perspectives - ENVIS

0 downloads 0 Views 118KB Size Report
Quality Control in the Development of Transgenic Crop Seed Products. Rita Hogan Mumm* and Donald S. Walters. ABSTRACT their plant products at various ...
PERSPECTIVES Quality Control in the Development of Transgenic Crop Seed Products Rita Hogan Mumm* and Donald S. Walters ABSTRACT

their plant products at various points in seed and food production (Beever and Kemp, 2000). Grain handlers must develop the capability to evaluate the presence or proportion of transgenic grain in truckloads and other grain lots as nontransgenic grain takes on identity-preserved status. Likewise, food processors need to effectively monitor supply chains within identity preservation systems to maintain compliance with the labeling requirements demanded by some markets. Furthermore, the efficient evaluation of the type of any transgenic grain present in supply chains is important to ensure compliance with governmental regulation. The auditing requirement actually begins with the seed companies, which develop crop seed products for retail markets. Presently, seed certification requirements are expanding to incorporate verification of transgenic traits for which price premiums are often charged, impacting the production of commercial seed volumes to be sold to farmers (Thompson, 2000). At a more fundamental level, the process of TSPD figures prominently as the point at which newly-created transgenic plants are channeled into the research and development (R&D) pipeline and first introduced into the environment. A broad range of tests is conducted to evaluate commercial potential, safety, and the environmental impact of the transgenic materials in an effort to identify an appropriate source of the trait of interest, to obtain governmental approvals for commercialization, and to select new products for market launch. Because TSPD typically involves a number of candidate transgenic sources before a commercial source is finally identified, the authenticity of transgenic materials channeled through the R&D pipeline must be confirmed to ensure timely market launch and the integrity of new products. Quality control monitoring of the TSPD process is essential to protecting research investments, maintaining compliance with governmental regulations, and ensuring customer satisfaction.

In light of (i) public concern with the safety of genetically modified crops, (ii) the issue of food labeling, and (iii) governmental regulation of transgenic plants, quality control (QC) monitoring of the process of Transgenic Seed Product Development (TSPD) is essential to protecting research investments, maintaining compliance with governmental regulations, and ensuring customer satisfaction. The primary goal of QC monitoring in TSPD is to ensure the authenticity (transgenic event identity and purity) of seed materials used in product testing, in the development of regulatory data packages for governmental review, and to develop seed volumes for commercial release. Monitoring is performed to confirm the presence of the presumed transgenic event(s) and the absence of all others. Sophisticated QC strategies formulated to monitor the product development process and to maintain quality standards in the manufacturing industry can serve as a foundation in devising efficient strategies tailored to meet needs in the seed industry. A number of considerations in the design of an optimized QC monitoring plan for TSPD are discussed, including: costs vs. economic benefits of developing and commercializing nondefective products; effective timing of inspections; joint scheduling of inspections with other genetic analyses; assay methods that maximize screening efficiency, flexibility, and accuracy; sampling vs. 100% assessment of transgenic populations; the ability to bulk samples to minimize data point requirements; and the need to minimize or eliminate inspection errors. An overview of the general process of TSPD and an example QC monitoring strategy for corn (Zea mays L.) are provided.

T

ransgenic crop seed products are created through the use of molecular techniques that facilitate the introduction of new, useful traits to target crop species. DNA encoding for a trait of interest is identified and isolated from a source species (bacterial, viral, plant, or animal), and introduced to the target crop species via transformation. Typically, the gene to be introduced is modified and placed under the control of a designated promoter to facilitate optimal trait expression in the target species. Thus, transgenic crops have been genetically modified to express genes not found in their native gene pool or to express native genes in a novel manner. Public concern with the safety of transgenic crops, the issue of food labeling, as well as governmental regulation of transgenic plants by U.S. and international agencies has spawned much debate over measures required to assess genetic components of these plants and

The Process of Transgenic Seed Product Development The general process of TSPD encompasses a number of steps from transformation to new product launch, Abbreviations: BC1, generation resulting from the first backcross to the recurrent parent (BC2–BC6 follow the same convention with the specific generation designated); EPA, Environmental Protection Agency; FDA, Food and Drug Administration; PCR, polymerase chain reaction; QC, quality control; R&D, research and development; S1, first generation after self-pollination (S2–S3 follow the same convention with the specific generation designated); TSPD, Transgenic Seed Product Development; T0, generation resulting from regenerated transformed cells; T1, first generation after transformation; T2, second generation after transformation.

R.H. Mumm, GeneMax Services, 4002 Turnberry Drive, Champaign, IL 61822; D.S. Walters, DuPont Crop Genetics, Experiment Station, Room 402E/3224, Wilmington, DE 19880. Research supported in part by Exygen Research, 30158 Research Drive, State College, PA 16801. Received 30 Oct. 2000. *Corresponding author (rita.mumm@ genemaxservices.com). Published in Crop Sci. 41:1381–1389 (2001).

1381

1382

CROP SCIENCE, VOL. 41, SEPTEMBER–OCTOBER 2001

Fig. 1. The general process for Transgenic Seed Production Development (TSPD) in corn from transformation to new product launch involves a number of teams and functions, with seed and plant materials produced by Biotech Trait Development and Biotech Breeding serving as the basis for tests leading to commercialization decisions and government deregulation as well as new products per se.

and generally involves several functional teams across an agricultural seed company or other R&D organization. Figure 1 features a typical scenario for corn, with

developmental and commercialization activities performed by Biotech Trait Development, Biotech Breeding, Testing, Regulatory Compliance, and Supply Man-

MUMM & WALTERS: QUALITY CONTROL IN TRANSGENIC CROP SEED DEVELOPMENT

agement groups. While the activities listed are basic to the process, there is some malleability as to their sequence and the importance attached to their implementation. The structures of the teams involved in TSPD implementation may also be quite variable. A number of developmental and commercialization activities are typically done in parallel to minimize time to commercial launch of new products. Furthermore, the process of TSPD is generally structured to obtain as much information about the candidate transgenic sources as early as possible, so that resources can be focused on those that offer the most commercial promise. TSPD generally begins with production of a large number of candidate transgenic sources, or “events”, for a given trait. Each event is uniquely defined by both the actual DNA sequence that has been integrated within the target genome via transformation, as well as the chromosomal location(s) of the DNA insertion(s). Without the use of site-specific insertion technologies (e.g., recombination-based systems like Cre/lox) in transformation, the chromosomal location of the DNA insert is random. Likewise, the sequence of the integrated DNA is not generally under rigorous control, and often differs significantly from the sequence of the DNA fragment introduced in transformation. For example, transgenic inserts may contain several copies of the introduced genes, sometimes comprising a complex structure of rearranged fragments in various directional orientations (Register et al., 1994). Both the sequence and the position of the transgenic insert can affect trait expression. Since the vast majority of events will not meet the performance standards for commercialization, it is necessary to generate a large number of events to maximize the likelihood of producing and identifying at least one with commercial value. As site-specific insertion technologies are refined, the number of events needed to ensure success in TSPD will likely be reduced. Biotech Trait Development scientists perform the transformations and regenerate plants from transformed cells (T0 plants) (Fig. 1). Only those events for which regeneration results in fertile T0 plants can be advanced further in TSPD. Characterization of events in terms of the number of chromosomal insertion sites and the general structure and complexity of each (e.g., the number of copies of introduced genes or portions thereof, and orientation) is often initiated with the T0 or the T1 (first generation after transformation) generation. Although the underlying mechanisms are not fully understood, research has demonstrated that events comprising a single integration with one or two intact copies of the transgene (i.e., minimal deletion and rearrangement) tend to be more genetically stable and, hence, more reliable in terms of trait expression (Finnegan and McElroy, 1994). Nonreciprocal recombination, and both pre- and posttranscriptional modifications have been identified as sources of genetic instability, which can result in gene silencing in transgenic organisms (Bruening, 1998; Holliday and Ho, 1998; Waterhouse et al., 1998; Puchta et al., 1995). Thus, event characterization data, which are required for some government regulatory approvals, can also be used as an aid in identifying

1383

events with the greatest commercial promise. Fertile T0 plants are typically crossed to one or more elite inbred lines to initiate introgression of events into an array of commercially representative genetic backgrounds. T1 seed is produced and forwarded to the Biotech Breeding Team to expand and advance elite line conversion and, supplies permitting, to the Testing Team. If T1 seed quantities are insufficient to facilitate transfer to Testing, the Biotech Breeding Team generally produces seed to meet initial testing needs. The Testing Team conducts a series of tests to assess the commercial value of candidate events, focusing on trait efficacy as well as overall agronomic performance (Fig. 1). The tests result in a stepwise reduction in the number of events advanced through TSPD, enabling effective resource allocation to the events with the greatest commercial promise. Trait efficacy evaluations focus on whether the trait of interest is expressed at commercially useful levels in appropriate plant tissues at suitable times in plant development. These evaluations may be preceded by “event sorting”, exercises aimed at identifying and discarding events obviously lacking desired trait expression. Event sorting may result in the elimination of up to 50% of the events screened, those representing the lower performance spectrum. Event sorting is typically conducted at the T0, the T1, or the T2 (second generation after transformation) stage, the specific timing depending on a number of factors, including the nature of the specific trait of interest and the tissues and life stage of the plant involved in trait expression, the type of accessible screening assays, the vigor of T0 plants and their general ability to withstand selective pressures, the number of T0 plants produced per event, and the seasonal time lag if any until efficacy testing can be performed. It is easy to see how the nature of the trait and the intended products (e.g., insect or disease resistance, enhanced grain quality, environmental stress resistance, or novel biosynthetic products) strongly impacts the design of the testing plan. For example, some introduced traits such as disease and insect resistance may be previewed in T0 plants prior to reproductive stages, facilitating early event sorting. Traits affecting grain composition, such as enhanced oil or modified amino acid levels, may require the production of seed before any event sorting or efficacy evaluation is possible. Some environmental stress resistance traits may preclude event sorting prior to efficacy testing unless methods can be established to effectively simulate stress conditions in screening environments in a highly repeatable fashion. Drought tolerance is one such example. Events are evaluated for overall agronomic performance to test for any unwanted effects due to either the biological activity of the transgene or the chromosomal location of the transgenic insertion. For example, an event could involve an insertion within a gene influencing grain yield, knocking out normal expression of that gene and resulting in substandard yield. Typically, final agronomic performance evaluations are conducted using candidate products grown in a large number of locations representative of prospective regional markets to determine whether plant performance is maintained

1384

CROP SCIENCE, VOL. 41, SEPTEMBER–OCTOBER 2001

across the potential range of field environments. These tests are generally referred to as Wide Area Hybrid Tests. The candidate products typically represent a range of genetic backgrounds to facilitate a comprehensive assessment of potential interactions between the event and other genes comprising the crop genome. Often, the candidate hybrids are compared directly with genetically similar, nontransgenic hybrids. Preliminary agronomic performance evaluations may be conducted with materials (e.g., testcross hybrids) produced using incomplete conversions of elite lines (see text in the second paragraph following in this section for a discussion of the conversion process). However, in these preliminary evaluations, only relatively large effects associated with events are generally detectable, given the lack of genetic uniformity in such conversions. Testing, as described above, is the primary mechanism for identifying and selecting the event to commercialize, but regulatory studies conducted to assess the safety and environmental impact of events figure prominently in decisions as well. For example, the results of studies on molecular characterization and genetic stability of the transgenic inserts are often considered key criteria for event advancement. Regulatory studies complement evaluations performed by the Testing Team to assess the commercial viability of events and, while these are conducted somewhat in parallel, the latter provides guidance as to which few events to include in the more costly regulatory studies conducted downstream (e.g., environmental studies and expression studies). Furthermore, the Testing Team provides additional feedback, as well as valuable field performance data as regulatory packages are assembled for U.S. governmental agencies, including USDA, EPA, and FDA, and agencies of other countries for which import approval or approval for seed product launch is sought. Likewise, Testing provides feedback to the Biotech Breeding Team, directing continued advancement of only the most commercially promising events. The primary objective of Biotech Breeding is to produce improved versions of proprietary inbreds through backcross breeding, which facilitates event evaluation in genetic backgrounds representative of potential commercial products and the development of candidate products (Fig. 1). In corn, one or both parents of hybrids may be targeted for conversion depending upon the anticipated degree of dominant gene action of the transgene and the seed company’s proposed strategy for combining transgenic traits in future products. Traditionally, following genetic theory, seven consecutive generations of crossing to a recurrent parent (i.e., the BC6 generation) are required to recover conversion lines that are ≈99% genetically similar to the target inbred. However, the use of molecular markers to select within early backcross populations can effectively reduce the number of generations necessary to recover the recurrent parent germplasm and, thereby, accelerate product development (Johnson and Mumm, 1996). Virtual recovery of the recurrent parent germplasm is followed by up to three generations of self-pollination to fix transgenic trait expression in the conversion lines.

Most of the seed for both testing and regulatory studies (other than perhaps T1 seed) is typically provided by the Biotech Breeding Team (Fig. 1). Ideally, seed for testing and regulatory studies is produced using finished, nonsegregating backcross conversions (as described above). However, to hasten TSPD and focus on the most commercially promising events, seed produced at any or all generations in the conversion procedure is often used for evaluations. Inbred or testcross seed, which is either segregating or nonsegregating for the trait of interest, can be utilized for most types of evaluations provided that experimental designs are tailored to account for limitations. One notable exception is Wide Area Hybrid Testing. Finished, nonsegregating conversions are required to produce candidate product hybrids and to conduct final event evaluations. In concert with the production of the hybrid seed for Wide Area Testing, BC6S3 seed (or its genetic equivalent) is produced for consignment to Supply Management (Fig. 1). The Supply Management Team increases quantities of finished inbred conversions, and also produces commercial volumes of hybrid seed in keeping with marketing objectives. These activities are often conducted in parallel with Wide Area Hybrid Testing, the results of which ultimately identify the event to be commercialized and the specific new product offerings. Thus, Supply Management may be dealing with a number of events or candidate products that ultimately will not proceed to commercialization. Market launch of transgenic products conferring a new trait is contingent upon results from Wide Area Testing, the availability of commercial seed volumes, and governmental approvals for deregulated status of the selected event (Fig. 1). Clearly, the division of labor and technical specialization necessary for successful TSPD within a seed company mandates a high level of coordination and communication between the functional teams involved to ensure that product development goals are approached effectively. Also key is the authenticity of the seed and plant materials which serve as the basis for decisions leading to commercialization and represent the actual source of new products per se. In the process of TSPD, the Biotech Trait Development and Biotech Breeding Teams typically produce the seed and other plant materials for efficacy and performance testing, for regulatory studies, and for product release (through Supply Management). Clearly, QC auditing of materials advanced through and distributed by these two groups is essential to the successful commercial release of tested, valued, and government-approved transgenic seed products and, ultimately, to the success of the seed company at large.

The Goals of Quality Control in Transgenic Seed Product Development The primary goal of QC in TSPD is to ensure the transgenic composition of seed materials advanced through Biotech Trait Development and Biotech Breeding, and seed materials forwarded to other functional teams within the company. The authenticity and accurate la-

MUMM & WALTERS: QUALITY CONTROL IN TRANSGENIC CROP SEED DEVELOPMENT

beling of advanced and transferred materials are essential to efficient operations and timely commercial launch. A parallel but related goal is to evaluate the genetic stability of each event across generations. That is, to confirm that both the sequence of the integrated DNA and the resulting trait phenotype does not change over time. These efforts are directed at minimizing risks involving field performance or market suitability that might impede successful commercialization of the transgenic seed product. A large body of knowledge related to process monitoring and acceptance sampling has been developed and refined for the manufacturing industry. With appropriate modification, this knowledge can be tapped to design an optimal system in terms of both efficiency and cost to achieve QC goals in TSPD. The backcross breeding procedure, often used in TSPD, can be considered a “process”, not unlike those modeled in industrial QC settings. That is, this procedure is a stepwise progression, consisting in our corn example of a series of ear-to-row or row-to-row plantings of successive generations incorporating the new trait, from the T0 generation to finished conversions. It results in the creation of materials ultimately sold as products, as well as those serving as the basis for commercialization decisions. Consider the following scenario, which is typical of TSPD in grain corn: For development of each potential transgenic product, a number of genes may be utilized in transformation to create a pool of candidate events. For each of these genes, a number of vectors may be assembled, comprising various combinations of promoters and other genetic elements that influence expression and targeting of the transgene product. In addition, one or more selectable markers and their associated genetic regulatory elements may be included in the vector or used to produce other vectors for cotransformation. Finally, for each vector or set of vectors used in transformation or cotransformation, a large number of events are produced. As a result, the pool of events representing a given trait is generally sizeable at the T0 phase. The number of events decreases as events are advanced through breeding, and further decreases prior to parent seed increase and hybrid production in accordance with feedback from testing. Typically, it is only at the final stages of TSPD that the focus narrows to a single event for commercial launch. As the number of events decreases, generally the number of elite lines being converted for each event is increased. Once the event to be commercialized is identified, breeding often encompasses a large number of target elite lines. While this scenario details the scope for a single trait, often within a particular crop program at a seed company, a large number of transgenic traits may be under development simultaneously. For example, a seed company may be developing corn products conferring various herbicide resistances, insect resistances, disease resistances, stress tolerances, grain quality characteristics, and agronomic traits. Many or all of the events associated with a large array of traits may be advanced in the same nursery locations and even in the same plantings. Within a given nursery devoted to Biotech Breeding,

1385

a row or set of rows represents a particular conversion at a given generation of breeding, and each conversion is considered a separate entity or “item”. A “defective” item is one in which the transgenic composition does not meet expectations. That is, the presumed event is absent, has been replaced by another event, or is present along with a contaminant event(s) in at least one plant or seed. Several potential sources of contamination or error can result in a defective conversion, including pollination errors, pollen migration, planting errors, errors in creating the field computer record or nursery book, errors in packeting seed or labeling seed packets prior to planting or seed storage, misplaced row markers in field plots, and volunteer transgenic plants in nursery rows. Once contamination or error occurs, its effect may be carried forward through subsequent generations of that particular conversion line, with the proportion of defective plants in any one generation equal to x, where 0 ⱕ x ⱕ 1 (note that the proportion may be zero if the defective plants in the previous generation were not used as parents or did not produce seed). A finished conversion containing any proportion of defective plants is considered defective. On the other hand, a “good” conversion is one in which the genetic composition does meet expectations, with authenticity represented both in terms of the particular event contained therein (event identity) and the proportion of transgenic plants (purity). The primary aim of QC is to confirm the presence of the presumed transgenic event(s) and the absence of all others in a given conversion. The technical ability to distinguish between different events that confer the same trait is essential to fulfillment of this objective. The technical ability to confirm the absence of events associated with other traits is also crucial and, to meet this objective, phenotypic assays may be useful. However, phenotypic assays may not always be feasible in the nursery environment or efficient for timely product development, especially when the assays cannot be performed on seedling plants. Furthermore, phenotypic assays do not have capability to reveal the presence of silenced transgenes or nonfunctional transgenic events. Thus, assays based on genotype rather than phenotype may offer more flexibility, efficiency, and accuracy. The related goal of confirming the genetic stability of events necessitates data collection through successive generations. With efficient system design, this goal can be accommodated in the auditing procedures established to meet the primary QC goal. While the focus of this discussion is QC monitoring in TSPD, a seed company may wish to consider QC monitoring of new nontransgenic lines in product development to confirm the absence of transgenic events, particularly if the same nursery or seed storage facilities are used for both transgenic and nontransgenic materials in development.

The Cost Versus the Economic Benefits of Quality Control Auditing The cost versus the economic benefits of QC auditing in TSPD can be evaluated in a straightforward manner.

1386

CROP SCIENCE, VOL. 41, SEPTEMBER–OCTOBER 2001

Deming (1986) and others have shown that costs are minimized by treating all items identically, either inspecting all or processing all without inspection, for models wherein product characteristics are identically and independently distributed and the overall inspection cost is the sum of individually and identically determined costs for each of the items encountered. The product characteristic considered in TSPD is the presence of defective plants in a given generation of a particular conversion. The QC monitoring process must be sensitive enough to detect contamination or error introduced randomly. Monitoring costs are assumed to be uniformly distributed across conversions for a given vector (or set of vectors in the case of cotransformation). Adapting models developed by Deming (1982, 1986) and Vander Wiel and Vardeman (1994), the cost/benefit ratio for QC inspection in TSPD depends on: (i) P, the probability of a conversion being defective; (ii) wD, the probability that a defective conversion passes inspection; (iii) 1 ⫺ wD, the probability that a defective conversion fails inspection; (iv) wG, the probability that a good conversion fails inspection; (v) kI, the cost to inspect a conversion; (vi) kDU, the cost when a defective conversion is advanced without inspection; (vii) kDF, the cost when a defective conversion fails inspection; (viii) kDP, the cost when a defective conversion passes inspection; and (ix) kGF, the cost when a good conversion fails inspection. If p ⬎ pC, then inspection is warranted, where pC is given by: pC ⫽ [kI ⫹ (wG)(kGF)]/K, where K ⫽ (1⫺wD)(kDU ⫺ kDF) ⫹ (wD)(kDU – kDP) ⫹ (wG)(kGF) The critical value pC is a function of K; larger values of K lead to smaller values of pC and an increased likelihood that inspection will be cost effective. Note that (1 ⫺ wD ) represents the sensitivity of the test (the probability of detecting defectives) and that the specificity of the test (the probability of correctly identifying good conversions) is equivalent to (1 ⫺ wG ). Consider the case where the sensitivity and the specificity both equal 1. That is, QC auditing results in no inspection errors. Then wD ⫽ wG ⫽ 0, and the equation reduces to pC ⫽ kI/(kDU – kDF) This special case demonstrates that the key element of the cost analysis is the difference between the cost of a defective item being advanced without inspection and the cost of a defective item being detected through inspection. In TSPD, this cost differential is likely to be substantial because the cost of a defective item being advanced without inspection can be enormous. The success and timetable for testing, regulatory submissions to authorizing agencies, and the consignment to Supply Management all depend on the quality and authenticity of the materials obtained from Biotech Trait Development and Biotech Breeding. If a defective conversion is advanced to a late stage of TSPD without inspection, the

most important cost factor is likely to be the time to market lost and the associated loss of revenue and market share that could have been realized with a timely commercial launch. A number of devastating scenarios could result from advancement of defective materials in TSPD. Contamination or misidentification of an event early in TSPD could compromise all elite line conversions. An error that results in an initial misidentification of the most commercially promising events could invalidate years of effort directed toward new product launch. Contamination or misidentification of seed produced for safety testing could compromise timely government deregulation and delay product commercialization for years. In the highly competitive seed industry, delays in product introduction could be disastrous to a seed company, particularly when high-demand products are marketed by competitors. Worse still may be the consequences of commercializing a product that contains an event not approved by regulatory agencies, whether in place of or in addition to the intended event. The resulting loss of public confidence and professional image could be severe and long lasting. Beyond these, numerous other less devastating, but costly, genetic purity problems can arise in the absence of a diligent quality control effort. Other considerations affecting the cost differential include costs for regulatory studies and efficiency and agronomic performance testing conducted with misidentified materials, nursery row costs and production costs incurred by advancing misidentified materials, and the cost of wasting resources on materials without commercial potential, especially when dealing with resources of fixed size like continuous nurseries. The cost of releasing a defective product or advancing such to final stages of TSPD due to lack of inspection may be difficult to precisely estimate but, clearly, is very great. In the instance where there is some level of inspection error in the QC monitoring process, the full equation must be applied to accurately reflect the cost/benefit relationship. The inability to detect defective plants 100% of the time diminishes the impact of the difference between the cost of a defective item being advanced without inspection and the cost of a defective item being detected through inspection. Furthermore, it introduces the likelihood that the second term of K will be negative. These effects lead to smaller values of K, larger values of pC, and a decreased likelihood that inspection would be cost effective. Practically speaking, the accuracy of inspection must be determined by verifying assay results with materials of known transgenic status (Irwig et al., 1994). Furthermore, estimation of p requires some process monitoring data to be available. If such data indicate that p vacillates widely between values greater than and less than pC, then process improvement rather than process monitoring is warranted (Vander Wiel and Vardeman, 1994).

Quality Control Strategies and Assays Once the need for QC auditing is established, the particular strategy to be implemented must be deter-

MUMM & WALTERS: QUALITY CONTROL IN TRANSGENIC CROP SEED DEVELOPMENT

1387

Table 1. Example of a quality control (QC) auditing strategy for corn, featuring fixed interval inspections conducted in concert with other analyses either required by Transgenic Seed Product Development (TSPD) or that optimize the system. Generation

QC Assay/Purpose

Complementary Assay/Purpose

T0 or T1 BC1 or BC2

Develop event fingerprint to establish reference for future audits Ensure event identity and purity of materials used in breeding and for testing and regulatory studies Evaluate event identity and purity of conversions prior to selfing and production of additional test materials Evaluate event identity and purity of finished conversions prior to consignment to Supply Management and production of seed for Wide Area Hybrid Testing Confirm genetic stability of the given event

Initiate event characterization to assess insert complexity

BC4 or BC5 or BC6 BC6S2 or BC6 S3

mined. Here again, the cost vs. the economic benefit of inspection is the major driver in strategic design. From a business standpoint, the most critical juncture for inspection is prior to consignment of transgenic conversions to Supply Management (i.e., before seed quantities are increased to prepare for and produce commercial volumes). However, other inspections upstream may be cost-effective additions, especially inspections staged just prior to product development steps that require greater research investments (e.g., seed production for expensive regulatory studies, expansion of the number of nursery rows devoted to each conversion) or that are associated with event selection (e.g., seed production for testing) or that impact commercialization timelines and the schedule for new product launch (e.g., regulatory studies required for timely government approval). The cost-effectiveness may be further optimized by staging QC inspections to be done in conjunction with other genetic analyses that may be required by TSPD or that optimize the system. By conducting respective analyses with the same plant materials, duplication of expenses and efforts associated with sample collection, shipping, and preparation can be avoided. The following example features a fixed interval QC approach for TSPD in corn wherein inspections are conducted in four predetermined generations, three of which are scheduled in concert with other assays (Table 1). The first QC inspection is performed in the T0 or T1 generation to generate unique identifiers for individual events as a reference for future monitoring. This inspection fits conveniently with event characterization procedures generally performed early in TSPD to assess insert complexity. A second QC inspection conducted in an early backcross generation provides assurance of the event identity and purity of materials advanced through breeding and used to produce seed for testing and regulatory studies. A third QC inspection performed in a late backcross generation provides assurance of the event identity and purity of conversion lines just prior to selfing generations which demand a significant increase in the number of nursery rows. Often, final backcross generations are used to produce seed for testing and regulatory studies, further supporting an inspection in this timeframe. This inspection can be effectively conducted in concert with an analysis to evaluate the quality of individual conversions. A molecular marker analysis to estimate the recovered proportion of the recurrent parent germplasm is useful in predicting the overall agro-

Assess genetic similarity to elite line target used as recurrent parent in backcrossing Develop baseline whole-genome fingerprint to capture the genetic profile of the finished conversion which will serve as a parent of potential commercial product

nomic performance of the conversion relative to its nontransgenic counterpart (note that this analysis differs from the marker analysis related to accelerated backcross conversion mentioned in “The Process of Transgenic Seed Product Development” section, in being retrospective rather than proactive regarding recovery of recurrent parent germplasm). A fourth QC inspection is performed just prior to production of materials for Wide Area Hybrid Testing and consignment of finished conversions to Supply Management. This inspection can be conducted in concert with a molecular marker assay to capture a baseline whole-genome profile, which serves as a unique identifier of the commercial candidate hybrid parent. This profile is useful in screening subsequent seed lots and in patenting the new line, should the candidate product be commercialized. The genetic stability of events can be assessed using data from each QC audit. Observations across generations can be reviewed as available or collectively at the time of the last inspection. An alternative to the fixed interval sampling strategy discussed in the previous example is variable sampling, in which the auditing intervals are determined in response to results obtained with previous inspections. Consider, for example, that a relatively large number of defective conversions found within a given nursery could reflect a systematic error in pollinating, harvesting, storing or inventorying the previous nursery, or in packeting or planting the current nursery. Depending on the actual source of the error, there may be an increased likelihood of contamination in the next generation (i.e., both proportion of defective plants per conversion and number of defective conversions) which could impact the authenticity of seed produced for testing or regulatory studies or remnant backcross seed inventoried for possible future use. Thus, above a predetermined threshold, the number of defective plants or conversions in a given generation could trigger either more intense inspection at the next sampling interval or the addition of supplemental sampling intervals (more on this topic is beyond the scope of this article; however, interested readers are referred to Reynolds, 1996; Taylor, 1996; and Baxley, 1995.) Regardless of the particular strategy selected for QC auditing, the key considerations in devising an approach are (i) assurance of event identity and purity of materials used in product testing and the development of regulatory data packages, and (ii) assurance of event identity and

1388

CROP SCIENCE, VOL. 41, SEPTEMBER–OCTOBER 2001

purity of conversions to be increased for commercial product seed volumes. Furthermore, the QC strategy should be tailored to the specific product development pipeline that is to be monitored, taking into account probable trouble spots, critical data collection needs (those involving zero tolerance of mistakes), sampling and analysis efficiencies that might be possible by coupling QC inspections with other analyses, and any other relevant considerations specific to the product in question. Like auditing strategies, assay methods must be focused on achieving objectives in the most cost-effective manner. Assay methods that maximize screening efficiency and minimize the number of data points required can be adopted. The backbone technology for QC in TSPD must be DNA-based analysis rather than proteinbased, because protein-based analyses cannot discriminate between individual events produced using the same vector in transformation or even between events produced using vastly different vectors that incorporate the same gene to confer the trait of interest. Protein-based assays, such as ELISAs or other types of trait metrics may have use in verifying transgene expression, but these analyses must be supplemented by DNA assays to confirm the identity of transgenic event in question. In some instances, it may be acceptable to screen for trait expression within a given population, and then analyze a smaller subset of positive samples for event identity. Conversely, from the standpoint of genetic stability, it is also important to verify that those plants that contain a given transgenic event actually express the desired trait. Failure to do so may be an indication of transgene silencing. DNA assays can be hybridization-type or PCR-type technologies. Hybridization assays, such as Southern blot assays, do not require sequence information, as the vector used in transformation or portions thereof can be used straightaway as a probe in QC auditing. In contrast, PCR assays require sequence information for the design of primers that will amplify regions between them. Generally, sequence information pertinent to the chromosomal regions immediately adjacent to the transgenic insert is used; hence, PCR-based assays can be event-specific. Although a high level of specificity can also be achieved with hybridization assays, PCR-based assays are generally more amenable to high throughput and are less expensive to perform. Nonetheless, development of PCR assays is not typically undertaken until the number of candidate events has been winnowed down to a very low number, because a significant resource investment is needed to generate the sequence information required to fabricate event-specific primers. If the necessary sequence information is unavailable, hybridization assays (mainly Southern blot analyses) are typically the only available means to address the question of event identity. Of course, with the rapid pace of development in DNA technology, this situation will probably change sooner rather than later. As an alternative to inspection of every plant of each conversion line in generations targeted for QC auditing, inspection can be performed with a subset of representative plants. Even if all plants or seeds representing a

given conversion are not assayed, 100% assessment can be effectively achieved if all of the individuals to serve as parents of the next generation are inspected (commonly in corn, only a portion of the total number of available plants per conversion are pollinated and then harvested for planting the next generation.) The specific number of plants to be inspected will depend on a number of factors, including lab turnaround time, the number of plants required to produce seed quantities for the next generation according to the breeding scheme used in the conversion procedure, and the likelihood of a row failure (i.e., the likelihood that seed will not be produced on any plants in that row under typical nursery practices and typical local weather patterns). Bulking of samples may further minimize the number of required data points. The ability to bulk depends on the sensitivity and specificity of the assay that can be attained with various numbers of pooled plant samples per row. If, for example, one defective plant in a bulk of at most ten plants can be detected in an assay, then up to ten plants per row (or generation of a given conversion) can be pooled for inspection. Lab turnaround enabling two consecutive assays to be performed prior to pollination (preferred) or prior to nursery harvest would be ideal in that, should the bulk fail inspection, individual plant samples which comprise the bulk could be assayed to identify appropriate parents (i.e., nondefective plants) to use in producing the next generation. In discussing the benefits of inspection, the impact of inspection errors must be considered. The functional form of pC (shown above) indicates that inspection errors typically push the cost/benefit relationship in the direction of doing no inspection (Vander Wiel and Vardeman, 1994), and must be controlled to maximize the value provided by inspection. Inspection errors can arise due to errors in the collection, handling, and pretreatment of samples, failure of analytical equipment, departures from established laboratory protocol, errors in scoring assay results, and inaccuracy inherent to the assay itself (Houba et al., 1996). Most of the potential for errors arises through field and laboratory practices and systems. Obviously, dependability in these areas and a high degree of coordination between the field and laboratory is crucial. In closing, the advent of transgenic crops has generated QC concerns in seed product development that have not traditionally existed for conventional seed products. Sophisticated QC strategies formulated to monitor the product development process and maintain product quality standards in the manufacturing industry can be used as a foundation for the development of QC strategies for the biotechnology seed industry. Designing an optimized QC monitoring strategy in TSPD requires integration of a host of considerations related to the costs and economic benefits of developing and commercializing nondefective products, the nature of the trait under consideration, analytical tools available, the urgency and pace of the product development stream, characteristics of the crop plant, peculiarities of the transformation system used to generate the transgenic materials, seed and grain purity standards imposed by

MUMM & WALTERS: QUALITY CONTROL IN TRANSGENIC CROP SEED DEVELOPMENT

governmental regulators, and so on. Clearly, there is both room for and a need for more detailed consideration of each of the QC monitoring factors presented here as this industry moves forward. REFERENCES Baxley, R.V., Jr. 1995. An application of variable sampling interval control charts. Journal of Quality Technology 27(4):275–282. Beever, D.E., and C.F. Kemp. 2000. Safety issues associated with the DNA in animal feed derived from genetically modified crops: A review of scientific and regulatory procedures. Nutrition Abstracts and Reviews, Series B: Livestock Feeds and Feeding 70 (3):175–182. Bruening, G. 1998. Plant gene silencing regularized. PNAS 95:13349– 13351. Deming, W.E. 1982. Quality productivity and competitive position. MIT Center for Advanced Engineering Study, Cambridge, MA. Deming, W.E. 1986. Out of the crisis. MIT Center for Advanced Engineering Study, Cambridge, MA. Finnegan, J., and D. McElroy. 1994. Transgene inactivation: Plants fight back! Biotechnology 12:883–888. Holliday, R., and T. Ho. 1998. Evidence for gene silencing by endogenous DNA methylation. PNAS 95:8727–8732. Houba, V.J.G., I. Novozamsky, and J.J. van der Lee. 1996. Quality aspects in laboratories for soil and plant analysis. Communications in Soil Science and Plant Analysis 27(3&4):327–348. Irwig, L., P.P. Glasziou, G. Berry, C. Chock, P. Mock, and J.M. Simp-

1389

son. 1994. Efficient study designs to assess the accuracy of screening tests. American Journal of Epidemiology 140 (8):759–769. Johnson, G.R., and R.H. Mumm. 1996. Marker assisted maize breeding. In Proceedings of the 51st Corn and Sorghum Conference, Chicago, IL. 11–12 Dec. 1996. American Seed Trade Association, Washington, DC. Puchta, H., P. Swoboda, S. Gal, M. Blot, and B. Hohn. 1995. Somatic intrachromosomal homologous recombination events in populations of plant siblings. Plant Molecular Biology 28(2):281–292. Register, J.C., D.J. Peterson, P.J. Bell, W.P. Bullock, B.F. Evans, A.J. Greenland, N.S. Higgs, I. Jepson, S. Jiao, C.J. Lewnau, J.M. Sillick, and H.M. Wilson. 1994. Structure and function of selectable and non–selectable transgenes in maize after introduction by particle bombardment. Plant Molecular Biology 25:951–961. Reynolds, M.R., Jr. 1996. Shewhart and EWMA variable sampling interval control charts with sampling at fixed times. Journal of Quality Technology 28(2):199–210. Taylor, W.A. 1996. Quick switching systems. Journal of Quality Technology 28(4):460–471. Thompson, D. 2000. Certification perspectives related to increased GMO seed and grain production. In Proceedings from the Thirtysixth Annual Corn Breeders’ School, University of Illinois at Urbana-Champaign. 6–7 March 2000. Urbana, IL. p. 19–27. Vander Wiel, S.A., and S.B. Vardeman. 1994. A discussion of all-ornone inspection policies. Technometrics 36(1):102–109. Waterhouse, P.M., M.W. Graham, and M. Wang. 1998. Virus resistance and gene silencing in plants can be induced by simultaneous expression of sense and antisense RNA. PNAS 95:13959–13964.