliable indication of whether or not tumor cells survived treatment.
`
`Most xenografts that grow sc are amenable to use as an advanced-stage model,
`although for some tumors, the duration of the study may be limited by tumor necro-
`sis. As mentioned previously, this model enables the investigator to measure clinically
`relevant parameters of antitumor activity and provides a wealth of data on the effects
`of the test agent on tumor growth. Also, by staging day, the investigator is ensured
`that angiogenesis has occurred in the area of the tumor, and staging enables no-takes
`to be eliminated from the experiment. However, the model can be costly in terms of
`time and mice. For the slower-growing tumors, the passage time required before suf-
`ficient mice can be implanted with tumors may be at least 3—4 wk, and an additional
`2—3 wk may be required before the tumors can be staged. In order to stage tumors,
`more mice than needed for actual drug testing must be implanted, often 50%, and
`sometimes 100% more.
`
`3.3. Early Treatment and Early Stage sc Xenograft Models
`
`Early treatment and early stage sc models are similar to the advanced-stage model,
`but because treatment is initiated earlier in the development of the tumor, the models
`are not suitable for tumors that have less than a 90% take-rate or have a >10%
`
`spontaneous regression rate. We define the early treatment model as one in which
`treatment is initiated before tumors are measurable, i.e., < 63 mg, and the early stage
`model as one in which treatment is initiated when tumor size ranges from 63—200
`mg. The 634mg size is used as an indication that the original implant of approx 30 mg
`has demonstrated some growth. Parameters of toxicity are the same as those for the
`advanced-stage model; parameters of antitumor activity are similar. Percent T/C
`values are calculated directly from the median tumor weights on each observation day
`instead of as changes (A) in tumor weights, and growth delays are based on the time in
`days after implant for the tumors to reach a specified size, e.g., 500 or 1000 mg.
`Tumor-free mice are recorded, but may be designated no-takes or spontaneous re-
`gressions if the vehicle-treated control group contains more than 10% mice with
`similar growth characteristics. A no-take is a tumor that fails to become established
`and grow progressively. A spontaneous regression (graft failure) is a tumor that, after
`
`

`

`1 12
`
`Part II / In Vivo Methods
`
`a period of growth, decreases to 50% or less of its maximum size. Tumor regressions
`are not normally recorded, since they are not always a good indicator of antineo-
`plastic effects in the early stage model. For those experiments in which treatment is
`initiated when tumors are 100 mg or less, only a minimal reduction in tumor size may
`bring the tumor below the measurable limit, and for some small tumors early in their
`growth, reductions in tumor size may reflect erractic growth rather than a true reflec-
`tion of a cell killing effect. The big advantage of the early treatment model is the
`ability to use all implanted mice. The latter is the reason for requiring a good tumor
`take-rate, and in practice, the tumors most suitable for this model tend to be the
`faster-growing ones.
`
`