See related article Waldman et al., Cancer Res 1995;55:5187–90.

Visit the Cancer Research 75th Anniversary timeline.

By the early to mid-1990s it was becoming increasing clear that cancer was in large part a loss-of-function disease, caused by inactivating mutations in tumor suppressor genes. Several of the most important tumor suppressors had already been identified, including p53, Rb, APC, and others. However, once these new recessive cancer genes were discovered, it was often difficult to determine their function, primarily because of a lack of tools for loss-of-function functional genetics in human cells. The tools that were available at the time—antisense RNA, ribozymes, and dominant negative alleles—were frustratingly inefficient, inconsistent in their efficacy, or only applicable to proteins existing in dimers (or in the case of p53, tetramers). Although it was possible to create knockouts in mice, the same was not true for cultured human cells, thought in large part to be due to the inefficiency of homologous recombination in human cells.

My interest in this problem had been piqued as an undergraduate at Yale in the Department of Molecular Biophysics and Biochemistry, when I took a seminar course from John Sedivy (now at Brown University) in which he described his long standing interest in knocking out genes in cultured somatic (i.e., non-ES) cells. His particular focus at the time was in developing techniques for targeted deletion of the myc oncogene in cultured murine somatic cells. The “word on the street” was that knocking out genes in mammalian somatic cells was a bit of a holy grail. I found this potential challenge particularly appealing because I had initially trained in Escherichia coli genetics with Susan Gottesman at NIH, where I had routinely used P1 transduction to easily introduce inactivating mutations into the endogenous allele of genes. I found it shocking that similar approaches for loss-of-function genetics were not available in cultured mammalian cells. After graduating from Yale in 1991, I started up in the MD/PhD program at Johns Hopkins (Baltimore, MD), having prearranged to do my PhD with Bert Vogelstein and hoped that it might be possible to work on the development of somatic cell gene targeting approaches for cancer genetics.

Fortunately, upon my arrival in the laboratory, a scientific challenge emerged that was perfect for the application of potential human somatic cell gene targeting approaches—evaluating the importance of p21WAF1/CIP1 as a p53 effector in human cells. Wafik El-Deiry (now at Fox Chase Cancer Center, Philadelphia, PA), a postdoctoral fellow who worked at the bench next to mine, had just reported in Cell the discovery of WAF1 as a gene whose transcription was dramatically upregulated by p53, contained several p53-binding sites in its promoter, and whose ectopic expression in human cells resulted in potent growth suppression (1). In a remarkable coincidence, in the same issue of Cell, Steve Elledge at Baylor (Houston, TX; now at Harvard, Boston, MA) had published the discovery of p21 as a novel cdk inhibitor that bound and inactivated cdk2-containing complexes (2). When Steve and Bert read off the first few amino acids of their respective proteins to each other over the phone, it became clear that they were studying the same protein from two completely different and complementary angles, resulting in the fundamental discovery that the major p53-regulated gene encoded an inhibitor of cyclin-dependent kinases that had the potential to stop cell-cycle progression in its tracks. However, this hypothesis—that p21 was the key effector of the effects of p53 expression on cell-cycle arrest— required loss-of-function genetics to formally test. We knew that several groups had already geared up to create p21 KO mice. Bert and I decided that I should try to develop human somatic cell gene–targeting approaches for studying the role of p21WAF1/CIP1 in p53-mediated tumor suppression.

To test the role of p21 in p53 responses, we knew we needed a p53 wild-type colon cancer cell line that had an intact radiation-induced G1 arrest. We were limited to cancer cell lines, as hTERT-immortalization of primary human cells would not be reported for another five years. Our preference was a colon cancer cell line, as the focus of the laboratory was primarily colon cancer. And, most importantly, we needed a near-diploid cell line with two (and only two) copies of p21, as aneuploidy would make sequentially knocking out all the copies of a gene virtually impossible. After asking around, someone in the laboratory [I think it might have been either Jennifer Pietenpol (now at Vanderbilt, Nashville, TN) or Li Kuo Su (now at Cancer Cell)] mentioned that HCT116 was the fastest dividing, most robust cancer cell line they had ever grown in culture, and was reported to be diploid and have wild-type p53 (3). So, we went with HCT116 cells, which to this day remains the most robust cell line with which I have ever worked. We only later learned that HCT116 cells were especially well chosen for this project because they had a mutation in the MLH1 mismatch repair gene and were therefore mismatch repair deficient, which is thought to result in an enhanced efficiency of homologous recombination.

In an initial attempt to knockout p21, I created a targeting vector modeled after the standard mouse ES cell–targeting vectors that were in use at the time—two 2–5 kb homology arms (the longer the better!) flanking a neoR gene, with a thymidine kinase–negative selectable marker outside one of the homology arms. We spent some time pondering whether it was necessary to use so-called “isogenic” DNA for the homology arms (i.e., cloned from HCT116 cells themselves rather than from a human genomic DNA library derived from an anonymous donor). Isogenic DNA was (and still is) generally used to make mouse knockouts, as it was thought that occasional strain-specific polymorphisms could lower the efficiency of homologous recombination. After some thought, we decided against the use of isogenic DNA, as unlike the laboratory mouse, humans are not inbred, so the maternal and paternal alleles would have different polymorphisms. Therefore, we would need to create allele-specific knockout vectors, not just cell line–specific knockout vectors. Also, it would be impractical to use isogenic DNA if we hoped to eventually use the same targeting vector for multiple cell lines. So, I created the first targeting vector from genomic DNA contained in a commercially available BAC clone, transfected it into HCT116 cells, selected G418-resistant clones by limiting dilution in 96-well plates, and screened several hundred by Southern blot. No hits.

Because the frequency of homologous integration seemed quite low, we decided it might be more efficient and simpler to screen by PCR instead of by Southern blot. So, I modified the targeting vector by reducing the length of one of the homology arms (making it possible to PCR across it), and again transfected HCT116 cells and screened several hundred more clones. No hits. At this point I was getting quite anxious, because we knew that several laboratories were making p21 KO mice (see ref. 4 for the first of several p21 KO mice), and we felt that despite the fact that our approach was different, we would be viewed by journal editors as in competition with these mouse groups, so we did not want to get “scooped” (a concern that would be realized in about a year).

Then a remarkable and still classic article was published in Science by Shirasawa and colleagues reporting heterozygous knockout of either the mutant or wild-type allele of K-Ras in human colon cancer cell lines HCT116 and DLD1 (5). They showed that HCT116 and DLD1 could proliferate in culture in the absence of their oncogenic allele of K-Ras, but that deletion of mutant K-Ras led to loss of anchorage-independent growth and loss of tumorigenicity in immunodeficient mice. To create these cells, they created a so-called “promoterless” (aka “promoter-trap”) targeting vector, in which the selectable marker gene lacked its own promoter and was instead driven by the promoter of the gene they were trying to knockout. The obvious advantage of this approach was that the neoR gene would only be expressed (and create a G418R colony) if it landed next to a promoter. Therefore, most random integration events would fail to create a drug-resistant colony. While this would not increase the number of homologous integration events, it would reduce the so-called “denominator problem,” enhancing the all-important ratio of homologous recombination events to random integration events. A similar approach had recently been used by John Sedivy in his attempts to knockout myc in mouse somatic cells, and he had published a report attesting to the efficiency of this approach (6).

On the basis of these reports, I built a third-generation targeting vector that was identical to the first targeting vector except with a promoterless neoR gene. I transfected HCT116 cells, and, interestingly, the vector made orders of magnitude fewer G418-resistant colonies than the previous promoter-containing vectors, suggesting an enrichment was going on—promising. I then went on to screen approximately 100 individual colonies by Southern blot. I still remember sitting at the phosphorimager waiting for the image to appear and seeing two bands (heterozygous KO!) instead of one band (no KO) in about a third of the lanes. I ran to get Bert and we excitedly looked at the image of the KOs. It was a thrilling moment, and I still remember it fairly vividly 22 years later.

Now, how to target the remaining wild-type allele? This was before the advent of FLOXed neoR genes that could be removed easily by transient expression of cre. We decided to remake the targeting vector, replacing the neoR gene with several other resistance genes (we tried hygromycin and histidinol). Histidinol did not work at all (I am not sure it really ever worked for anybody and is no longer in use). Hygromycin worked pretty well, although not as well as G418—in the case of hygR, it was necessary to remove the drug after all the untransfected cells had died because the amount of hygR protein expressed by the resistance gene when driven by the p21 promoter was insufficient to confer long-term resistance. But, we screened approximately 100 hygR clones and achieved the same high efficiency of KO (∼30%)—as expected, half the KOs had integrated into the already targeted allele (replacing G418R with HygR) and half targeted the remaining wild-type allele, leading to a complete, homozygous knockout. At the time, we thought that this was the first homozygous knockout of any diploid gene in human cells, although I much later learned that a group at Oxford in the United Kingdom had reported the complete (i.e., biallelic) knockout of an immunologic-related gene in HT1080 osteosarcoma cells two years earlier (7).

Once the cells were made, we wanted to do some quick experiments and submit a paper. The key question at the time was whether p21 was required for the effects of p53 expression on enforcing a G1 cell-cycle arrest, which was thought to be the primary phenotypic effect of p53 expression. So, I cultured the isogenic sets of parental and p21 KO cells, walked them over to the radiation oncology center to irradiate, and then studied the cells via flow cytometry. Lo and behold, the G1 arrest was completely absent in the cells lacking p21. I performed some similar related experiments, and Bert and I planned some figures together over a celebratory lunch at the Hopkins outpatient center, and we submitted a letter to Nature. Rejected without review. Their rationale: we had been scooped by Phil Leder's laboratory (in collaboration with Steve Elledge), who had reported a p21 KO mouse in Cell several months earlier and had stated that p21 was partially (but not completely) responsible for the p53-induced G1 arrest (4). We then quickly resubmitted the paper to Nature Genetics, which also rejected it without review, with the same reasoning. We then submitted to Cancer Research, where it was quickly accepted and published as an Advances in Brief (now called Priority Reports), which to this day I think is a great option for the rapid publication of important findings in cancer biology (8).

After the initial finding that p21 was required for the p53-dependent G1 checkpoint was published in Cancer Research, we went on to use p21 KO cells to make several additional observations about the role of p21 in p53 responses and cell-cycle control. We found that when treated with radiation or DNA-damaging chemotherapeutics, the p21 KO cells were not just deficient in the G1 checkpoint, but underwent multiple rounds of S phases in the absence of mitosis or cytokinesis, causing the cells to eventually undergo apoptosis (9). Because of this propensity for endoreduplication, we found that p21 modulated sensitivity to radiation and other DNA-damaging drugs both in vitro and in vivo, which led to the use of these isogenic cells as the basis of high-throughput screens to identify novel anticancer agents (10). In a related observation, Nelly Polyak (now at Harvard) found that p21 KO cells were more susceptible to apoptosis induced by ectopic expression of p53, which provided early evidence for the now pervasive notion that p21 serves as a brake on p53-induced apoptotic pathways (11). John Sedivy used the human p21 KO vectors described in the Cancer Research article to create the first ever knockouts in primary human cells, demonstrating that p21 was required for Hayflick limit-induced senescence (12). Then, Fred Bunz (now at Hopkins) used similar gene-targeting approaches to create isogenic sets of p53 KO HCT116 cells and used them together with the p21 KO cells to demonstrate roles for p53 in multiple cell-cycle checkpoints and in the role of p53 in regulating sensitivity to anticancer agents (13, 14).

In addition to catalyzing work on the p53/p21 pathway, the p21 KO study published in Cancer Research inaugurated a host of studies using gene targeting to knockout a variety of other cancer-related genes in HCT116 cells to discern their biochemical functions and phenotypic effects. Subsequent genes targeted included SMAD4, B-Raf, K-Ras, B-Catenin, PTEN, PIK3CA, Securin, PUMA, DNMT1, DNMT3B, Akt1, Akt2, BLM, 14-3-3Sigma, and others (reviewed in refs. 15, 16). Bert Vogelstein's postdoc Chris Torrance decided to try to use isogenic sets of gene-targeted human cells for drug screening, and soon thereafter moved his home to the United Kingdom and started Horizon Discovery, a now publicly traded company devoted to the creation and licensing of gene-targeted human cells to pharmaceutical companies for target-based drug discovery (17).

Since the initial publication of this paper in Cancer Research about 20 years ago, there have been a number of subsequent technical advances that have enhanced the efficiency and ease of gene targeting in cultured human cells. As described above, the initial advance (reported in 1993 by Shirasawa and colleagues; ref. 5) was the use of so-called promoterless targeting vectors, which made it possible to enrich for KOs by substantially reducing the number of nonhomologous recombination events resulting in neoR colonies. The only downside to this promoterless approach was that it required expression of the gene being targeted. This, however, was largely a theoretical concern, as it is unlikely that one would be interested in targeting a gene that was not expressed.

The second major advance was made several years later, in 1998, when David Russell at the University of Washington discovered that targeting vectors built-in an adeno-associated virus (AAV) backbone and delivered via infection drove homologous recombination events at high frequency (18). This was likely due, at least in part, to the fact that AAV has a single-stranded DNA genome, which may be particularly efficient at strand invasion, a key early step in the homologous recombination pathway. Since its initial discovery, this AAV-based strategy has been used extensively over the past 18 years to knockout genes, to correct naturally occurring mutations in tumor suppressor genes, and to add sequence features such as epitope tags to endogenous alleles of genes in human cells (examples in refs. 19–21).

The third major advance was the substantial efforts by a number of groups to develop technologies for creating targeted double-strand breaks in the human genome, with the idea that these breaks would drive high efficiency recombination with a recombination substrate. These advances coincided with a change in the name of this general approach from “gene targeting” to “gene editing.” These approaches, based on zinc-finger nucleases (ZFN) and transcription activator-like effector nucleases (TALEN), displayed variable efficacy, but were quite promising and in active development until the discovery of the most recent technique for the creation of targeted breaks in the human genome—Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR). For the first time ever, CRISPR has made the creation of homozygous knockouts of gene in human cells almost trivial, both because of its remarkably high efficiency (both alleles are often targeted simultaneously in the same cell!) and because aberrant double-strand break repair leads to the very efficient creation of frameshift (truncating) mutations (reviewed in ref. 22). Currently, CRISPR is used primarily to introduce inactivating frameshift mutations, as the efficiencies of CRISPR-driven recombination with ectopically supplied recombination templates remain variable (as with ZFNs and TALENs). However, it seems likely that it is only a matter of time before CRISPR will also be easily and routinely used to create knockin mutations with high efficiency as well.

This article was one the first reports of a knockout in human cells and inaugurated the use of gene editing for the study of tumor suppressor pathways in human cells. Together with work by Shirasawa and colleagues and Hanson and colleagues (refs. 5 and 6), it demonstrated the importance of using a promoterless enrichment for enhancing the efficiency of gene targeting in human cells and led to several observations that proved important for the p53 field, such as the fact that p21 was required for p53-induced cell-cycle arrest and that p21 serves as a brake on p53-induced apoptosis. In the 21 years since the publication of this article in 1995, the field of human gene targeting has been renamed “gene editing” and has progressed substantially with the advent of AAV-based gene targeting, through the use of ZFNs and TALENs, and most recently with the use of CRISPR. It will be exciting to see where targeted genomic modification in human cells will take us next.

No potential conflicts of interest were disclosed.

1.
El-Deiry
WS
,
Tokino
T
,
Velculescu
VE
,
Levy
DB
,
Parsons
R
,
Trent
JM
, et al
WAF1, a potential mediator of p53 tumor suppression
.
Cell
1993
;
75
:
817
25
.
2.
Harper
JW
,
Adami
GR
,
Wei
N
,
Keyomarsi
K
,
Elledge
SJ
. 
The p21 Cdk-interacting protein Cip1 is a potent inhibitor of G1 cyclin-dependent kinases
.
Cell
1993
;
75
:
805
16
.
3.
Brattain
MG
,
Fine
WD
,
Khaled
FM
,
Thompson
J
,
Brattain
DE
. 
Heterogeneity of malignant cells from a human colonic carcinoma
.
Cancer Res
1981
;
41
:
1751
6
.
4.
Deng
C
,
Zhang
P
,
Harper
JW
,
Elledge
SJ
,
Leder
P
. 
Mice lacking p21CIP1/WAF1 undergo normal development, but are defective in G1 checkpoint control
.
Cell
1995
;
82
:
675
84
.
5.
Shirasawa
S
,
Furuse
M
,
Yokoyama
N
,
Sasazuki
T
. 
Altered growth of human colon cancer cell lines disrupted at activated Ki-ras
.
Science
1993
;
260
:
85
8
.
6.
Hanson
KD
,
Sedivy
JM
. 
Analysis of biological selections for high-efficiency gene targeting
.
Mol Cell Biol
1995
;
15
:
45
51
.
7.
Porter
AC
,
Itzhaki
JE
. 
Gene targeting in human somatic cells. Complete inactivation of an interferon-inducible gene
.
Eur J Biochem
1993
;
218
:
273
81
.
8.
Waldman
T
,
Kinzler
KW
,
Vogelstein
B
. 
p21 is necessary for the p53-mediated G1 arrest in human cancer cells
.
Cancer Res
1995
;
55
:
5187
90
.
9.
Waldman
T
,
Lengauer
C
,
Kinzler
KW
,
Vogelstein
B
. 
Uncoupling of S phase and mitosis induced by anticancer agents in cells lacking p21
.
Nature
1996
;
381
:
713
6
.
10.
Waldman
T
,
Zhang
Y
,
Dillehay
L
,
Yu
J
,
Kinzler
K
,
Vogelstein
B
, et al
Cell-cycle arrest versus cell death in cancer therapy
.
Nat Med
1997
;
3
:
1034
6
.
11.
Polyak
K
,
Waldman
T
,
He
TC
,
Kinzler
KW
,
Vogelstein
B
. 
Genetic determinants of p53-induced apoptosis and growth arrest
.
Genes Dev
1996
;
10
:
1945
52
.
12.
Brown
JP
,
Wei
W
,
Sedivy
JM
. 
Bypass of senescence after disruption of p21CIP1/WAF1 gene in normal diploid human fibroblasts
.
Science
1997
;
277
:
831
4
.
13.
Bunz
F
,
Dutriaux
A
,
Lengauer
C
,
Waldman
T
,
Zhou
S
,
Brown
JP
, et al
Requirement for p53 and p21 to sustain G2 arrest after DNA damage
.
Science
1998
;
282
:
1497
501
.
14.
Bunz
F
,
Hwang
PM
,
Torrance
C
,
Waldman
T
,
Zhang
Y
,
Dillehay
L
, et al
Disruption of p53 in human cancer cells alters the responses to therapeutic agents
.
J Clin Invest
1999
;
104
:
263
9
.
15.
Waldman
T
,
Lee
C
,
Nishanian
TG
,
Kim
JS
. 
Human somatic cell gene targeting
.
Curr Protoc Mol Biol
2003
;
Unit
9.15.
16.
Rago
C
,
Vogelstein
B
,
Bunz
F
. 
Genetic knockouts and knockins in human somatic cells
.
Nat Protoc
2007
;
2
:
2734
46
.
17.
Torrance
CJ
,
Agrawal
V
,
Vogelstein
B
,
Kinzler
KW
. 
Use of isogenic human cancer cells for high-throughput screening and drug discovery
.
Nat Biotechnol
2001
;
19
:
940
5
.
18.
Russell
DW
,
Hirata
RK
. 
Human gene targeting by viral vectors
.
Nat Genet
1998
;
18
:
325
30
.
19.
Sur
S
,
Pagliarini
R
,
Bunz
F
,
Rago
C
,
Diaz
LA
 Jr
,
Kinzler
KW
, et al
A panel of isogenic human cancer cells suggests a therapeutic approach for cancers with inactivated p53
.
Proc Natl Acad Sci U S A
2009
;
106
:
3964
9
.
20.
Solomon
DA
,
Kim
T
,
Diaz-Martinez
LA
,
Fair
J
,
Elkahloun
AG
,
Harris
BT
, et al
Mutational inactivation of STAG2 causes aneuploidy in human cancer
.
Science
2011
;
333
:
1039
43
.
21.
Kim
JS
,
Bonifant
C
,
Bunz
F
,
Lane
WS
,
Waldman
T
. 
Epitope tagging of endogenous genes in diverse human cell lines
.
Nucleic Acids Res
2008
;
36
:
e127
.
22.
Cong
L
,
Zhang
F
. 
Genome engineering using CRISPR-Cas9 system
.
Methods Mol Biol
2015
;
1239
:
197
217
.