Context.—Fluorescence in situ hybridization (FISH) is a molecular cytogenetic assay that is commonly used in laboratory medicine. Most FISH assays are not approved by the US Food and Drug Administration but instead are laboratory-developed tests that use analyte-specific reagents. Although several guidelines exist for validation of FISH assays, few specific examples of FISH test validations are available in the literature.

Objective.—To provide an example of how a FISH assay, using an analyte-specific reagent probe, may be validated in a clinical laboratory.

Design.—We describe the approach used by an individual laboratory for validation of a FISH assay for mixed lineage leukemia (MLL) gene.

Results.—Specific validation data are provided illustrating how initial assay performance characteristics in a FISH assay for MLL may be established.

Conclusions.—Protocols for initial validation of FISH assays may vary between laboratories. However, all laboratories must establish several defined performance specifications prior to implementation of FISH assays for clinical use. We describe an approach used for assessing performance specifications and validation of an analyte-specific reagent FISH assay using probes for MLL rearrangement in interphase nuclei.

As the fields of cancer cytogenetics and molecular cytogenetics continue to expand, new tests are being developed to identify an increasing number of chromosome abnormalities that are important for the diagnosis, prognosis, and monitoring of neoplastic diseases. Molecular cytogenetic methods, such as fluorescence in situ hybridization (FISH), are more sensitive than routine metaphase chromosome analysis because they can detect a specific abnormality in both dividing and nondividing cells and have a lower threshold for detecting small populations of abnormal cells.1,2 Thus, FISH testing has become part of routine clinical practice in many laboratories that deal with neoplastic specimens.

Introduction of a new FISH test requires assessment of performance characteristics of the assay and validation of the test by the laboratory. If the FISH test has been approved by the US Food and Drug Administration, the laboratory must independently verify the performance characteristics of the assay. However, most FISH probes used in clinical testing are not US Food and Drug Administration–approved but instead are analyte-specific reagents. Although the technical performance of these probes has been evaluated by the manufacturer prior to marketing, they must be validated as a laboratory-developed test prior to use in clinical testing. General guidelines on validation of FISH tests are available and provide an excellent introduction to issues that are particular to this methodology.37 Validation of FISH probes should include the following steps: (1) familiarization and planning, (2) assessment of performance characteristics (accuracy, precision, sensitivity and specificity, reference ranges, and reproducibility), and (3) implementation of the test. This article describes an example of how to use these steps to validate FISH probes for detecting aberrations in the mixed lineage leukemia (MLL) gene in interphase nuclei using an analyte-specific reagent manufactured by Abbott Molecular, Inc (Des Plaines, Illinois).

Prior to initiating laboratory studies, it is important to understand the biology of the target locus and its disease association(s). For example, a literature review of the MLL gene at 11q23 shows that it encodes a transcriptional regulatory factor that is a major regulator involved in hematopoietic and embryonic development that acts through regulation of HOX gene expression.8 ,MLL is the most promiscuous gene yet observed in cancer with rearrangements involving more than 50 gene partners. Abnormalities involving MLL account for approximately 10% of the chromosome abnormalities observed in acute leukemia of both lymphoid and myeloid lineages.811 

The laboratory personnel involved in testing and scoring should become familiar with the concepts underlying the probe labeling, testing strategy, and result reporting. A break-apart probe strategy is particularly appropriate for detecting MLL gene rearrangements because chromosomal breakpoints occur in a relatively narrow region of the gene regardless of translocation partner. The dual-color analyte-specific reagent probe set available from Abbott Molecular consists of differentially labeled 5′ and 3′ probes that flank the common breakpoint region in the MLL locus. A chromosome translocation disrupts the normally juxtaposed probes, creating 2 widely separated signals that can be distinguished as different colors. Test results are reported qualitatively as either negative or positive for MLL rearrangement, but the number of negative or positive cells can also be included.

Laboratorians should be familiar with the established practice and/or laboratory standards and guidelines and regulations that are applicable to the testing methodology. Helpful references include the general FISH standards and guidelines from the American College of Medical Genetics (www.acmg.net, accessed September 2010), the Clinical Laboratory Improvement Act and Amendments, and the College of American Pathologists standard checklist items that pertain to FISH.1214 An approved guideline for FISH assay verification and clinical validation prepared by the National Committee for Clinical Laboratory Standards (now Clinical and Laboratory Standards Institute) is available15 and a second edition is anticipated.

During the familiarization phase of probe validation the laboratory performs practice assays with the FISH probe set to assess its performance characteristics. The goal is for laboratory personnel to gain experience with the specific variables that affect hybridization and scoring of the break-apart probe. Most probes come with a package insert that describes a hybridization protocol optimized by the manufacturer for the particular probe set. In addition, several manuals are available that include protocols and troubleshooting guidance for FISH.1618 If the laboratory is not familiar with specific probe designs (break-apart, single fusion, double fusion, etc), several normal and abnormal specimens should be tested to gain experience with the reagents, techniques, protocols, fluorescence filters, and hybridization characteristics. For laboratories already familiar with the break-apart probe design and interpretation, several specimens should be assayed with the new probe to establish its specific hybridization characteristics and to set criteria for scoring for various signal patterns.

The MLL dual-color break-apart probe set used here as an example is available from Abbott Molecular and consists of a probe to the 5′ proximal portion of the gene and upstream flanking sequences (350 kb) labeled with SpectrumGreen and a probe to the 3′ distal portion of the gene and downstream flanking sequences (190 kb) labeled with SpectrumOrange. The latter signal may appear more red than orange and is referred to as red (R) in the descriptions of normal and abnormal probe signal patterns that follow. In a normal cell, the probe signals usually appear as 2 yellow fusion signals (FF) due to overlap of the closely spaced 5′ green and 3′ red fluorescent signals on each chromosome 11 (Figure, A). In a cell with MLL rearrangement the proximal and distal probe sequences are physically separated (usually as a result of chromosome translocation), and the probe signal pattern commonly appears as 1 green (G) signal (corresponding to the 5′ MLL locus remaining on the derivative chromosome 11), 1 R signal (corresponding to the 3′ MLL locus translocated to a partner chromosome), and 1 F signal (corresponding to the normal chromosome 11); this pattern (Figure, B) can be designated FRG for data recording purposes.

During the familiarization phase, normal cases may be used to optimize the conditions for hybridization, prehybridization and posthybridization washing, and analysis. All signal patterns observed in the normal samples should be carefully examined and recorded for future reference. Depending upon the specific probes, a normal signal pattern produced by dual-color break-apart probes may appear as a yellow F signal, as separate side-by-side RG signals with no space in between, or as separate, side-by-side RG signals spaced only a short distance apart. This variability is due in part to differing degrees of DNA condensation in normal interphase nuclei. Becoming familiar with the range of signal separation associated with a particular break-apart probe set and with the frequency of unexpected signal patterns encountered in known normal cells is critical to avoid false-positive interpretations during the next stage of the validation. Some abnormal cases known to have MLL rearrangement by either metaphase chromosome analysis or another laboratory assay should also be studied during this phase to gain familiarity with the range of abnormal signal patterns.

Criteria for evaluating the quality of hybridization and staining, selecting appropriate interphase cells for scoring, and recording signal pattern interpretations should be developed. For example, probe signal intensities should be sufficiently robust that all hybridized signals are easily recognized without interference from nonspecific hybridization or background staining. Interphase nuclei should be single, nonoverlapping, round cells or have the shape and size appropriate for the cell type targeted in the analysis. Data should be recorded in a standardized fashion. After formulating these criteria, training should be provided to the laboratory personnel involved in scoring cells for analysis prior to the next step of the validation.

After technologists and the laboratory director are familiar with the MLL FISH assay and have developed criteria for scoring normal and abnormal cells, the formal process for determining the performance characteristics of the assay begins. Prior to analysis, FISH slides are evaluated to ensure that the assay has met criteria for acceptable technical performance. Characteristics examined include adequacy and consistency of signal strength (brightness), lack of background and/or cross-hybridization signals, and presence of appropriate control (internal or external) signals. Only slides that are considered acceptable for analysis are further evaluated.

Assay performance characteristics include accuracy, precision, sensitivity, and specificity. Metaphase cells are used to localize the probe and determine its analytical sensitivity and specificity; male cells are used to rule out cross hybridization of the probe with the Y chromosome. The metaphase cells can be obtained from bone marrow from a pooled sample (prepared from 5 karyotypically normal males with similar mitotic indices) or from 5 separate karyotypically normal male samples (20 cells scored per sample). Probe localization and analytical sensitivity and specificity can be assessed simultaneously. The discussion that follows focuses on the use of MLL FISH as a qualitative test in bone marrow specimens.

Accuracy

Whether the qualitative MLL FISH assay is considered accurate is determined both by the performance of the probes and by the performance of the assay as a diagnostic test. It is important to confirm that the probes hybridize to their expected chromosomal loci (probe sensitivity) and only to their expected chromosomal loci (probe specificity) and to demonstrate positive (abnormal) results in a high proportion of cases known to have MLL rearrangement (assay sensitivity) and negative (normal) results in a high proportion of cases known to be negative for MLL rearrangement (assay specificity). Assays that are both highly sensitive and highly specific have a high degree of accuracy. The accuracy of the MLL FISH assay was determined as follows.

Probe localization and its sensitivity and specificity were determined by recording the location (11q23 target versus nontarget) of MLL probe signals in 100 intact metaphase cells. Five representative metaphase cells were captured for documentation of probe localization. MLL probe signal localization to 11q23 was determined by evaluating chromosome morphology using reverse 4′,6-diamidino-2-phenylindole dihydrochlorine (DAPI) imaging of the hybridized chromosome and/or sequential G-banding followed by FISH.

To determine the assay sensitivity and specificity, interphase signal patterns were examined in known normal and abnormal specimens (see discussion under “Reference Range” later).

Precision (Reproducibility)

Precision is defined as the closeness of agreement between independent test results. For qualitative FISH assays, such as the MLL rearrangement detection assay being described here, precision may be considered equivalent to reproducibility.7 One approach that can be used for determining reproducibility is to perform replicate assays of a normal specimen and a specimen with a known proportion of abnormal cells during several days. A comparison of the mean, standard deviation, and range of results between replicates determines the level of reproducibility. Alternatively, as in this example, once a laboratory has become very familiar with different FISH probes, a good estimate of precision and/or reproducibility for qualitative interphase FISH assays can be provided by the analytical sensitivity and specificity.3 In addition, an estimate of the interobserver and intraobserver variability encountered in scoring a FISH assay obtained while collecting data to determine probe and assay sensitivity and specificity (described later) provides another measurement of reproducibility.3 Assigning an acceptable level of reproducibility for a FISH assay can be problematic due to the relatively small number of cells analyzed in each replicate, the presence of variant signal patterns, and varying degrees of mosaicism.

Analytical Sensitivity

Considerations in determining analytical sensitivity in FISH studies include the fidelity with which the probe hybridizes to its expected chromosome target (probe sensitivity) and the performance of the assay as a diagnostic test with few false-negative results in comparison to a gold standard (assay sensitivity). Probe sensitivity is defined as the percentage of chromosome targets that have the expected normal signal pattern. Assay sensitivity is determined by examining interphase signal patterns in known normal and abnormal specimens (see discussion under “Reference Range” later).

MLL probe sensitivity was determined by calculating the proportion of 11q23 target sites that had the expected probe signal. As seen in Table 1, the expected FF probe signal pattern was localized to 11q23 in 100 of 100 metaphase cells examined (200 of 200 chromosome 11q23 targets), yielding a probe sensitivity of 100%. The American College of Medical Genetics FISH Guidelines (www.acmg.net, accessed September 2010) recommend that probe sensitivity should be at least 98%.

Analytical Specificity

Considerations in determining analytical specificity in FISH studies include the fidelity with which the probe hybridizes only to its expected chromosome target (probe specificity) and the performance of the assay as a diagnostic test with few false-positive results in comparison to a gold standard (assay specificity). Probe specificity is defined as the percentage of probe signals that hybridize to the correct locus and to no other location. Assay specificity is determined by examining interphase signal patterns in known normal and abnormal specimens (see discussion under “Reference Range” later).

MLL probe specificity was determined by examining the location (target versus nontarget) of all MLL probe signals in 100 intact metaphase cells followed by calculating the proportion of the probe signals that hybridized to the 11q23 locus. Hybridization of a probe to a nontarget locus is evidence for decreased probe specificity; this is manifest as an F or separated R or G signals at a site other than 11q23. For example, if 199 of 200 MLL probe signals scored are located at 11q23, the specificity (number of signals at expected target/total number of signals) would be 99.5%. As seen in Table 1, no cells with an abnormal signal pattern were identified, corresponding to a probe sensitivity of 100%. The American College of Medical Genetics FISH Guidelines (www.acmg.net, accessed September 2010) recommend that probe specificity should be at least 98%.

Reportable Range

This parameter is not applicable to qualitative assays.

Reference Range (Normal Cutoff)

The reference range, as defined by the American College of Medical Genetics FISH Guidelines (www.acmg.net, accessed September 2010), is the range of test values expected to occur in 95% of healthy individuals. The upper cutoff for normal results in a FISH assay can be determined by calculating the 95% confidence interval for probe signal patterns found in normal control samples that are representative of the sample type to be analyzed.

All MLL probe signal patterns in a defined number (see later) of interphase cells from each of 20 normal bone marrow samples were analyzed and recorded. Scoring criteria were used as defined in the familiarization phase. Because most laboratories do not have access to bone marrow specimens from normal volunteer donors, these samples may be difficult to obtain. As a practical alternative, karyotypically normal bone marrow specimens referred for assessment of nonneoplastic conditions that would be unlikely to harbor a MLL rearrangement (or other probe being validated) may be used for the validation.

It is useful for 2 individuals to independently score the same slides, each recording the consecutive signal patterns encountered in the first 100, 150, 200, and 250 cells. When counts are totaled, this allows for construction of a 200, 300, 400, and 500 cell database that may be valuable later should unanticipated circumstances arise in the interpretation of clinical specimens (eg, need to score additional cells to resolve discrepant results close to a cutoff that could alter the diagnostic interpretation). Table 2 illustrates one way of documenting the number of cells with various MLL signal patterns that were observed in a 200 cell analysis in each of 20 normal control specimens. It is important to record results of all probe signal patterns encountered during this phase of the validation because a different normal cutoff value can then be calculated for each signal pattern. This also provides a quality control tool with which to compare the reproducibility of results between individual readers.

The normal cutoff value for the MLL FISH assay was calculated using the beta inverse function with a probability of 95% [ = BETAINV(probability,alpha,beta,[A],[B]) available in Microsoft Excel (Microsoft Corp, Redmond, Washington)]. Variables to be entered into the function equation include probability, a desired level of probability associated with the beta distribution; alpha and beta, parameters of the distribution; A, an optional lower bound of the distribution; and B, an optional upper bound of the distribution. To complete the equation, probability equals 0.95, alpha equals the number of false-positive nuclei +1 found in the normal specimen with the greatest number of false-positive nuclei for any given signal pattern, and beta equals the total number of cells scored. For example, to calculate the normal cutoff value for the MLL signal pattern FRG listed in Table 2, note that of the 20 samples in the reference group, sample 2 has the highest number of false-positive FRG signals (4 + 3  =  7). The beta inverse function as calculated in Microsoft Excel is therefore  = BETAINV(0.95, 7+1, 200) or 6.3%. Because a given test result is considered abnormal if the percentage of nuclei with that aberrant signal pattern exceeds the 95% confidence interval of the normal reference distribution, a minimum of 13 of 200 (6.5%) cells would be required for an FGR signal pattern to be considered an abnormal result. The signal patterns and respective calculated normal cutoffs for the MLL probe set when 200 cells are scored are FRG (6.3%), FG (2.3%), FR (1.5%), F (7.4%), FFF (5.5%), FFR (2.3%), FFRG (1.5%), and FFFF (5.1%). For other less common abnormalities, such as deletion of the distal portion of the MLL gene, a minimum of 5 cells (2.5%) with a signal pattern of FG would denote an abnormal clone.

Normal cutoffs can also be calculated assuming a Gaussian distribution of results in the normal reference population for 2 or 3 standard deviations. The primary difference between the cutoffs generated by the Gaussian or beta inverse function is in the case where the number of false-positive cells is zero. Using the Gaussian statistical method, should a particular signal pattern be encountered in a patient sample that was not observed in the normal control data set, a single cell with that pattern would be considered abnormal. The beta inverse formula does not allow a cutoff of zero.19 

Ten specimens, including normal cases and cases with MLL rearrangement confirmed by another method, were then examined in a blinded fashion for detection of different MLL abnormalities in patient samples. Specimens known to have 11q23 rearrangement by cytogenetics were selected as a gold standard to validate the cutoff values. It is also helpful to test cases with known chromosome abnormalities such as variant translocations, interstitial or terminal deletions, or aneuploidy for chromosome 11 that are expected to generate variant MLL FISH signal patterns. Examples of MLL FISH signal pattern results from a test set of 4 normal cases, 3 cases with 11q23 rearrangements, 2 with loss of chromosome 11, and 1 with gain of chromosome 11 are presented in Table 3. The FISH probe results correlated well with the cytogenetic results in all cases; in this example the assay sensitivity and specificity were each 100%.

The assay may be put into clinical use only after completion of the assessment of probe performance and validation studies. Although such an extensive validation does not need to be performed on each new probe lot, each lot must be shown to have comparable sensitivity and specificity prior to being put into use for clinical testing.5,14 This may be accomplished by analyzing a patient sample with both the old and new lots to document equivalent results. Ongoing quality control of reagents and equipment and continued monitoring of assay performance is required. Comments recorded on assay worksheets concerning the size and brightness of probe signals, presence or absence of background staining and/or cross hybridization, and hybridization efficiency can be monitored for acceptability as one method of ongoing assay performance verification. Similarly, ongoing monitoring of interobserver reproducibility, accomplished in part by having 2 technologists read every case, can help detect changes in the assay performance or loss of consistency in applying scoring criteria. Participation in external proficiency testing programs is an important form of ongoing quality assurance.

1.
Secker-Walker
LMM
.
General report on the European Union concerted action workshop on 11q23
.
Leukemia
.
1998
;
12
(
27
):
776
778
.
2.
Cohen
N
,
Novikov
I
,
Hardan
I
, et al.
Standardization criteria for the detection of BCR/ABL fusion in interphase nuclei of chronic myelogenous leukemia patients by fluorescence in situ hybridization
.
Cancer Genet Cytogenet
.
2000
;
123
(
2
):
102
108
.
3.
Wiktor
AE
,
Van Dyke
DL
,
Stupca
PJ
, et al.
Preclinical validation of fluorescence in situ hybridization assays for clinical practice
.
Genet Med
.
2006
;
8
(
1
):
16
23
.
4.
Test and Technology Transfer Committee
.
Technical and clinical assessment of fluorescence in situ hybridization: an ACMG/ASHG position statement. I. Technical considerations
.
Genet Med
.
2000
;
2
(
6
):
356
355
.
5.
CLSI
.
User Verification of Performance and Precision of Trueness; Approved Guideline. 2nd ed
.
Wayne, PA
:
Clinical and Laboratory Standards Institute
;
2006
.
CLSI document EP15-A2.
6.
Von Bergh
A
,
Emanuel
B
,
Van Zelderen-Bhola
S
, et al.
A DNA probe combination for improved detection of MLL/11q23 breakpoints by double-color interphase-FISH in acute leukemias
.
Genes Chromosomes Cancer
.
2000
;
28
(
1
):
14
22
.
7.
Jennings
L
,
Van Deerlin
VM
,
Gulley
ML
.
Recommended principles and practices for validating clinical molecular pathology tests
.
Arch Pathol Lab Med
.
2009
;
133
(
5
):
743
755
.
8.
Bernard
OA
,
Berger
R
.
Molecular basis of 11q23 rearrangements in hematopoietic malignant proliferations
.
Genes Chromosomes Cancer
.
1995
;
13
(
2
):
75
85
.
9.
Rubnitz
JE
,
Behm
FG
,
Downing
JR
.
11q23 rearrangements in acute leukemia
.
Leukemia
.
1996
;
10
(
1
):
74
82
.
10.
Young
BD
,
Saha
V
.
Chromosome abnormalities in leukemia: the 11q23 paradigm
.
Cancer Surv
.
1996
;
28
:
225
245
.
11.
Pui
CH
,
Chessells
JM
,
Camitta
B
,
Baruchel
A
.
Clinical heterogeneity in childhood acute lymphoblastic leukemia with 11q23 rearrangements
.
Leukemia
.
2003
;
17
(
4
):
700
706
.
12.
Public Law 100-578, Clinical Laboratory Improvement Amendments of 1988, October 31, 1988
.
Fed Regist
.
1992
;
57
(
40
):
7002
7186
.
13.
Schwartz
MK
.
Genetic testing and the Clinical Laboratory Improvement Amendments of 1988: present and future
.
Clin Chem
.
1999
;
45
(
5
):
739
745
.
14.
Commission on Laboratory Accreditation
.
Laboratory Accreditation Program: Cytogenetics Checklist
.
Northfield, IL
:
College of American Pathologists
;
2009
.
15.
NCCLS
.
Fluorescence In Situ Hybridization (FISH) Methods for Medical Genetics; Approved Guideline
.
Wayne, PA
:
National Committee for Clinical Laboratory Standards
;
2004
.
NCCLS Document MM7-A.
16.
Barch
MJ
,
Knudson
T
,
Spurbeck
JL
,
eds
.
AGT Cytogenetics Laboratory Manual. 3rd ed
.
Philadelphia, PA
:
Lippincott-Raven
;
1997
:
557
590
.
17.
Fluorescence in situ hybridization
.
Nat Methods
.
2005
;
2
(
3
):
237
238
.
18.
Rooney
DE
,
ed
.
Human Cytogenetics. 3rd ed
.
Oxford, England
:
Oxford University Press
;
2001
:
175
209
.
19.
Ciolino
AL
,
Tang
ME
,
Bryant
R
.
Statistical treatment of fluorescence in situ hybridization validation data to generate normal reference ranges using Excel functions
.
J Mol Diagn
.
2009
;
11
(
4
):
330
333
.

Author notes

From the Department of Pathology and Laboratory Medicine, Emory School of Medicine, Atlanta, Georgia (Dr Saxe); the Department of Pathology and Laboratory Medicine, University of Kansas Medical Center, Kansas City (Dr Persons); the Department of Pathology and Laboratory Medicine, Medical University of South Carolina, Charleston (Dr Wolff); and the Department of Clinical Pathology, Cleveland Clinic, Cleveland, Ohio (Dr Theil).

The authors have no relevant financial interest in the products or companies described in this article.