US 5641670

I'll search for authoritative information on this patent number.

US Patent 5,641,670 — Analyst Summary

⚠️ Sourcing / confidence note first

I searched for this patent number specifically and took care not to substitute similar numbers (e.g., 5,641,609, 5,641,607, 5,641,515 appeared in numeric-neighbor results and are unrelated — one is a Brazilian state payroll notice). Two things I want to flag up front:

1. I could not retrieve verbatim claim text for this patent from the sources available to me. My claim overview below is derived from the patent's own abstract and specification ("Summary/Definitions" matter on the Google Patents record), not from a verbatim reproduction of the claims. Treat claim wording as paraphrase, not quotation.
2. I found no 2026 CAFC or PTAB docket activity for 5,641,670. See the separate section below.

---

1. Bibliographic data

Family members cited in the record (relevant to priority/continuation strategy): US 5,733,746; US 5,733,761; WO 93/09222; WO 94/12650; WO 95/31560; EP 0649464; EP 0759082. A later continuation, US 2003/0082675 A1, claims priority to a 1999 application.

One discrepancy worth noting: a secondary Japanese-language reference (JP 2002504361 A) cites this patent as "issued June 24, 1996." The authoritative US record (and the PCT search-report family annexes) give June 24, 1997. I am reading the identifier literally and reporting the discrepancy rather than silently correcting the secondary source.

---

2. Abstract (verbatim, per the USPTO/Google Patents record)

"The invention relates to constructs comprising: a) a targeting sequence; b) a regulatory sequence; c) an exon; and d) an unpaired splice-donor site. The invention further relates to a method of producing protein in vitro or in vivo comprising the homologous recombination of a construct as described above within a cell. The homologously recombinant cell is then maintained under conditions which will permit transcription and translation, resulting in protein expression. The present invention further relates to homologously recombinant cells, including primary, secondary, or immortalized vertebrate cells, methods of making the cells, methods of homologous recombination to produce fusion genes, methods of altering gene expression in the cells, and methods of making a protein in a cell employing the constructs of the invention."

---

3. Plain-language overview of the independent claims

This patent is the foundational "gene activation" / "transkaryotic therapy" patent — the mechanism was later described in the literature as introducing a promoter or other exogenous regulatory sequence into a patient's chromosomal DNA ex vivo to switch on a dormant endogenous gene, then reintroducing the activated cells. (Sources: US 20090258826 A1 and JP 2007502132 A, both of which cite US 5,641,670 and 5,733,761 for exactly this proposition.)

The independent claims fall into roughly four groups, described here in plain language as they are characterized in the patent's own Summary/Definitions sections:

(A) The DNA construct claim(s). A DNA construct (targeting vector) made of four elements:

• (a) a targeting sequence — DNA homologous to a chosen site in the cell's chromosomal DNA, so it will recombine there;

• (b) a regulatory sequence — e.g., a promoter/enhancer (the patent's examples use the mouse metallothionein-I promoter);

• (c) an exon — DNA that will be copied into the mature mRNA (may or may not encode amino acids); and

• (d) an unpaired splice-donor site — a splice signal with no matching acceptor in the construct, so splicing is forced to an acceptor in the endogenous gene.

  In plain terms: a vector that carries a "on switch" plus a splicing hook, designed to land next to a gene you want to turn on.

(B) A variant construct claim with a complete splice cassette. Same idea, but with six elements: targeting sequence, regulatory sequence, exon, splice-donor site, an intron, and a splice-acceptor site. This version brings its own intron rather than relying on acceptor sites already present in the target gene.

(C) Method claims — producing a protein. A method in which the construct above is introduced into a cell; homologous recombination (gene targeting) at the preselected site creates a new (chimeric) transcription unit in which the introduced regulatory sequence, exon, and splice-donor site are operatively linked to the endogenous gene; the cell is then maintained under conditions permitting transcription, splicing, and translation, and the protein is produced — in vitro (cell culture, for manufacturing) or in vivo (gene therapy). The patent expressly contemplates activation of a normally silent gene, increased expression of a gene already expressed, altered regulation/induction, or reduced/eliminated expression.

(D) Cell claims — homologously recombinant (HR) cells. A cell (primary, secondary, or immortalized; vertebrate, typically mammalian, ideally human) whose genome contains the introduced regulatory sequence + exon + flanking splice-donor site operatively linked to an endogenous gene (e.g., linked to the endogenous gene's second exon). The specification also describes an amplification embodiment: including an amplifiable selectable marker (e.g., dhfr, ada, CAD) so the activated endogenous gene can be co-amplified under drug selection (the EPO/dhfr examples report ~70-fold increases in hEPO output).

Practical significance claimed: because the construct supplies its own promoter and splice-donor, the inventors argue you can position the "on switch" anywhere from immediately adjacent to ~30 kb upstream of a gene, or inside an intron — without the constraint of avoiding stray ATG codons in a long leader sequence. This makes it possible to activate very large genes, or genes whose full coding region had never been cloned.

---

4. Search of USPTO / CAFC 2026 dockets — result: no hits

I ran targeted searches for this patent number against Federal Circuit docket material and PTAB/UPC listings. I found no 2026 CAFC appeal, no 2026 PTAB proceeding, and no 2026 UPC proceeding referencing US 5,641,670.

• This is consistent with the record showing expiration on June 24, 2014 (anticipated expiration), which would make active 2026 US infringement litigation very unlikely.
• The Google Patents record carries a "Family has litigation" flag pointing to a third-party (Darts-IP) litigation dataset, indicating historical worldwide litigation in this family — but I cannot confirm any specific case identity, party, or 2026 status from that flag alone.
• Do not confuse this with a case that did surface in my search: Genzyme Corp. & Mount Sinai School of Medicine v. Transkaryotic Therapies, Inc., 346 F.3d 1094 (Fed. Cir. 2003) — that appeal concerned the '804 patent and the claim term "chromosomally integrated," not US 5,641,670. It involves the same assignee but a different patent.
• I also could not independently verify the well-known historical EPO-related litigation involving the '670 family. I am not asserting it; I simply don't have authoritative confirmation in front of me.

Uncertainty statement: If you need a negative-clearance or litigation-history confirmation, it should be run against a live, authoritative litigation database (Darts-IP, Docket Navigator, PTAB API, CAFC docket) rather than inferred from a Google Patents status flag. My searches are indicative, not exhaustive.

---

5. Quick reference — what you should verify against the primary source

I'll research litigation involving US patent 5641670.

Let me search more specifically for litigation databases and known cases.

Let me search litigation-specific databases directly for this patent number.

Summary of findings

I could not identify any litigation specifically asserting US Patent No. 5,641,670.

I want to be explicit about this rather than give you a case list that would be wrong. Here is what my searches did and did not surface.

What the sources show for 5,641,670 itself

• The Google Patents record for US5641670A ("Protein production and protein delivery"; inventors Treco, Heartlein, Selden; original assignee Transkaryotic Therapies, Inc.) carries a "Family has litigation" flag with a Darts-IP link. That link points to patent family 22918596 — i.e., it is a family-level flag. Google Patents does not identify which family member(s) were litigated, so this alone does not establish that 5,641,670 itself was asserted or challenged in any proceeding.
• The record also shows the patent's legal status as Expired – Lifetime (anticipated expiration June 24, 2014).
• All other hits for the literal string "5641670" were citations of 5,641,670 as prior art in other patents' search reports and family annexes (e.g., WO2001068882A3, EP3883606B9, various CN/other family pages). Those are prior-art citations, not litigation.

Important caution against a likely false positive

My searches heavily surfaced the well-known EPO litigation — Amgen Inc. v. Hoechst Marion Roussel, Inc. (now Aventis Pharmaceuticals, Inc.) and Transkaryotic Therapies, Inc., D. Mass., Fed. Cir. Nos. 314 F.3d 1313 (2003) and 469 F.3d 1039 (2006). That case did NOT involve US 5,641,670. The patents in suit were Amgen's 5,547,933; 5,618,698; 5,621,080; 5,756,349; and 5,955,422 (Lin EPO patents). Because these are "similar numbers" to 5,641,670, I am flagging them and excluding them from the answer to your specific question, per your instruction.

(The relationship is contextual only: TKT — assignee of 5,641,670 — was the defendant in Amgen's EPO suit, and TKT's gene-activation technology, of which 5,641,670 is an example, is what was accused. But 5,641,670 was not a patent asserted in that case.)

Litigation found for US 5,641,670

No case asserting, invalidating, or otherwise adjudicating US 5,641,670 was identified in my searches.

Caveats and recommended verification

1. My search sequence was truncated before I could query the litigation databases directly by patent number (Unified Patents litigation portal, CourtListener/RECAP, Docket Navigator, Darts-IP family 22918596, PACER). The family-level "has litigation" flag on Google Patents should be resolved at the source.
2. Because the patent expired in 2014 and its asserted family members date to the 1990s, any litigation, if it exists, would most likely be historical (1990s–2000s) rather than a currently pending matter.
3. The most reliable next step is to pull Darts-IP family 22918596 (per the Google Patents link) and run a patent-number search on the Unified Patents litigation portal and CourtListener filtered to exactly 5,641,670 — which I was not able to complete here.

I cannot state with high confidence that no such litigation ever existed; I can state that none appeared in the searches I was able to run, and that the only litigation-related signal attached to this patent is a family-level flag whose member(s) it does not identify.

I'll start by checking the structured PTAB data provided and then run targeted searches for any AIA trial proceedings involving this patent.

Let me verify whether the "670 patent" hit relates to US 5,641,670 or a different patent, and check this patent's litigation/reexamination history.

One more confirmation search to be sure no IPR/PGR/CBM exists on this patent and to check whether it was ever asserted in litigation.

I've reached my search-tool limit. Here is the analysis based on the structured ODP block and the searches completed.

---

Proceedings overview

Zero AIA trial proceedings on file. The USPTO Open Data Portal returns no Inter Partes Review, Post-Grant Review, or Covered Business Method review for US 5,641,670, and my independent searches surfaced none — so the count is 0 active / 0 claims invalidated / 0 claims sustained / 0 settled / 0 institution denials. The bottom-line defensive posture for a defendant is therefore not "the patent has been hardened by IPR" and not "the claims are canceled," but something arguably better: the patent expired on 2014-05-13 (Google Patents lists anticipated expiration 2014-06-24), it was never subjected to a post-grant validity challenge, and there is a full, unestoppel-ed universe of § 102/§ 103 art available to you in District Court. Anybody asserting this patent today is asserting an expired patent, which caps the exposure to past damages inside the § 286 six-year lookback and forecloses injunctive relief.

⚠️ Naming trap — do not confuse this patent with US 9,990,670. My searches repeatedly surfaced PTAB papers (including IPR2024-00012, Home Depot U.S.A., Inc., filed 2023-11-27) referring to "the '670 patent." That shorthand refers to US 9,990,670, Innovaport LLC's in-store product-location patent asserted against IKEA, Home Depot, and Walmart. It is a different patent, different owner, different technology (retail software, § 101 litigation before Judge Conley in W.D. Wis.). It has nothing to do with US 5,641,670. Any docket-skimming that conflates the two will produce a completely wrong estoppel analysis.

Proceedings

None. There is no PTAB proceeding to report at the proceeding level — no petitioner, no panel, no Institution Decision, no Final Written Decision, no settlement, and no CAFC appeal arising from a PTAB case. I am not going to invent a proceeding number to fill the template.

What the surrounding record does show (relevant context, not PTAB activity)

• Patent status: US 5,641,670 ("Protein production and protein delivery"), inventors Treco, Heartlein & Selden, original assignee Transkaryotic Therapies, Inc. (Cambridge, MA); current assignee Shire Human Genetic Therapies Inc. Application 08/243,391 filed 1994-05-13; granted 1997-06-24; earliest priority 1991-11-05. Now expired — no live claim scope remains. Source: https://patents.google.com/patent/US5641670/en
• Litigation of record: Google Patents flags "Family has litigation." The litigation I can confirm from public reporting is the Amgen v. Hoechst Marion Roussel / Transkaryotic Therapies EPO dispute. Critically, that case was Amgen asserting its own patents (US 5,547,933, 5,618,698, 5,621,080, 5,756,349, 5,955,422) against TKT's Dynepo product — it was not an assertion of TKT's '670 patent. See Amgen Inc. v. Hoechst Marion Roussel, Inc., 314 F.3d 1313 (Fed. Cir. 2003), https://www.courtlistener.com/opinion/[780346](/patent/780346)/amgen-inc-plaintiff-cross-v-hoechst-marion-roussel-inc-now-known-as/ and the 2006 remand/vacatur, reported at 2006-08-03. So the "litigation" flag on this patent family reflects TKT being a defendant, not this patent being enforced.
• Reexamination: I found no ex parte or inter partes reexamination certificate for 5,641,670 in the surfaced material. Note this is a negative finding from limited searching, not a confirmed absence — the CRU file is not visible through the ODP block provided.

Strategic summary

Claim status. Because no AIA trial ever reached a Final Written Decision, there are no claims canceled and no claims held patentable by the Board. Every claim of 5,641,670 is "untested" at the PTAB — but that framing flatters the patent, because the patent expired in 2014. Practically: the claims are unadjudicated in the post-grant forum, and simultaneously unenforceable prospectively due to expiry. If a demand letter cites any claim of this patent, the correct counter is not "the claim is dead from IPR" — it is "the patent is expired, so you get nothing but possibly time-barred back damages under § 286, and you must prove infringement of a claim you've never had construed anywhere."

Estoppel landscape — wide open. With no IPR/PGR/CBM, 35 U.S.C. § 315(e)(2) estoppel never attached to anyone. No petitioner, no privy, no real party in interest is barred from raising any prior-art ground, and there is no IPR record for a patent owner to use as a roadmap or a District Court to use as a shortcut. Your invalidity case can be run in the district court (and via a de novo ex parte reexamination request at the CRU, subject to the § 325(d) "same or substantially the same art" discretion) on whatever art you choose, without the estoppel shadow that normally hangs over defendants in the TKT/Shire biotech families.

Pattern signals. No serial petitions, no repeat petitioner, no defensive aggregator (Unified Patents or similar) in the chain, and no patent-owner PTAB appeal practice evident for this patent. The absence is itself a signal: US 5,641,670 is an expired 1991-priority Gene Activation patent that was never worth a post-grant attack. Well-asserted patents of this vintage in the Amgen/TKT EPO wars did attract decades of litigation — but always on Amgen's patents, never this one.

Recommended next steps

1. Treat any assertion of US 5,641,670 as a stale-patent play. Confirm the expiration date on the face of the patent and via USPTO Patent Center (https://patentcenter.uspto.gov) and Google Patents (https://patents.google.com/patent/US5641670/en). An expired patent cannot support prospective injunctive relief, and damages are limited to infringing acts within six years before suit (35 U.S.C. § 286).
2. Do not search PTAB E2E for "670" generically. Go to https://ptacts.uspto.gov/ptabweb/ and query the exact patent number 5,641,670 (or the application number 08/243,391) so you don't collide with the Innovaport US 9,990,670 proceedings (IPR2024-00012 et al.).
3. No estoppel to work around. Because there is no FWD, the full prior-art universe remains available in District Court; there is also no § 315(e)(1) bar on a new IPR/PGR — though filing one against an expired patent is of doubtful practical value and would run into discretionary-denial practice (Fintiv/§ 325(d)) in the current regime.
4. Confirm the reexamination and litigation record through USPTO Patent Center's "Reexaminations" tab and a PACER/CourtListener search on the '670 patent number before relying on the negative finding above — my tools returned no reexam certificate, but I could not exhaustively verify.

Bottom line for a defendant: no PTAB activity exists on US 5,641,670. The patent survived no IPRs because it never faced one — and it never faced one because it expired in 2014. Defensive posture is strong, but the strength comes from expiry and the clean estoppel slate, not from a Board cancellation you can quote. If opposing counsel hands you an FWD for "the '670 patent," check the number: it is almost certainly Innovaport's US 9,990,670, not this one.

I'll research the ownership chain for US 5641670 using the sources you specified.

I'll research US patent 5641670 and its prior-art citations.

Let me get the complete citation list from the patent's front page.

Let me retrieve the full reference list and the patent's claims.

Bottom line up front

I was able to identify US 5,641,670 and assemble its front‑page "References Cited" list, but I must flag two limits immediately:

1. I did not obtain the citation list by querying USPTO Patent Public Search / PatFT / PatentCenter directly. The tool calls I could run returned Google Patents, PubChem's patent record, EveryPatent, and Justia. The reference list below is transcribed from those secondary renderings of the printed patent's front page, and it should be re‑verified against the official USPTO "References Cited" before being relied on in a legal filing.
2. The exact claim text and claim numbering of US 5,641,670 was not retrieved directly. I verified the substance of the claims from the authoritative full text you supplied (abstract + specification) and from the closely related continuation US 6,270,989 (whose claims are visible in search results). Where I map references to claim numbers, treat the numbers as the claim family/subject matter rather than a verified verbatim claim set.

With those caveats, here is the analysis.

---

1. The patent under review

Family context (matters for the prior‑art date analysis): the international counterpart is PCT/US1995/006045 → WO 95/31560 A1, published 1995‑11‑23, which lists US 5,641,670 in its family (source: the WO 1999005268A1 international search report family table retrieved in search). Sibling/continuation members include US 5,733,761, US 5,733,746, US 6,270,989, US 6,240,218, and US 6,200,778. Note that several of these were filed in 1995–1996 and therefore have reference lists that mix the 1991 priority art with later art — a point that explains an anomaly in the list below.

Critical date issue: the listed priority date is 1991‑11‑05, but the application was filed 1994‑05‑13. Several references on the front page are dated 1992–1995 (e.g., Barinaga 1994, Crystal 1995, Marshall 1995). Those post‑date the listed priority date and therefore cannot be § 102 prior art against these claims unless the priority claim is invalidated or the effective filing date is shown to be 1994‑05‑13. Their presence is best explained by an IDS or by reference‑list carry‑over from the continuation families; it should be confirmed and, if they truly pre‑date nothing, treated as non‑prior art.

---

2. What the claims require (the anticipation yardstick)

From the authoritative full text you supplied, the construct claim family requires:

• (a) a targeting sequence (homologous to a preselected genomic site);
• (b) a regulatory sequence (e.g., mMT‑I or CMV promoter);
• (c) an exon; and
• (d) an unpaired splice‑donor site — expressly defined in the specification as "a splice‑donor site which is present in a targeting construct and is not accompanied in the construct by a splice‑acceptor site positioned 3′ to the unpaired splice‑donor site," such that it "results in splicing to an endogenous splice‑acceptor site."

The specification also (col. text you supplied) recites alternative construct formats adding an intron + splice‑acceptor site, and the method claims require transfection → homologous recombination → transcription under the exogenous regulatory sequence → splicing to an endogenous acceptor → translation.

Anticipation therefore requires a single reference disclosing all four elements (a)–(d) in one construct, or the full method steps. That is the correct yardstick for the § 102 analysis below.

---

3. References cited on the face of US 5,641,670

Table A — U.S. patent documents

Table B — Foreign patent documents

Note: WO 93/09222 and WO 94/12650 — both frequently described as "activate an endogenous gene / amplifiable gene by homologous recombination" art — appear in the background discussions I retrieved but I could not confirm them on the front‑page reference list of 5,641,670; WO 94/12650 post‑dates the listed 1991 priority in any event. Do not attribute them to this patent's citation list without checking the printed patent.

Table C — Non‑patent literature cited (transcribed from PubChem's record of US‑5,641,670‑A and EveryPatent's rendering)

---

4. § 102 anticipation — overall assessment

My assessment: no reference on the cited list appears to be a single‑reference anticipation of the principal construct claim or the principal method claim. Specifically:

• The four‑element construct (targeting sequence + regulatory sequence + exon + unpaired splice‑donor site) — I found no cited reference disclosing all four in one construct. US 5,272,071 (A3), WO 90/11354 (B10), WO 91/06666/06667 (B14/B15), WO 92/19255 (B21) and WO 92/20808 (B20) each teach subsets (targeting sequence + regulatory element; or targeting + amplifiable marker; or in‑situ modification), and are best characterized as § 103 combinations rather than § 102 art. The specific, definitional limitation — an unpaired donor, i.e., one deliberately lacking a downstream acceptor within the construct so that splicing must run to an endogenous acceptor — is the element the cited art does not show.
• The amplification‑related dependent claims are the most exposed. B14/B15 (WO 91/06666, WO 91/06667), B21 (WO 92/19255) and A/CD12 (Kellems; Schimke) directly teach targeting an amplifiable marker adjacent to an endogenous gene and co‑amplifying. If any granted claim is directed to a construct comprising a targeting sequence, an amplifiable marker gene and a regulatory sequence, without more, those references warrant a close § 102 look (and are strong § 103 art).
• The negative‑selection dependent claims are squarely met by the background art (Mansour C5; Lupton C14) — but as art teaching the selection technique, not the targeting construct.
• The references dated 1992–1995 (C19–C23) should be removed from any § 102/§ 103 chart absent a determination that the effective filing date is 1994‑05‑13. If the 1991‑11‑05 priority holds, they are not prior art at all.

---

5. What I could not verify — do not treat as established

Per your instruction to interpret identifiers literally and not auto‑correct:

1. The identity of "0236059" (Sep. 1987) — resolved literally as printed only; likely a truncated EP number, unverified.
2. The titles of US 4,656,134 and US 4,822,736 — I did not verify these and have deliberately not asserted titles.
3. The assignee and full title of US 5,272,071 — Google Patents lists the inventor as "Chappel"; I did not independently verify the title or assignee.
4. The complete and exact claim set of US 5,641,670, including whether pREPO18 or any other named plasmid is recited in a claim. A third‑party review (TKT's plans for turning on endogenous genes, Expert Opin. Ther. Patents 8(3):325, via Tandfonline) states that pREPO18 is "the only named construct cited as a claim" — I could not verify this against the granted claims and flag it as unverified.
5. The official USPTO reference list — this analysis rests on secondary transcriptions. If you need a filing‑grade result, pull the "References Cited" from USPTO Patent Public Search / PatentCenter for 5,641,670 and diff it against Tables A–C above; discrepancies (especially in Table A and the ambiguous B6 entry) should be resolved in favor of the official record.

Also worth noting for completeness: this family carries a litigation flag on Google Patents (Darts‑IP family link), and the trade press items captured in my searches reference Amgen v. HMR/TKT EPO patent litigation around 1997. That litigation concerns the EPO patent family generally; I did not verify that US 5,641,670 itself was asserted, so treat that as contextual, not established.

I'll research the prior art cited in this patent and the key references in the field to build the § 103 analysis.

Let me research the specific prior-art patents cited in the specification.

Obviousness Analysis — US 5,641,670 under 35 U.S.C. § 103

---

0. Sourcing, scope, and caveats (read first)

0.1 — What the "Prior Art section" of this page actually contains. The authoritative Google Patents text you supplied does not render a discrete "Prior Art / Citations" table. The only prior-art field present is the five-term keyword string: cells, gene, dna, cell, targeting. The prior-art references that can be grounded in the supplied text are therefore:

Everything below is keyed to that universe. I have supplemented it with secondary-source verification of references 1–4 and 5 (cited with URLs). I did not retrieve verbatim claim text for this patent in the earlier pass and I have not done so now — so the limitations analyzed below are the ones recited in the patent's own Abstract and Summary, which I treat as substantially mirroring the independent claims but which remain paraphrase, not quotation. If verbatim claims differ materially, this analysis must be re-run.

0.2 — Legal framework. The application (08/243,391) was filed May 13, 1994 with a listed priority date of Nov 5, 1991. It is therefore governed by pre-AIA § 103(a). The analysis below applies the Graham v. John Deere factors and the KSR Int'l v. Teleflex framework (which governs any present-day validity challenge). The critical date for prior-art qualification is November 5, 1991 (with the actual U.S. filing date of May 13, 1994 controlling the § 102(b) one-year grace period).

0.3 — One thing to exclude. WO 93/09222 (PCT/US92/09627, "Selden et al.") is a family member of this patent — same inventors, same priority. It is cited in later documents alongside the true prior art, but it is not prior art to the '670 patent and must not be used in this analysis. (Confirmed by the citation cluster in US 2005/0129669 and US 6,910,193, which list it next to Chappel, Sherwin and Skoultchi — the listing makes it look like art, but the inventorship identity disqualifies it.)

---

1. The claims in issue and the limitations that matter

Grouped as the summary sets out:

• Claim group A (construct): targeting sequence + regulatory sequence + exon + unpaired splice-donor site, arranged so that the targeting sequence directs integration such that the regulatory sequence, exon and splice-donor are operatively linked to the endogenous gene.
• Claim group B (construct, cassette variant): same, plus intron + splice-acceptor site.
• Claim group C (method): introduce construct → homologous recombination at preselected site → maintain under conditions permitting transcription/splicing/translation → produce protein (in vitro or in vivo).
• Claim group D (cell): a homologously recombinant primary, secondary or immortalized vertebrate cell bearing the group-A elements operatively linked to an endogenous gene.

The three limitations that carry the entire § 103 argument are:

1. The mechanism is regulatory-sequence insertion to drive an endogenous coding region — nothing more exotic than promoter/enhancer knock-in.
2. The exogenous exon supplies the 5′ end of the transcript and, optionally, the initiation codon and signal peptide.
3. The splice-donor site is unpaired — the construct deliberately contains no matching acceptor, forcing the splicing machinery to use a splice-acceptor site resident in the target gene.

The patentee's own statement of the problem is in the specification and is an admission usable against interest: "ATG start codons are found frequently within mammalian DNA (approximately one occurrence per 48 base pairs)… transcription cannot simply initiate at any position upstream of a gene and produce a transcript containing a long leader sequence preceding the correct ATG start codon, since the frequent occurrence of ATG codons in such a leader sequence will prevent translation of the correct gene product and render the message useless."

That paragraph is the whole invention-in-one-line: prior-art promoter knock-in worked only if you landed the promoter in the narrow window where no upstream ATG intervened. The claimed fix is to buy your own 5′ exon plus a splice donor, so the leader is defined by splicing rather than by genomic distance.

---

2. Level of ordinary skill in the art (PHOSITA), as of Nov 5, 1991

A Ph.D.-level molecular biologist or an M.D./Ph.D. with 2–3 years of post-doctoral mammalian cell-culture and recombinant-DNA experience. By late 1991 that person:

• knew homologous recombination in mammalian cells was real and reproducible (Smithies et al. 1985, Nature 317:230-234 — β-globin locus; Thomas & Capecchi 1987, Cell 51:503-512; Doetschman et al. 1987);
• knew how to enrich targeted clones using positive/negative selection (Mansour/Thomas/Capecchi 1988);
• had a review-level command of the whole field (Capecchi 1989, Science 244:1288-1292, cited repeatedly in the patent itself);
• knew the splice-site consensus sequences (Jackson 1991, NAR 19:3715-3798, cited in the patent as the definitional source for both donor and acceptor);
• routinely used promoterless and enhancer-trap reporter cassettes in ES cells, and amplifiable markers (dhfr/MTX) for gene amplification in CHO cells (this was 1980s-era technology).

That is a highly capable PHOSITA operating in a field where the basic tools were already in place. That matters to KSR: the more predictable the art, the broader the permissible combination.

---

3. The prior art, element by element

3.1 WO 90/11354 — Institut Pasteur (priority FR 8903630, 20 Mar 1989; PCT/FR90/00185)

Source: https://patentimages.storage.googleapis.com/b4/6f/f0/c21127b5318dbb/WO1990011354A1.pdf

This is the single most dangerous reference, and the analysis should not be written without it. Its abstract states the invention is "the targeted insertion of a foreign DNA in a site selected in the genome of eucaryote cell… The DNA to be inserted may comprise either a coding sequence, or a regulating sequence. The flanking sequences are chosen so as to allow via homologous recombination, according to the case, either the expression of the coding sequence of the DNA to be inserted under the control of regulating sequences of the receiver gene, or the expression of a coding sequence of the receiver gene under the control of the regulating sequence of the DNA to be inserted."

Claim 2 as published (French text on the PDF) expressly recites that the inserted DNA contains "soit une séquence codante soit une séquence régulatrice" — i.e., a regulatory sequence — flanked by two homology arms.

Read plainly, WO 90/11354 discloses claim-group A's core mechanism and claim-group C's core step: pick a gene, flank a regulatory sequence with homology arms, transfect, recombine, and get expression of the endogenous coding region under the new regulatory element.

What WO 90/11354 does not disclose: any exon, any splice-donor site, the "unpaired" configuration, a selectable marker, amplification, or primary/secondary (non-immortalized) human cells.

3.2 WO 91/06667 / EP 0 452 484 — Cell Genesys (Skoultchi), PCT/US90/06436, priority 6 Nov 1989, EP published 23 Oct 1991

Source: https://patent.public.lu/fo-eregister-view/search/details/591900640_EPV/ (EP 0 452 484, "PRODUCTION OF PROTEINS USING HOMOLOGOUS RECOMBINATION," applicant Cell Genesys Inc., inventor Skoultchi, A. I.)

Published 13 days before the '670 priority date and more than two years before the '670 U.S. filing date. Qualifies as a § 102(b) printed publication. Same genus as Pasteur: use homologous recombination to put a regulatory element at a chosen locus so the endogenous gene is expressed. The patent itself cites it in the Background as art.

3.3 U.S. Pat. No. 5,272,071 (Chappel)

Cited in the '670 Background as art for "inserting a regulatory sequence upstream of the coding region of an endogenous gene." A family member (US 6,270,989, at https://patentimages.storage.googleapis.com/f8/f0/7e/4b3590ca0b1f61/US6270989.pdf) characterizes the four references collectively as follows:

"U.S. Pat. No. 5,272,071 and foreign patent applications WO 91,06666, WO 91,06667 and WO 90,11354 describe homologous recombination methods for inserting a regulatory sequence upstream of the coding region of an endogenous gene. In these methods, only a very small number of positions for promoter insertion are acceptable for expression, limited by the frequent occurrence of ATG start codons…"

⚠️ Flag: I could not verify Chappel's own filing/issue dates or verbatim claims in this session, and I will not assert them. Whether it is § 102(a)/(b) art or § 102(e) art only, it is combinable either way.

3.4 WO 91/06666

Cited in the Background but not identified to an applicant or inventor in any source I retrieved. Flagged as unverified. Treat its contribution as cumulative of 3.1–3.3 unless independently pulled.

3.5 Mansour, Thomas & Capecchi, Nature 336:348-352 (1988)

Sources: https://pubmed.ncbi.nlm.nih.gov/[3194019](/patent/3194019)/ ; full text at https://collections.lib.utah.edu/ark:/87278/s60p1hnh

Discloses, with full structural detail: (i) a replacement-type targeting vector bearing 10–15 kb of homology to the target gene; (ii) a neo^r^ marker inserted into an exon; (iii) a linked HSV-tk gene positioned so homologous (but not random) integration discards it; (iv) positive/negative selection enriching ~2,000-fold. Also expressly notes that "the procedure should be applicable to any gene" and works even for genes with "less than one copy per cell" of transcript — i.e., essentially silent genes.

⚠️ Note the nuance: Mansour's promoterless-marker logic and its "land inside an expressed gene" screening are conceptually adjacent to, but do not disclose, the '670's splice-donor architecture.

3.6 Jackson, NAR 19:3715-3798 (1991)

The patent's own cited authority for the donor consensus (A/C)AG GURAGU and acceptor consensus YYYYYYYYYNYAG. This reference is what makes the predictability argument work: it establishes that a person of skill could predict, from sequence alone, where a given donor would splice and that splicing follows rules.

⚠️ Date flag: a 1991 journal volume must be checked against the Nov 5, 1991 critical date. If the relevant issue published after Nov 5, 1991, it drops out of the § 102(a) window and can only be used as evidence of the state of the art (or of what a PHOSITA knew), not as § 102(b) art. The splicing rules were, in any event, textbook knowledge well before 1991 (Breathnach & Chambon 1981), so the argument survives without Jackson.

3.7 The amplification art

dhfr/MTX-mediated gene amplification was standard by the early 1980s (Alt et al. 1978; Kaufman & Sharp 1982). The specification's own list of amplifiable markers (dhfr, ada, CAD) and selective agents (methotrexate, alanosine, PALA) confirms these were off-the-shelf.

---

4. The obviousness case — specific combinations

★ Combination 1 (primary): WO 90/11354 + Mansour 1988 + Jackson 1991 (+ Hamer & Walling 1982 for the mMT promoter)

This is the combination that most cleanly maps onto independent claim groups A and C.

Motivation to combine (why a PHOSITA would have done this):

1. Same field, same problem, same solution direction. WO 90/11354 and WO 91/06667 are directly on point: put a regulatory element into a genome by homologous recombination so an endogenous gene is expressed. A PHOSITA working on protein production or somatic gene therapy in 1991 would start there.
2. The known deficiency of the simple insertion approach was documented in the art and conceded in the patent itself. The '670 specification states the ATG-density problem (1 ATG per ~48 bp) that makes long 5′ leaders untranslatable. A PHOSITA confronting that problem has a motivation to modify the construct.
3. The fix is the routine tool of the trade. Decoupling a coding region from a genomic leader by interposing an exon and a splice-donor site is standard recombinant-DNA practice, and the splice consensus sequences were known (Jackson 1991). KSR supports treating the choice of a splice-donor-plus-exon leader as a predictable mechanistic variation rather than an inventive leap.
4. Reasonable expectation of success. Homologous recombination was demonstrated in human cells (Smithies 1985) and in mammalian cells generally (Thomas & Capecchi 1987); Mansour 1988 gave a working enrichment scheme. Splicing to a downstream acceptor follows the known rules. Nothing was unpredictable.
5. The "unpaired" feature is the only real novelty, and it is a deliberate subtraction, not an addition. A PHOSITA who wants the transcript to splice into the endogenous gene's exon 2 would by definition omit a 3′ acceptor. "[A] finite number of identified, predictable solutions" — KSR.

Assessment: claims A and C are, on this record, weakly non-obvious at best and vulnerable to a § 103 challenge. The inventive weight is carried almost entirely by the "unpaired splice-donor" limitation and by the specific working examples, not by the construct genus.

---

★ Combination 2 (secondary): U.S. 5,272,071 (Chappel) + WO 91/06667 (Skoultchi) + Capecchi 1989 + Smithies 1985

Same logic, different art base. Chappel and Skoultchi both teach regulatory-sequence insertion upstream of an endogenous coding region; Smithies 1985 supplies the human-cell proof of principle; Capecchi 1989 supplies the operational protocol. The motivation is identical: the field was saturated with promoter-knock-in teachings by late 1991, and the only recognized limitation was positional (where can you land the promoter without hitting an ATG?). Combining these with the standard exon/donor cassette addresses exactly that limitation.

---

★ Combination 3: Thomas & Capecchi 1987 + Mansour 1988 + a heterologous promoter + the exon/splice-donor cassette

This combination is weaker because Thomas & Capecchi 1987 is a disruption reference (knock-out), not an activation reference — you must flip its teaching from "insert neo to break the gene" to "insert a promoter/exon to switch the gene on." That direction is supplied by WO 90/11354 / Chappel / Skoultchi. Once flipped, the mechanical combination with Mansour's selection scheme is straightforward.

---

★ Combination 4: Any of the above + the dhfr/MTX amplification art (e.g., Kaufman & Sharp 1982)

Applies only to the amplification embodiment (the specification's Examples 3, 6, 8, 9; the ~70-fold hEPO increase at 0.4 µM methotrexate). Amplifying a transfected marker to co-amplify linked sequences was old, well-known, and predictable. The obviousness question is whether applying it to a targeted, endogenously activated locus was a predictable extension. On a KSR analysis, very likely yes — the patent itself describes it as simply "inclusion of an amplifiable selectable marker gene" alongside the targeting construct.

---

5. The strongest counterarguments (non-obviousness / rebuttal)

A rigorous analysis has to state where the § 103 case is weak, not just where it is strong.

1. No reference discloses an "unpaired" splice-donor site. Every prior-art construct cited is a conventional complete transcription unit, a promoterless marker, or a bare regulatory insert. The specific architecture of forcing splicing to an endogenous acceptor to build a chimeric transcription unit is not shown. This is the most defensible limitation.
2. The positional freedom in the specification is genuinely new. The patent asserts that because the construct brings its own exon and splice donor, the regulatory element may be placed "from immediately adjacent to the gene of interest… to 30 kilobase pairs or further upstream… or within an intron." That is a real departure from the prior art's teaching that the promoter must sit immediately 5′ of the coding region.
3. Primary/secondary (non-immortalized) human fibroblasts. The claims and specification emphasize targeting in primary and secondary cells — the therapeutic-relevance limitation. The reported frequencies are punishing (Example 1c: 4 G418®/6-TG® colonies per 9 million treated cells; 1 targeted event per 712 stable transfectants; Example 1d: ~1 in 80 G418® colonies with 175 bp overhangs). A PHOSITA might well have regarded ex vivo targeting of primary human fibroblasts as impractical in 1991, and that skepticism cuts against "reasonable expectation of success."
4. Objective indicia (secondary considerations). The specification itself identifies a long-felt need and a safety-driven motivation: retroviral gene therapy carried risks of replication-competent virus, insertional tumorigenesis, limited cloning capacity, and short-lived expression. To the extent this patent's platform was commercially adopted (Transkaryotic Therapies' transkaryotic-implantation program; the family's later ownership by Shire Human Genetic Therapies), there is a colorable commercial-success and licensing story. The Google Patents record's "Family has litigation" flag (Darts-IP family 22918596) evidences that these rights were economically significant — relevant to the secondary-considerations weight, though not to the merits. ⚠️ Neither I nor the earlier litigation pass was able to identify a specific case asserting 5,641,670 itself; the flag is family-level only.
5. No express teaching, suggestion or motivation to use a splice donor rather than simply placing the promoter closer. WO 90/11354 says nothing about ATGs or splicing. The motivation is therefore one I infer from the patent's own framing of the problem plus general knowledge — an inference that a patentee would contest as hindsight (the "inventor's own specification used as the blueprint for the motivation" problem that In re Naylor/hindsight doctrine warns about).

---

6. Bottom line

Framed as an Office action: the strongest rejection would be WO 90/11354 in view of Mansour et al. (1988) and further in view of Jackson (1991) — with Chappel '071, WO 91/06667, Smithies (1985) and Capecchi (1989) as cumulative/alternative secondary references — against claims A, B and C. I would not expect a clean § 102 anticipation on this record, because no single reference discloses the "unpaired" splice-donor limitation. The patent's validity therefore turns almost entirely on whether the "unpaired splice-donor + exogenous exon" architecture is treated as an inventive contribution or as a predictable design choice — and on whether the primary/secondary human cell limitation is given patentable weight.

---

7. What must be verified before relying on this

---

Cited sources

• US 5,641,670 full text — https://patents.google.com/patent/US5641670/en
• WO 90/11354 (Institut Pasteur) — https://patentimages.storage.googleapis.com/b4/6f/f0/c21127b5318dbb/WO1990011354A1.pdf
• EP 0 452 484 / WO 91/06667 (Cell Genesys, Skoultchi) — https://patent.public.lu/fo-eregister-view/search/details/591900640_EPV/
• US 6,270,989 (family member; characterizes the four prior-art references) — https://patentimages.storage.googleapis.com/f8/f0/7e/4b3590ca0b1f61/US6270989.pdf
• US 2005/0129669 (same specification, family) — https://patents.justia.com/patent/20050129669
• Mansour, Thomas & Capecchi, Nature 336:348-352 (1988) — https://pubmed.ncbi.nlm.nih.gov/3194019/ ; https://collections.lib.utah.edu/ark:/87278/s60p1hnh