The answer to the question depends on what is meant by lupus, and what is meant by ANA.
To start with the definition of lupus: systemic lupus erythematosus (SLE) is one of many similar illnesses that affect young women and that cause arthritis, rashes, low white blood cell counts, low platelet counts, and kidney disease. (I'll ignore, for this discussion, discoid [skin], drug-induced and neonatal lupus.) The features that unequivocally diagnose SLE are (1) high titer anti-double stranded DNA antibody, (2) anti-Sm (Smith) antibody, (3) biopsy-proven kidney disease, or (4) biopsy-proven skin disease. [The American College of Rheumatology criteria are used to group similar patients together for research and treatment trials. They are classification criteria and are not meant to be used to make a diagnosis.]
The antibody tests must be accompanied by symptoms, since antibodies alone, with no symptoms, do not diagnose the disease. If a patient has symptoms plus diagnostic antibodies or biopsy proof of disease, the patient has SLE. If the patient lacks all four, the diagnosis is presumptive, even in the presence of illness.
There is no real consensus among physicians how best to describe symptomatic patients who lack all four of the unequivocal features. This writer diagnoses "lupus-like" illness, "mixed connective tissue disease," "undifferentiated connective tissue disease," and "forme fruste (meaning incipient, or hidden form) lupus". Each has specific and separate meaning and describes different forms of illness. Most summaries report anti-DNA antibody in about half of patients, anti-Sm in about one-third, and ANA in 87-94% of patients with unequivocal SLE. Urowitz and Gladman cover this topic extensively in a textbook chapter. In another textbook, Edworthy offers a complicated algorithm for diagnosis of lupus, using ANA titer, related antibodies, complement levels and symptoms to derive "diagnostic certainty". Still other text books discuss the uncertainties of serological testing without specifying frequencies of antibody negativity. In this writer's practice, as many as one-quarter of patients referred for suspect SLE fall into the not-quite-lupus category, while persons who have any one of the four unequivocal features listed above have SLE.
Some but not all physicians diagnose lupus, even ANA-negative lupus, in cases in which, to this author, another term would be preferable. Because the medical community does not agree, whether to diagnose a patient as having lupus or lupus-like disease is not a question of the physician's being right or wrong; it is a question of a physician's style. In either case, treatment options are the same.
Now, to the question of the meaning of ANA. The antinuclear antibody test is performed with different techniques in different laboratories; some techniques are more sensitive than others, such that one laboratory may find a (usually weak) positive test while another finds it negative. Some laboratories dilute 10-fold for screening (1:10), some as much as 100-fold (1:100), and some not at all. A commonly used method starts with a dilution of 1:10, then doubles with every successive dilution, so the next specimen tested is 1:20, then 1:40, 1:80, etc., the highest number positive being what the laboratory reports (for instance, 1:1280). Most lupus patients have sera that react at very high dilutions, essentially always more than 1:80, often more than 1:5120. Depending on the laboratory's reporting habits, a test that is positive at 1:10 or even 1:40 may be called either negative or weakly positive. Most laboratories count 1:80 and higher as clearly positive. The point is that ANA-negative does not always precisely mean completely negative. Another point is that speckled (as opposed to diffuse or peripheral) ANA patterns do not read well in automated immunofluorescence tests, so may be reported as lower titer or negative than they would be if they were hand read by an experienced technician.
There is a further complication. The ANA is used to screen for lupus, not to diagnose it. This means that, for practical purposes, if the ANA is negative, lupus does not exist and no further testing need be done; indeed, some laboratories will not further screen sera that are ANA negative. If the ANA is positive, that means only that lupus is possible and that tests for antibody to double-stranded DNA, Sm (Smith), Ro/SSA (Sjogren's syndrome A), La/SSB (Sjogren's syndrome B), and RNP (ribonucleoprotein) must be performed to determine whether lupus is or is not present. Because of a variety of technical factors, it is possible to have a negative ANA but a positive specific antibody test, though this is very uncommon; for diagnosis a positive test for a specific antibody is more important than is a negative ANA test. Thus an "ANA-negative" person with strongly positive antibody to Sm would unequivocally have lupus (the author has seen many such patients). Because of this possibility, this writer, evaluating a new patient for lupus, usually simultaneously (to save time) tests for ANA, anti-DNA, anti-Sm, anti-Ro/SSA, anti-La/SSB, and anti-RNP, as well as for other abnormalities relative to the patient's specific condition. That is, he does not completely discard the diagnosis if the ANA test is negative, though the diagnosis is much less likely.
Now another complication: all of the above applies to the evaluation of new symptoms in an untreated patient at a specific time. In a lupus patient-or in a patient with suspect lupus-tests change over time, either improving or worsening. Furthermore, tests often revert to normal during treatment. If the question, "Can an ANA-negative person have SLE?" is asked of a patient with established or treated disease, it may be impossible to confirm or refute the diagnosis-on the basis of blood tests alone-made years earlier, or in a patient taking high doses of prednisone or immunosuppressive drugs. In our recent cross-sectional study of atherosclerosis in SLE studying treated patients with disease of variable duration, at the time of study, 42% had anti-DNA antibody and 10% anti-Sm. We did not specifically tabulate frequency of ANA.
The answer to the question, "Does ANA-negative lupus exist?" is technically "yes", with a large number of buts, and ifs, and whens. Another answer is that the question is not very important. It is never critical to say definitively that a given patient does or does not have lupus. What is important is to evaluate the current symptoms, to put the symptoms into an overall context that includes blood tests, duration of symptoms, other illnesses, and medications, and to develop a treatment plan based on the total information rather than on a blood test alone.
 Urowitz MB, Gladman DD. Antinuclear antibody-negative systemic lupus erythematosus. In: Lahita, RG. Systemic Lupus Erythematosus, 3rd Edition. San Diego: Academic Press; 1999. p. 319-324.
 Edworthy SM. Clinical manifestations of systemic lupus erythematosus. In: Ruddy S, Harris ED III, Sledge CB, editors. Kelley's Textbook of Rheumatology, 6th ed. Philadelphia: W.B. Saunders; 2001. p. 1105-1124.
 Koopman WJ. Arthritis and Allied Conditions. Philadelphia: Lippincott Williams & Wilkins; Philadelphia. p 1445-1502.
 Lahita RG, The clinical presentation of systemic lupus erythematosus. In: Lahita, RG. Systemic Lupus Erythematosus, 3rd ed. San Diego: Academic Press; 1999. p. 325-336.
 Fritzler MJ. Immunofluorescent antinuclear antibody test. In: Rose NR, de Macario E, Fahey JL, Friedman H, Penn GM, editors. Manual of Clinical Laboratory Immunology. Washington: American Society for Microbiology; 1992. p. 724-729.
 Roman MJ, Shanker B-A, Davis A, Lockshin MD, Sammaritano L, Simantov R, Crow MK, Schwartz JE, Paget SA, Devereux RB, Salmon JE. Prevalence and correlates of accelerated atherosclerosis in systemic lupus erythematosus: a case-control study. N Engl J Med;349:2399-2406