PAPERS
`
`Non-invasive detection ofmalignancy by identification ofunusual
`CD44 gene activity in exfoliated cancer cells
`
`Yasuhiro Matsumura, Damian Hanbury, Joseph Smith, David Tarin
`
`Abstract
`Objective-To investigate non-invasive detection
`of cancer by testing for unusual CD44 gene activity
`in a clinical sample as an indicator of exfoliated
`tumour cells.
`Design-Case-control study.
`patients
`unselected,
`Subjects-44
`consecutive
`with bladder cancer and 46 people with no evidence
`ofneoplasia.
`Main outcome measure-Presence or absence of
`large CD44 gene products containing exon 6 deriv-
`atives in urine samples.
`Results-Novel abnormalities in the pattern of
`expression of this gene, seen specifically in tumour
`tissue, led to cloning of a newly recognised coding
`region in it (exon 6). This was tested as a probe for
`detection of exfoliated malignant cells in naturally
`voided urine. CD44 gene products extracted from
`the urine and amplified with polymerase chain
`reaction contained predicted electrophoretic band
`of 735 base pairs in 40 ofthe 44 patients with bladder
`cancer (sensitivity 91%). Products from 38 of the 46
`people with no evidence of neoplasia showed no
`such band (specificity 83%).
`Conclusions-Unusual activity of the CD44 locus
`in neoplasia and malignancy is confirmed, and
`techniques for the analysis of such activity can
`enable non-invasive investigation of patients for
`primary or recurrent bladder cancer or for other
`tumours that shed neoplastic cells into body fluids.
`
`Introduction
`Early detection of cancer is desirable since many
`cancers can be cured by surgical resection if diagnosed
`at an early stage,' 2 and this has led to the introduction
`of cancer screening programmes in many countries.
`Therefore, there is considerable incentive to find ways
`of detecting cancers early in their life history when they
`might still be localised and easily resected or destroyed.
`Bladder cancer is a common form of malignancy,
`and treatment would be greatly helped by early
`detection of primary and recurrent tumours. Current
`methods of screening by looking for microscopic
`non-specific, and urine
`cytology,
`haematuria are
`though specific, does not have the sensitivity to be used
`as a routine test for exclusion of bladder neoplasia.
`Radioimmunoassay for tumour markers such as
`carcinoembryonic antigen is not suitable for detecting
`cancer in its early stages though it can sometimes be
`used to monitor patients for increasing tumour burden
`or tumour recurrence.4 Diagnoses by urography and
`cystoscopy are effective but time consuming, expen-
`sive, invasive, and uncomfortable. They are the main
`methods for investigating symptomatic disease but are
`not suitable for routine screening for asymptomatic
`primary or recurrent tumours. Also, the workload of
`follow up cystoscopies after diagnosis and treatment to
`
`enable early detection of any recurrence is a consider-
`able burden to urology departments.
`There is therefore a need for a sensitive, reliable, and
`non-invasive method of detecting bladder cancer. We
`recently reported abnormal pattems of activity of the
`CD44 gene in tumours,5 and we have now cloned and
`sequenced a piece of complementary DNA corre-
`sponding to a new exon of this gene from a fresh
`tumour. In the present study we used this sequence
`data to confirm our previous findings of overexpression
`of the new exon in tumours and investigated the
`method's potential for detecting bladder cancer.With
`improved assay methods we could detect abnormal
`activity of the gene, and hence the likely presence of
`small numbers of exfoliated cancer cells, in naturally
`voided urine from patients with bladder cancer if the
`sample contained viable cells.
`
`Subjects and methods
`COMPLEMENTARY DNA SEQUENCING
`In our earlier study we noted that when comple-
`mentary DNA from tumour samples was amplified
`with primers P1 and P4 one of the polymerase chain
`reaction products consistently obtained was 1650 base
`pairs in size,' larger than the maximum predicted size
`of 1500 base pairs that would result from inclusion of
`transcripts from all then known exons at the insertion
`site.69 We concluded that there was an extra, unidenti-
`fied exon in the variably expressed region of the gene
`that is overexpressed in tumour tissue. We reasoned
`that this segment of complementary DNA might
`provide a sensitive and specific probe for tumour
`diagnosis and therefore isolated and characterised it.
`By using suitably chosen combinations of primers
`(data not shown), we found this exon to lie 5' to all the
`other variably expressed ones and thus to be between
`them and the 5' section (that is, upstream of the
`insertion site) of the standard portion of the gene (see
`fig 1).
`We then designed primers P3 and E4 to amplify
`
`El
`
`E3
`
`-4-
`
`F71 F-1FglF-o
`E2
`
`~~~~~~-4-
`
`ii
`E4 E5
`
`F12F
`
`Pi
`
`P2
`
`P3
`Standard
`
`SI
`
`Form
`
`3'
`
`<- 324 -
`base pairs
`
`158 .
`base pairs
`Insertion site
`(position 782-783 base pairs)
`FIG 1-Map of CD44 gene products together with positions ofprobes
`and primers. Newly identified exon labelled exon 6 in accordance with
`nomenclature ofScreaton et al'2
`
`Nuffield Department of
`Pathology, University of
`Oxford, John Radcliffe
`Hospital, Oxford OX3 9DU
`Yasuhiro Matsumura, senior
`clinical research scientist
`David Tarin, Nuffield reader
`in pathology
`
`Department ofUrology,
`Churchill Hospital, Oxford
`OX3 8LJ
`Damian Hanbury, senior
`registrar
`Joseph Smith, consultant
`surgeon
`
`Correspondence to: Dr
`Tarin.
`
`BMJ 1994;308:619-24
`
`BMJ VOLUME 308
`
`5 MARCH 1994
`
`619
`
`Geneoscopy Exhibit 1063, Page 1
`
`

`

`I2
`
`3 4
`
`5 6 7
`
`-o
`'I
`-O
`I',
`
`z17
`
`1.35-
`1.08 I
`0.87 "
`0.60.'
`
`Standard
`
`744 base pairs
`
`615 base pairs
`
`482 base pairs
`
`between the 5' end of the standard portion and exon
`1 1, using four tissue samples of breast cancer and three
`samples of normal breast tissue. As expected, we
`observed clear bands of 615 base pairs and 744 base
`pairs in the gels of the polymerase chain reaction
`products from cancer samples but not from normal
`samples (fig 2, upper panel). These were truncated
`versions of the bands of 1500 base pairs and 1650 base
`pairs seen in tumours and not in normal tissues in our
`earlier study with primers P1 and P4.' The larger band
`was therefore cut out from the gel, and the constituent
`Cancer
`Normal
`complementary DNA eluted. This was cloned into the
`% agarose)
`FIG 2-Ethidium bromide stained gel (12
`after 30 TA vector (Invitrogen), and transformed bacteria were
`amplification cycles of complementary DNA
`from four samples of
`..
`hybridised with the E4 probe. Positive clones were
`if normal breast tissue
`breast cancer (lanes 1-4) and three samples c
`analysed by the dideoxy method'" to obtain the base
`ter panel) only cancer
`(lanes 5-7). With primers P3 and E4 (uppi
`id 744 base pairs; with
`sequence of the new exon. Amplification with primers
`samples showed clear bands of 615 base pairs an
`iwed bands of482 base
`P1 and P4 produced a band of 482 base pairs, the
`primers P1 and P4 (lower panel) all samples shc
`pairs, the standardform ofCD44 transcript
`standard form of CD44 transcript, in all the samples
`(fig 2, lower panel), confirming that they contained
`satisfactory messenger RNA. Further studies with
`primers E3 and P4 indicated that no further exons lay
`downstream of E3 in the variably spliced region of the
`gene.
`
`Standard
`
`New exon (exon
`
`6)
`
`GCTACCACTTTGATGAGCAC
`CTAGTGCTACA
`I
`r S A T
`S
`T
`L M
`
`GCAACTGAGACAGCAACCA,
`AGAGGCAAGAA
`A T
`E
`T A T
`R Q E
`K
`
`ACCTGGGATTGGTTTTCATC
`;IGTTGTTT|CTA
`T W D W F
`S
`'V
`N L
`F
`L
`
`CCATCAGAGTCAAAGAATCA
`kTCITICACACA
`S
`P
`S
`E
`K N I
`H
`H
`L
`T
`
`ANALYSIS OF SOLID TISSUE SAMPLES
`We examined solid tissue samples from normal
`tissue and from tumours from patients with cancers of
`the breast, colon, and bladder. The procedures were as
`described earlier,5 and reaction products were probed
`with oligonucleotides E2, E4, and P2 to analyse
`expression of the new exon, expression of exon 11, and
`to check the quality of the complementary DNA,
`respectively.
`
`FIG 3-Sequence ofnew exon
`identified in CD44 gene.
`Underlining indicates potential
`0-glycation sites
`
`ACAACACAAATGGCTGGTA(
`T Q M AH
`T
`
`A
`6 7 8 9 0lil 121314 15
`
`1.35I-
`I .0
`087
`0.60-
`
`1.35 -
`1.08-
`0.87-
`0.60-
`
`Exon 7
`
`c
`2 3 ,4
`
`5
`
`URINE ANALYSIS
`We refined the experimental protocol with an initial
`series of urine samples from eight patients with known
`bladder cancer and eight with no evidence of the
`disease and then applied it to a larger series of subjects.
`Subjects-We examined urine samples from 90
`6 7 8
`subjects: 44 patients with bladder carcinoma proved by
`biopsy, 12 patients with inflammation of the bladder
`--
`....(cystitis), and 34 asymptomatic volunteers. Of the 46
`~people without neoplasia, 30 were males aged 18-77
`-s _
`1
`Fand 16 were females aged 7-76.
`; --- _
`~Initial analysis-About50m ofntrlyvie
`-F |
`:at1 1urine was obtained from each subject and transported
`
`-
`
`-
`
`~~of each sample was centrifuged, and the viability and
`quantity of cells in the sample were assessed by fluores-
`
`_diacetate and ethidium bromide." The remainder of
`lthe sample was centrifuged to pellet suspended cells for
`~extraction of messenger RA.
`... ...Molecular analysis-he actiity of the CD44 gene
`
`M _
`
`...
`
`,f ~~RNA to produce a complementary DNA copy of any
`X recently
`functioning
`genes
`followed
`by
`specific
`;4 amplification of the complementary DNA molecules
`.:::- i !pthat correspond to genes ofinterest with the polymerase
`chain reaction. This method allows study of the
`;L |
`| | 11 |amounts and sizes of the products assembled by a given
`-
`* |gene and therefore provides a snapshot of its activity at
`._*l |
`the time of sampling. Details of the method are given in
`
`L
`
`..
`
`0'O0
`
`G0 owI, 4S
`
`000
`day), E4 (exposed for
`FIG 4-Autoradiographs offilters probed with oligonucleotides E2 (exposed for one o
`fter amplification with
`eight hours), and P2 (exposed for two hours) in rows I, II, and III respectively aj
`pheral blood leucocytes
`primers P1 and P4. Panel A shows results for normal breast tissue (lanes 1-4), peril
`ease (lanes 8-10), and
`from healthy volunteers (lanes 5-7), sternal bone marrowfrom patients with heart dies
`1-4, (normal tissue in
`breast cancer tissue (lanes 11-15); panel B shows non-neoplastic colon tissue (lanes
`ative colitis in lane 4))
`lanes I and 3 and inflamed colonic mucosa, from Crohn's disease in lane 2 and ulcerx
`tissue (lanes 1-4) and
`and tissue from primary colonic cancer (lanes 5-8); panel C shows normal bladder
`primary bladder catcer tissue (lanes 5-8)
`
`620
`
`Results
`DETAILS OF NEWLY IDENTIFIED CD44 EXON
`The overexpressed exon we cloned and sequenced
`from fresh human tumour tissue was 129 base pairs
`long and had 34% of seine and threonine residues and
`two potential 0-glycation sites (fig 3). Its sequence is
`identical to that recently isolated from a genomic
`library of the HT-29 colon cancer cell line by Screaton
`
`BMJ voLuME 308
`
`5 MARCH 1994
`
`Geneoscopy Exhibit 1063, Page 2
`
`

`

`et al."2 In accordance with their system
`of numbenng or
`CD44 exons, based on analysis of ger
`aomic DNA, we
`shall refer to it as exon 6.
`
`-10
`
`735 bae pairs
`
`44-base palrs
`
`17-35 base pairs
`
`484 base pairs
`
`th oligonucleotide E4
`FIG 5-Autoradiographs offilters amplified with primers El and ES and probed wit
`half). Tracks 15-28
`(upper half of each panel) and amplified with P1 and P4 and probed with P2 (lower
`nd 29-37 show results
`and 38-46 show results with urine samplesfrom bladder cancer patients. Tracks 1-14 at
`Ydy of plasmid DkNA
`with urine samples from non-neoplastic controls. Tracks 47-53 are serial dilution stu
`containing exons 6-11 (see appendix for details). Upper half of each panel shows only
`largest amplification
`pair band to confirm
`products, including 735 base pair band: lower half ofpanel shows standard 484 base
`presence ofsatisfactory messenger RNA
`
`TABLE I-Details of44 patients with bladder cancer
`
`Sex and
`Patient No age (years)
`
`No oflive cells per
`Histological stage and 10-4ml concentrated
`grade of tumours*
`urinet
`
`CD44 transcript
`levelt
`
`Lane in fig 5 if shown*
`
`smples
`Molecular analysis ofurine sa
`
`15
`16
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27,28
`39
`40
`41
`42
`43
`44
`45
`46
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`29
`30
`31
`32
`33
`34
`35
`36
`37
`38
`39
`40
`41
`42
`43
`44
`
`M75
`M62
`M69
`M76
`M53
`M57
`M76
`F69
`M44
`F56
`F62
`M69
`M81
`M51
`F72
`M80
`F50
`M62
`M64
`M47
`M69
`M74
`M65
`M62
`M64
`M78
`M75
`M76
`M56
`F71
`M87
`M64
`F87
`M71
`M65
`F81
`F81
`M56
`M59
`F64
`M66
`M84
`M72
`M78
`
`pT2 G3
`pTI G1 LO
`pT3 G3
`pT3G3L2
`pT2 G1
`pTl G1 LO
`pT3G2L1
`pTxG2LO
`pTaG1 LO
`pTI Gl L0
`pTl GI LO
`pT1 G2LO
`Necrotic debris
`pT3 G3 L2
`pT2 G2 LO
`pTl G3LO
`pTxG2LO
`pTaG1 LO
`pTl G2LO
`pTl G2 LO
`pTlG2LO
`pTI G3 LO
`pTl G1 LO
`pTl G1 LO
`pTis GI LO
`pTI G2L0
`pTl GlLO
`pTis GI LO
`pT2 G2 LO
`pTlG2LO
`pTl G2 LO
`pTl G3 LO
`pTl G1 LO
`pTl GI LO
`pTl G2LO
`pTI G1IL0
`pTl G2LO
`pTI G2LO
`pTl G2LO
`pTI G2LO
`pTl G2 LO
`pTl G3 L1
`pTa G1 LO
`pTl G1LO
`
`4-5
`1
`Many
`Many
`1
`4-5
`4-5
`2-3
`1
`4-5
`Many
`<1
`Many
`< 1
`4-5
`1
`2-3
`<1
`<1
`2-3
`<1
`< 1
`
`Many
`
`1
`5
`< 1
`
`Many
`1-2
`
`Many
`5-6
`
`2-3
`4-5
`Many
`2-3
`5-6
`Many
`2-3
`<1
`
`10-28
`10-20_10-'9
`10"-10-20
`10-22 10-
`10-19
`10-2_10-20
`10-2_10-20
`l0-22 10-21
`10-22 10-21
`10-2_10-"
`l0-22_10"-
`10-22 10-2
`10-22 10-21
`10-19
`10-20_10-19
`10-22 10-21
`10-21
`10-22_ 10-2
`10-22 l0-2
`> 10-7
`10-22 10-
`10-2'
`109
`10-20_10-19
`10-19
`10-22 10-21
`10-'9
`10-21_10-20
`10-20
`10-21_1020
`< 10-22
`< 1b22
`<10-22
`10-22 10-2
`10-21_10"20
`10-2110-20
`10-22102
`10-22 l0-
`1o-22
`0-
`10-21
`1022 l0-2
`10-21_10-20
`<10-22
`10-2210-2
`
`ancer (UICC): pTis-
`*Tumour stage and grade evaluated according to classification ofInternational Union against C
`ot extending beyond
`preinvasive carcinoma (carcinoma in situ), pTa=papillary non-invasive cancer, pTl -tumour ns
`edegree of
`the lamina propria, pT2-tumour invading superficial smooth muscle, pT3 -tumour invading d
`leep muscle, pTx-
`extent oftumour invasion cannot be assessed; G1 -high degree ofdifferentiation, G2-medium
`uperficial
`differentiation, G3-low degree ofdifferentiation; LO-no lymphatic invasion, LI -invasion of S
`lymphatics, L2 -invasion ofdeep lymphatics.
`tUrine samples concentrated 10-fold by centrifugation.
`tMole equivalents of DNA fragment used as a standard for quantification (see fig 5).
`
`ANALYSIS OF SOLID TISSUES FOR EXPRESSION OF TUMOUR
`RELATED EXONS
`When the polymerase chain reaction products were
`hybridised with radiolabelled E2 or E4 (see fig 1) all
`samples from carcinomas overexpressed several splice
`variants (fig 4, rows I and II), but the pattern of bands
`seen with each probe was different. Hence, the
`oligonucleotide probe for products of exon 6 was
`distinguishing
`neoplastic
`effective
`in
`and non-
`neoplastic samples but not significantly more sensitive
`than the E4 probe used previously, at least on samples
`from solid tissues. Subsequently, the same filters were
`stripped and hybridised with probe P2 to show that all
`samples, including normal tissues, produced the
`standard portion of CD44 (fig 4, row III). This
`confirmed that the differences observed between
`control and tumour samples when probed with E2 or
`E4 were not due to unequal loading of reaction
`products on these gels.
`The cumulative results of our studies on solid tissue
`samples are that cancers of the breast, colon, bladder,
`stomach, prostate, and thyroid in 47 different patients
`had significant overexpression of various exons of
`CD44 but corresponding normal tissues from 39
`people did not (partially shown in fig 4 and in our
`previous report5). Figure 4 shows that a few bands, all
`less than 1100 base pairs in size, were present in the
`lanes from the normal samples though the intensity of
`these bands was very weak. Bands larger than 1500
`base pairs in size were never seen in amplified gene
`products from normal tissue samples. We consider,
`therefore, that the higher the molecular size the more
`specific it is for the diagnosis of cancer.
`URINE ANALYSIS FOR MALIGNANCY WITH CD44 PROBES
`Table I shows details of the 44 patients with bladder
`cancer. If tumour cells in urine were to be expressing
`all the exons from 6 to 11 we predicted that our
`protocol should produce a 735 base pair band with
`primers El and E5 (see appendix). Figure 5 shows that
`there were clear differences between most tumour and
`non-tumour samples relative to the presence or absence
`of the 735 base pair product. The discrimination was
`not absolute, however, and we occasionally observed a
`735 base pair band in the amplification products from
`subjects with no clinical evidence of neoplasia (such as
`lanes 1, 10, 13, 32, and 35 in fig 5) and sometimes saw
`none in a patient with bladder cancer. The presence of
`the 484 base pair band showed that all the samples
`contained satisfactory messenger RNA.
`The serial dilution study shown in lanes 47-53 was
`included in this experiment to monitor the reliability of
`amplification of the target segment by the polymerase
`chain reaction under these conditions. It was possible
`to detect a signal down to 10-22 mole of the plasmid
`which was used as a standard. We therefore adopted
`this as the threshold for designating a sample as
`positive or negative for the presence of the 735 base
`pair template in the clinical samples. The overall
`results ofthe study showed that samples from 40 (9 1%)
`of the 44 patients with cancer were positive and that 38
`(83%) of the 46 from people without neoplasia were
`negative. The method therefore has a sensitivity of
`91%, specificity of 83%, a positive predictive value of
`83% (40/48), and a negative predictive value of 90%
`(38/42) for bladder cancer. Among the subjects re-
`ceiving the test the prevalence of the disease was 44%
`(40/90).
`The signal in individual samples can be roughly
`quantified by comparison with the serially diluted
`standard. For evaluation of the importance of a signal
`the results of molecular analysis should be related to
`the viable cell concentration in the urinary sediment
`after centrifugation. Table II summarises the results of
`such an analysis. Some samples from cancer patients
`
`BMJ VOLUME 308
`
`5 MARCH 1994
`
`621
`
`Geneoscopy Exhibit 1063, Page 3
`
`

`

`TABLE II-Results of molecular analysis of urine samples from 35
`patients with bladder cancer and 26 controls* by cell number in urine.
`Values are numbers ofsubjects unless stated otherwise
`
`Result of molecular analysist
`
`No of live cells per
`10- ml concentrated
`urinet
`
`Cancer patients
`
`Controls
`
`Positive
`
`Negative
`
`Positive
`
`Negative
`
`Many
`2-10
`1-2
`<1
`
`6
`14
`3
`8
`
`2
`1
`1
`0
`
`1
`3
`1
`1
`
`2
`6
`7
`5
`
`*Urne samples from remaining nine cancer patients and 20 controls which
`were received in the early stages ofthe study, were not assessed.
`tUrine samples concentrated 1 0-fold by centrifugation.
`tResult regarded as positive if - 10-1 mole of standard DNA segment
`present and regarded as negative if < 10-"1mole present.
`
`showed a strongly staining band of 735 base pairs after
`amplification despite containing less than one cell per
`1 0-4ml of 10-fold concentrated urine (for example,
`lanes 16, 19, and 38 in fig 5). Conversely, samples from
`subjects with no clinical evidence of neoplasia did not
`produce a strong band of 735 base pairs even when they
`contained more than 10 cells per 10Q4ml. Hence the
`presence and intensity of the band is not simply related
`to the total number of cells present.
`
`Discussion
`The products of CD44 gene activity appear to
`exercise many diverse functions in the cell and at the
`cell surface,'3 and disturbances in the gene activity seen
`in tumour cells probably stem from derangement of the
`mechanisms controlling alternative splicing of this
`gene. From the present work and from our earlier
`studies5 we believe that tumour cells produce orders of
`magnitude more of the unusual transcripts from the
`CD44 locus compared with non-neoplastic cells. Even
`a sprinkling of tumour cells in a sample can therefore
`be detected under optimal conditions.
`
`EFFECT OF RIBONUCLEASES
`Such a highly sensitive method could be of great
`clinical benefit, but ribonucleases are often present in
`high concentration in biological samples that contain
`many bacteria or dead or dying eukaryotic cells.
`Thus messenger RNA from a sample can be rapidly
`degraded during extraction if the quantity of inhibitor
`added is insufficient, giving a spurious negative or
`a weak positive result. In our study two of the
`four urine samples from patients with cancer that
`were negative for the 735 base pair band contained
`numerous cells because of coexisting haematuria or
`cystitis. The RNA extraction kit which we used cannot
`cope with more than 5 x 106 cells simultaneously
`because of the amount of ribonuclease released during
`the extraction procedure. For example, lanes 27 and
`28 in figure 5 are different aliquots from a single
`sample from a patient with severe haematuria, but one
`aliquot reproducibly gave a positive signal and the
`other did not. The only difference was the starting
`urine volume used for cell pelleting before extraction:
`20 ml for lane 27 and 1 ml for lane 28. The positive result
`was obtained with the smaller starting volume, which
`fits with the suggestion that the RNA was degraded
`by ribonucleases released by blood cells in the sample.
`It is also possible that inhibitors of the polymerase
`chain reaction, known to be present in blood, could
`be present in greater concentration in gross haema-
`turia.
`The most important requirement for this method is
`to get high quality intact RNA. In our experience RNA
`is degraded gradually even at - 70°C. RNA should
`therefore be extracted and transcribed to comple-
`mentary DNA when urine samples arrive and without
`
`freezing them, which releases ribonucleases from the
`cells. Concurrent cystitis should be treated before
`urine samples are taken for examination. People with
`persistent frank haematuria should be referred directly
`to a urologist for cystoscopy without having such a
`test.
`
`RELIABILITY OF METHOD
`Most of the control urine samples (83%) were
`completely negative for the 735 base pair band. It is
`unlikely that these results were caused by technical
`failure such as RNA degradation because the standard
`form CD44 transcript constitutively expressed in all
`cell types (482 base pairs) was obtained with each
`sample (fig 5, lower panel). Conversely, eight of the 46
`control samples were positive for the 735 base pair
`band. We do not have a definitive explanation for these
`cases, but, as they gave similar results on repeat testing
`of subsequent samples, cross contamination with RNA
`or cells from tumour samples can reasonably be
`excluded. Possibly, in normal samples there are slight
`differences related to cell type in the expression of
`CD44 variants which are usually below the threshold of
`detection but which can be seen in some subjects. In
`our study the threshold for detection was set at
`maximum sensitivity on the basis of the results of the
`pilot study. Results from further studies to refine our
`method indicate that adjustment of the positions of the
`primers and probes used for amplification and check-
`ing ofgels for an abnormal pattern of CD44 activity can
`give even greater specificity of diagnosis than that
`already achieved (unpublished observations). It should
`also be remembered that in an asymptomatic population
`there may be some people who have early, reversible,
`preneoplastic changes such as dysplasia or cellular
`atypia that could be detected by an highly sensitive
`method. Of the 40 patients with bladder cancer in our
`study who had positive results, two had preinvasive
`carcinoma in situ and 28 had histologically early
`stage invasive cancer (pTisG1 and pTlG1 respec-
`tively in table I). These results indicate that disorder
`in CD44 gene expression begins early in the process of
`neoplasia.
`The CD44 method achieved reasonable specificity as
`well as good positive and negative predictive value in
`this sample of people. We have reason to believe that
`sampling,
`improvements in urine
`transport, and
`laboratory technique will further improve its sensitivity
`and specificity of detection (unpublished data). Even
`so, the method does not directly visualise the presence
`ofmalignant cells or tissue, and a policy ofconfirmation
`of a presumptive diagnosis with the aid ofmicroscopical
`techniques would be advisable.
`
`RESULTS FROM OTHER STUDIES
`Studies on CD44 gene activity by ourselves and
`others have contributed to understanding the funda-
`mental mechanisms of neoplasia and the clinical
`detection of primary and recurrent malignancy.5'1'8
`Other genes have also been studied to investigate the
`feasibility ofdiagnosing tumours by means ofmolecular
`genetics. For example, Sidransky et al reported identi-
`fication ofp53 gene mutations in exfoliated cancer cells
`in urine in three patients.'9 In parallel studies on solid
`tissue samples they found that mutations in these genes
`were present in 61% of bladder tumours. Haliassos et al
`sought evidence of H-ras gene mutations in the urine of
`21 patients with bladder cancer and reported mutations
`at codon 12 in 10 (47%) ofthem.20 In the same patients
`the prevalence of H-ras mutations in their bladder
`tissues was 66%. Such findings, together with our
`results,
`evidence
`that molecular genetic
`provide
`methods have the potential to provide powerful new
`tools for the early detection and prognostic evaluation
`ofhuman tumours.
`
`BMJ VOLUME 308
`
`5 MARCH 1994
`
`622
`
`Geneoscopy Exhibit 1063, Page 4
`
`

`

`COMPARISON WITH OTHER METHODS
`DNA based techniques are ofundisputed importance
`for the detection of oncogene related mutations in the
`analysis of cancer aetiology and of familial predis-
`position to certain types of neoplasia. Our present
`results, however, indicate that RNA based diagnostic
`techniques, which give direct evidence of abnormal
`gene activity at the time of sampling, may have
`advantages in the diagnosis of existing neoplastic
`disease. Although the incidence of mutations in known
`oncogenes is quite high in certain types of cancer, in
`most malignancies it is insufficient to be of practical
`benefit in routine clinical practice. RNA based methods
`therefore deserve attention in further research. The
`method described here, analysing expression of CD44,
`has the advantages that it can be performed in a single
`day and that it reliably ascertained the presence of
`abnormal splice variants in exfoliated cells in urine
`samples from 91% of patients with bladder cancer.
`From data already published5 1418 we know that there is
`a strong association between such abnormal CD44
`expression and the presence of malignant cells in tissue
`samples. The detection of excessive and abnormal
`CD44 activity in a clinical sample should therefore
`raise strong suspicion ofthe presence ofneoplasia.
`The sensitivity of detection obtained with this new
`method in this study of an unselected series of bladder
`tumours ofall types and stages, including early papillary
`tumours, was greater than the values recently published
`for studies on similar unselected groups using urine
`microhaematuria.2'
`cytology3' 11
`For accurate
`or
`comparison, however, it would be necessary to study
`the results of all these methods on the same set of
`samples.
`
`CONCLUSION
`We suggest that this method could become a useful
`partner to cytology and histopathology in the clinical
`investigation of bladder malignancy. It is also likely to
`have wider applications because we have recently
`successfully used it to detect abnormal CD44 expression
`in exfoliated cells in stools from patients with colorectal
`cancer but not in ones from normal subjects (Matsu-
`mura and Tarin, unpublished observations). The
`technique is not difficult for trained perponnel, but it
`needs to be performed with care because messenger
`RNA is vulnerable to degradation and the reaction can
`amplify trace contaminants to detectable levels. It is
`non-invasive, and, as samples can be batch tested, it
`would be considerably cheaper than cystoscopy for
`initial investigation for primary or recurrent bladder
`neoplasia. We therefore present it as a potentially
`convenient and practical method for cancer screening
`and for investigation of patients with suspected neo-
`plastic disease. In order to make a confident assessment
`of the predictive value of the test in the general
`population, we now need more extensive clinical
`evaluation by means of double blind trials.
`We thank the following for help with this work: L Bao, D
`Cranston, D L Darling, J Davies, M Evans, G J Fellows,
`L Kaklamanis, D Lo, S Matsumura, J O'D McGee, T
`Shirakawa, E M Southern, L Summerville, and E Yap. We
`thank other colleagues in our departments for giving us
`samples and for general cooperation and encouragement.
`
`Appendix
`Urine samples were centrifuged at 2000 rpm for 10 minutes,
`and messenger RNA was immediately extracted from the cell
`pellet using Micro-Fast Tract (Invitrogen). Complementary
`DNA was synthesised from the messenger RNA template
`using the complementary DNA Cycle Kit (Invitrogen), and
`amplification was performed with appropriate primers and
`parameters using 2-5 units of Taq polymerase in 50 ,ul
`reaction mixture. In this study of urine samples amplification
`was performed across a shorter section of the CD44 comple-
`
`Clinical implications
`
`* Early stage bladder cancer can be effectively
`treated by resection, but it is often asymptomatic
`so that affected paients who could benefit are
`difficult to identify
`* With gene amplification techniques abnormal
`activity of the CD44 gene has been found in
`tumour tissue
`* In this study abnormal CD44 activity was
`confirmed with improved techniques in tumours
`from various tissues including breast, colon,
`and bladder
`* Abnormal CD44 gene activity was also found
`in exfoliated cells in urine samples from 91% (40/
`44) of patients with bladder cancer, while no
`such abnormality was found in samples from
`83% (38/46) of non-neoplastic controls
`* Technical improvements to the assay should
`substantially increase specificity of detection,
`and this non-invasive technique may be suitable
`for early detection of primary bladder tumours
`and monitoring for recurrent disease
`
`mentary DNA in each sample to increase sensitivity and
`specificity. Primers were designed to amplify across segments
`which included the new exon on the grounds that this is
`present in the largest transcripts typical of neoplasia.
`The primers used in the study, at 1 pmol/,ul in a 50/,ul
`reaction, were P1 (5'-GACACATATTGCTTCAATGCTT-
`CAGC),P2(5'-CCTGAAGAAGATTGTACATCAGTCAC-
`AGAC), P3 (5'-TGGATCACCGACAGCACAGAC), P4 (5'-
`GATGCCAAGATGATCAGCCATlTCTGGAAT), El (5'-
`TTGATGAGCACTAGTGCTACAGCA),E2(5'-CATTTG-
`TGTTGTTGTGTGAAGATG),E3(5'-AGCCGAGAGGAC-
`AGTTCCTGG),E4(5'-TGAGATTGGGTTGAAGAAATC),
`andE5 (5'-TCCTGCTTGATGACCTCGTCCCAT). Figure
`1 shows the positions of the primers and probes, and the
`predicted size of certain amplification products obtained with
`different primer combinations were 482 base pairs (primers
`P1 and P4, standard part only), 744 base pairs (P3 and E4 with
`exon 6), 615 base pairs (P3 and E4 without exon 6), and 735
`base pairs (El and E5, cancer related band consisting of exons
`6, 7, 8, 9, 10, and 11).
`The complete complementary DNA solution from each
`urine sample was divided equally into two tubes: one for
`reaction with primers El and E5, to amplify the complemen-
`tary DNA transcript that was of diagnostic value (735 base
`pairs); and one for reaction with primers P1 and P4, to
`amplify the standard form of CD44 (482 base pairs) as
`an internal control of messenger RNA quality and comple-
`mentary DNA synthesis. Both tubes underwent 35 cycles of
`amplification. The cycle conditions were: 94°C for one
`minute, 55°C for one minute, and 72°C for two minutes. After
`electrophoresis the amplified products were electrophoresed,
`blotted, and probed-with phosphorus-32 labelled probe E4
`to visualise the 735 base pair band or with P2 to view standard
`form CD44.
`The TA plasmid DNA containing a complementary DNA
`insert of the same composition as the one of diagnostic interest
`(containing transcripts from all exons from 6 to 11 inclusive)
`was used as an internal standard for quantification purposes.
`This calibration series consisted of the following serial
`dilutions of 10-17, 10.-8, 10-0', 10-20, 10-21, 10-22, and 10-23 mole of
`the plasmid DNA in 50,p1 of reaction solution (primers El
`and E5, probe E4). Conditions for hybridisation, high
`stringency washing, and autoradiography with x ray film
`(Kodak) were according to standard protocols.
`
`1 KantoffPW. Bladder cancer. CurrProbl Cancer 1990;14:233-92.
`2 Winawer SJ, Schotten-field D, Frehinger BS. Colorectal cancer screening. Jf
`Natl Cancer Inst 1991;83:243-53.
`3 Matzkin H, Moinuddin SM, Soloway MS. Value of urine cytology veraua
`bladder washing in bladder cancer. Urology 1992;39:201-3.
`4 Hedin A, Carlason L, Bergiund A, Hammarstrom S. A monoclonal antibody-
`enzyme inmmunoasaay for aerum carcinoembryonic antigen with increased
`specificity for carcinomas. Proc NadAcadSci USA 1983;80:3470-4.
`5 Matsumura Y, Tarin D. Significance of CD44 gene products for cancer
`diagnosis and diaease evaluation. Lancet 1992;340: 105