Inside a Genentech disclosure: a deliberately placed chemical handle, a two-part screening question, and the difference between a promising site and a usable conjugate.
US7521541B2 · Independent PatentOracle research · 20 September 2026
An engineered free cysteine provides a selected attachment site. But a location that looks exposed in a structural model is not automatically reactive enough to use: the disclosure reports that calculated accessibility did not correlate with measured thiol reactivity.
The original Figure 8 pairs a HER2-binding test with a biotin–streptavidin test. Together, they examine retained antigen binding and the availability of the engineered thiol for labeling. Biotin is the screening label in this figure, not the therapeutic payload.
Claim 1 concerns the engineered antibody. Claim 32 adds a drug, linker and further requirements. Keeping those pieces separate makes the invention easier to understand without treating attachment-site evidence as proof of payload release or a clinical outcome.
The central proposal is to alter a parent antibody by replacing one or more amino-acid residues with cysteine. The desired new cysteine is “free”: its sulfur-containing thiol group is not occupied in an intra- or intermolecular disulfide bond. That thiol can become a deliberately located chemical handle for a linker, a diagnostic label, or a drug payload. The stated motivation is that conventional conjugation at the often numerous lysines of an antibody can create a heterogeneous mixture: molecules differ in payload number and in the locations at which payload is attached. The invention instead seeks a selected, experimentally verified attachment site.
A useful engineered site must react. The patent defines a thiol reactivity value as the fraction of engineered free cysteine that reacts with a thiol-reactive reagent, normalized so complete reaction is 1.0 and no reaction is 0. Its stated range for the engineered antibodies is 0.6 to 1.0. This qualification matters because a newly introduced cysteine can oxidize and form an intermolecular dimer or an intramolecular disulfide; in either case it becomes unavailable for conjugation. The mechanism study is therefore not merely cysteine substitution, but structural proposal followed by experimental site selection.
The parent antibody is the starting antigen-binding protein. A cysteine-engineered antibody is derived from it by replacing residue(s) with cysteine. The resulting free cysteine supplies the reactive thiol. A thiol-reactive reagent is the chemical partner used either to measure that thiol or to add a component; the patent gives maleimide-containing reagents as examples. A final antibody–drug conjugate, or ADC, is a separate product: antibody (Ab) joined through linker (L) to drug (D), expressed as Ab-(L-D), with p stated as 1, 2, 3, or 4. Biotin used in a screening assay is an affinity label, not evidence that the attached material is a therapeutic payload.
A thiol is the sulfur-containing functional group on cysteine. The patent describes it as nucleophilic: it can form a covalent bond with an electrophilic group on a linker or drug-linker intermediate. It identifies maleimide and haloacetyl as useful thiol-reactive groups. In its stated comparison, reaction of cysteine thiol with a maleimide group is about 1,000 times higher than reaction by other protein functionalities, including lysine amino groups or the N-terminal amino group. This chemistry is the proposed basis for directing attachment to the engineered site rather than relying on broadly distributed natural residues.
The patent describes an initial model system, hu4D5Fabv8. Five positions were initially selected from crystal-structure information because they were remote from the antigen-binding surface: light-chain Ala43 and heavy-chain Ala40, Ser119, Ala121, and Ser122. Cysteine was introduced by site-directed mutagenesis. The brief drawing description calls Figure 1A a representation of hu4D5Fabv7; the later description calls it hu4D5Fabv8. This naming inconsistency remains unresolved. Both passages describe exemplary engineered heavy- and light-chain cysteine positions using sequential numbering. Its original sheet was not supplied among the selected images, so its spatial labels cannot be independently interpreted here. The described role of that structure is candidate selection, not proof of thiol reactivity.
Upper panel of original Figure 8. The crop preserves the source drawing and labels. Numbered blue marks are PatentOracle annotations.
Inspect original page 11 ↗The upper panel depicts the antibody fragment binding HER2. This part of the assay examines antigen binding after engineering and labeling.
The disclosed assay displays the Fab or ThioFab on phage. The phage is a screening carrier in this drawing.
The text describes detection with an anti-phage HRP antibody. This is the assay signal; it is not a depiction of drug delivery.
Lower panel of original Figure 8. This is the companion labeling test, not a second therapeutic mechanism.
Inspect original page 11 ↗The screening approach reacts cysteine-engineered material with a thiol-reactive affinity reagent. Here, the introduced biotin label is the handle used for the binding test.
The lower panel depicts biotinylated ThioFab binding streptavidin. The source compares this readout with HER2 binding when measuring thiol reactivity.
HRP detection of bound phage supports the paired assay. A useful attachment site must be assessed for labeling and retained binding, rather than inferred from structure alone.
The PHESELECTOR approach uses phage-displayed ThioFab variants. A biotinylation reagent reacts with a usable engineered thiol. Binding to HER2 tests whether the antibody fragment retains target recognition; binding to streptavidin tests whether biotin has been installed. Thiol reactivity is reported as the ratio of OD450 for streptavidin binding to OD450 for HER2 binding. Thus, a candidate is assessed both for a chemically available thiol and for preserved antigen binding rather than on either measurement alone.
The disclosure gives alternatives, not a single combined sequence. In one route, the engineered antibody thiol reacts with a linker reagent to form Ab-L, which then reacts with an activated drug. In the other, drug and linker first form D-L, which then reacts with the antibody cysteine. One reported ThioMab example used about a 50-fold excess of TCEP for 3 hours at 37°C; purification in 10 mM sodium acetate at pH 5; reoxidation with dilute 200 nM aqueous copper sulfate at room temperature overnight; and about a 10-fold excess of BM(PEO)-DM1 for about one hour at room temperature. Those are reported example conditions, not requirements for every embodiment.
The phage-ThioFab table reports high reactivity values for L-V15C (0.934), L-V110C (0.850), H-A88C (0.914), and H-A121C (0.925); these four were selected for further analysis. The patent expressly reports that calculated fractional surface accessibility did not correlate with measured thiol reactivity. In particular, partially exposed alanine or valine sites with 20% to 80% accessibility could perform better than cysteines installed at serine sites. The supported finding is practical: structure gave a rational starting set, but the assay was needed to identify reactive and stable attachment sites.
The supplied sheet is Figure 8, not Figure 1A. Its upper panel shows phage-associated antibody binding to immobilized HER2, with HRP shown as the reporting label. Its lower panel shows a biotin-bearing form binding immobilized streptavidin, again with HRP. Read upper then lower: first assess retained HER2 binding; then assess whether the thiol-reactive biotin-labeling step created streptavidin binding. The figure depicts the screening logic, not drug release, intracellular delivery, or a clinical result.
Claim 1 requires a cysteine-engineered antibody with a free cysteine amino acid having a thiol reactivity value of 0.6 to 1.0, plus a heavy-chain sequence selected from SEQ ID NOS: 11, 12, 13, and 15, where the indicated cysteine is the free cysteine. Claim 2 adds greater reactivity than the parent antibody. Claim 3 adds a process that determines thiol reactivity by reaction with a thiol-reactive reagent, also requiring greater reactivity than the parent. Claims 4 and 5 separately narrow the range to 0.7–1.0 and 0.8–1.0. Other dependent claims add preparation, phage or phagemid display, biotin/maleimide and streptavidin features, or antibody forms; they are not all cumulative requirements.
Claim 32 requires an ADC containing the specified cysteine-engineered antibody; a drug moiety selected from maytansinoid, auristatin, dolastatin, or calicheamicin; and a linker joining the drug to one or more free engineered cysteines. It specifies p as 1, 2, 3, or 4, requires engineering by replacement of one or more parent-antibody residues with free cysteine, and requires the parent and engineered antibody to selectively bind the same antigen. Claim 33 further limits that ADC through the replace-and-test process and greater reactivity than the parent. Subsequent claims separately narrow p, linker architecture, linker identity, payload identity, and antibody type.
The stated reason is control over conjugate composition. Attachment at many lysines can make a mixture varying in both payload count and attachment location. An engineered free cysteine is intended to create a selected chemical handle. But location cannot be assumed suitable simply because it is exposed: the new cysteine may oxidize, fail to react, or interfere with antibody structure or binding. The disclosed design therefore combines a structural hypothesis with direct reactivity and binding tests.
The selected evidence establishes a proposed site-specific attachment strategy, a comparative assay for thiol reactivity and retained binding, reported reactive variants, and examples of linker–drug construction. It does not establish one universal payload-release mechanism. Linker-dependent release is a distinct issue from choosing an engineered attachment site; dependent claims include particular linker features, but those should not be treated as a general release result. The evidence also does not establish clinical safety, efficacy, approval, or current implementation. A useful reading route is Figure 8 for the two-part screen, the Table 2 discussion for the accessibility-versus-reactivity result, then independent claims 1 and 32 for the distinct engineered-antibody and ADC requirements.
One historical patent, read through a bounded selection of 23 description and claim pages. The PDF includes the full explanation and all 105 original pages. This focused edition explains attachment-site selection and conjugation; it does not exhaust every sequence, embodiment or claim.
The PDF contains the explanation, annotations and every original page, with internal source links. The ZIP includes this web edition, the PDF, source and review records, and checksums. Downloaded files remain readable offline.
The automated source review is preserved. One naming inconsistency received a recorded Codex source correction after inspection of the original pages. The original figure crops and new annotations received separate Codex inspection. This is assisted source review, not independent human review.
The brief drawing description on original page 31 names hu4D5Fabv7, while the later discussion on page 47 names hu4D5Fabv8 for Figure 1A. The reading preserves that discrepancy instead of choosing a version. Figure 8 below is the independently reviewed screening schematic.
The source record separates reported experimental conditions, claim wording and interpretation. No clinical safety, efficacy or current product claim follows from this example.
The original patent is the source of the proposed capabilities. This independent example does not establish deployment, measured performance, a client relationship or endorsement by Genentech.
This curated example includes additional editorial work and annotations on the original drawings. Comparable bespoke work receives its own scope and quote.
Standard automated explanations and comparisons are separate services. The reading plan and price are confirmed before paid work begins.
Tell us which patents and questions matter to you. We’ll agree the annotation, review and delivery scope before quoting a bespoke edition.
Discuss a similar annotated report ↗