Reverse Genetics
Reverse genetics builds a replicating agent starting from a DNA copy of a genome, rather than deriving a genome from a cultured sample. It is the strongest available objection to the reading of virological method as self-referential, and it is set out here in detail because an objection should be met at its strongest.
This entry reports what two cases demonstrate, what their controls establish, and the four respects in which they depart from the experiment specified at The Missing Experiment.
An infectious clone is a DNA copy of a viral genome that can be introduced into cells, from which a replicating agent is then recovered. The technique inverts the direction of inference in the dependency chain: instead of deriving a sequence from a culture, it begins with the sequence and produces the agent.
This is the strongest available objection to the self-referential reading, and the two cases below are reported in detail because the argument should be met at its strongest rather than at its weakest.
Poliovirus (1981)
Racaniello and Baltimore inserted a complete cDNA copy of the poliovirus genome into the plasmid pBR322. Mammalian cells transfected with the plasmid produced infectious poliovirus. A construct lacking the first 115 bases produced none.
The negative control is the significant part. Altering the sequence while holding every other condition constant produced no agent. That is a method producing a negative result, keyed to the sequence.
Zika virus (2020)
A bacterial artificial chromosome carrying the sequence of ZIKV strain Paraiba/2015, under a cytomegalovirus promoter, was delivered into mice. The relevant details:
| Element | As performed |
|---|---|
| Host | IFNAR−/− A129 mice, 4–5 weeks old, six per group (three male, three female) |
| Routes | Subcutaneous in the footpad, intramuscular in the quadriceps, intraperitoneal in the peritoneal cavity |
| Inoculum (first design) | 1 × 10⁶ Vero cells per mouse, transfected with the plasmid |
| Inoculum (second design) | Plasmid complexed with Lipofectamine 2000, up to 100 µg per mouse in 200 µl Opti-MEM |
| Outcome | Rapid weight loss from about 6 days post-inoculation, severe paralysis, death by 8–10 days |
| Re-isolation | Agent recovered from sera of inoculated mice retained virulence in a second round of infection in new mice |
The control arms
The obvious objection is that injecting foreign material into an animal provokes an inflammatory response, so illness following injection does not identify the sequence as its cause. The study's controls address this objection directly, and the results do not support it:
- Empty-vector controls. Mice given empty plasmid with the same transfection reagent developed no disease. The vaccination experiment documents this control in three figures: the empty-vector arm itself (Fig. 7), the absence of ZIKV-specific antibodies in those mice where clone-inoculated mice developed them (Fig. 8A), and their survival to 21 days followed by death at 8–9 days after challenge with virulent rZIKV (Fig. 9B).
- The control animals were demonstrably susceptible. This is the strongest element of the design. Because the empty-vector mice died when subsequently given the agent, their failure to sicken from the empty vector is not explained by resistance, by the strain being unusually robust, or by a dose too small to affect any animal. A negative control is only informative if the control subjects could have shown the effect; here that was established rather than assumed.
- The transfection reagent was present in the control arms. Cationic lipid reagents are themselves inflammatory. That confound was included in the negative controls, and produced no disease.
- Dose-dependence. Doses of 100, 50, 25 and 12.5 µg were uniformly lethal; 6.25 µg largely was not. Injection trauma does not scale with the quantity of plasmid delivered.
- Route-dependence runs counter to the objection. Intraperitoneal delivery was the most lethal and intramuscular the least. A generic reaction to injected non-self material would not be mildest when injected into muscle.
- Second passage. Serum from affected mice caused disease in new mice. That material carries neither plasmid nor transfection reagent.
Taken together, these controls isolate the sequence as the operative variable. The objection "they injected it" is true as a description and insufficient as an explanation, because the empty-vector arm is that objection with the sequence removed, and it produced nothing in animals shown to be capable of the outcome.
Documentation status. Within the rescue experiments, the corresponding control results are reported as "data not shown". Eight results across the paper carry that label, clustering on control arms, transfection-quality checks and clinical signs. Those particular arms cannot be inspected by a reader, and nothing resting on them alone should be treated as demonstrated. The control claim above rests on the vaccination-experiment figures, which can be inspected.
This distinction is worth drawing wherever a control is invoked: an unpresented control is a statement by the authors about what they observed, not a result a reader can check. The two carry different weight, and papers routinely present them in the same sentence.
What this establishes, and what it does not
The distinction matters, because the first claim is routinely allowed to stand in for the second:
| Claim | Status |
|---|---|
| This sequence is sufficient to produce a lethal wasting and paralytic syndrome in an interferon-receptor-deficient mouse when delivered by injection, and a single residue change attenuates that outcome | Demonstrated, with controls |
| A purified particle transmits the human disease between immunocompetent hosts by the natural route | Not addressed by this experiment |
The first row is deliberately specific about the syndrome, the host and the route, because each of those qualifications is load-bearing and each is routinely dropped when the result is summarised.
The departures from the specification in The Missing Experiment are nameable, and none of them is incidental:
- The host was engineered to be susceptible. A129 mice lack the type-I interferon receptor. The model is used because the agent does not reliably produce disease in immunocompetent mice. An experiment in a host built to be unable to mount a standard response cannot establish what happens in one that can.
- No natural route was used. The agent is described as primarily mosquito-borne, which deposits material intradermally during blood-feeding. Footpad, quadriceps and peritoneal injection reproduce none of this. The paper describes its routes as those "typically used" for the agent — an indication that the standard is convention, not natural exposure.
- The starting material is a DNA sequence, not a purified particle. Steps 4 and 5 of the chain — physical purification and characterisation of a particle — are not performed and not required by this method. The technique bypasses the purification gap rather than closing it.
- Transmission was not tested. Re-isolation and re-inoculation by injection is not transmission between hosts.
- The syndrome produced is not the disease the model describes. What was measured in the mice is wasting and paralysis: body weight loss, survival, viremia, and signs recorded as tremors, disorientation, hind limb weakness and severe paralysis. The humane endpoint was a loss of more than 20% of body weight or hind limb paralysis, so survival curves record animals meeting euthanasia criteria. The paper's own introduction describes the human disease as asymptomatic or a mild febrile illness in the majority of cases, with congenital malformation in pregnancy and Guillain-Barré syndrome as the serious outcomes. Neither rash, fever, conjunctivitis, nor any congenital effect was measured. Paralysis has some kinship with the rare neurological complication, but it is the dominant outcome in the mouse and an uncommon one in people: the frequencies are inverted. Against the criterion "observe whether the same disease develops", this is a different syndrome in a different species.
- Sequence specificity was tested narrowly. Three arms were compared: the full sequence, the same sequence carrying a single valine-to-alanine substitution at residue 117 of the NS2A protein, and an empty vector. The point mutation attenuated the outcome, which demonstrates fine-grained dependence on the sequence and is stronger evidence than an empty-vector comparison alone. No scrambled, unrelated or arbitrary construct was tested. The gradient therefore runs from the full sequence, through the same sequence minus one residue, to inert DNA, and the headline contrast is between a replicating construct and a non-replicating one rather than between this sequence and another. This matters in this host in particular: A129 mice are selected for use because the absence of type-I interferon signalling makes them broadly permissive, so whether a different replicating construct would produce a comparable outcome is a question the design cannot answer.
Consequence for the dependency-chain argument
Reverse genetics runs the inference in the opposite direction to the diagnostic chain, and on that basis is sometimes offered as answering the self-referential reading. The two concern different activities: the chain this entry maps runs from clinical sample to case count, and an infectious clone appears nowhere in it. Nothing in the chain is overturned by this work, and nothing in this work depends on the chain.
What it does bear on is causation, which is why it is addressed here. It does not satisfy the specification in The Missing Experiment, and it leaves the purification, natural-route and host-competence gaps untouched.
What would close the remaining gap: the same sequence-level controls, applied to a purified particle introduced to an immunocompetent host by the claimed natural route, with transmission to further hosts measured rather than inferred.
Epistemic Status
The Zika study is observed — its full text was read, including routes, doses, host strain, control arms, dose titration and passage results. The poliovirus result is reported from an identified source rather than from the 1981 paper itself.
Whether these two cases generalise to other agents is unresolved. Whether comparable control arms exist for any given agent is a separate, checkable question that this entry does not answer.
See Also
References
- Racaniello VR, Baltimore D. "Cloned Poliovirus Complementary DNA Is Infectious in Mammalian Cells." Science 214:916–919 (1981).
- Ávila-Pérez G, Nogales A, Park J-G, et al. "In vivo rescue of recombinant Zika virus from an infectious cDNA clone and its implications in vaccine development." Scientific Reports 10:512 (2020). doi:10.1038/s41598-020-57545-2