PaCO2 decreases, and PaO2 increases in pregnancy.
The key feature of gas exchange in pregnancy is the increased metabolic rate of the combined mother/foetus organism, with a resulting increase in the total O2 consumption and CO2 production. This part of the equation makes sense. One would, therefore, expect the mother to develop a persisent hypoxia and mild respiratory acidosis, which would lead to a slight increase in respiratory rate and a raised HCO3-.
However, this is not observed. Instead, expectant mothers blow off vast amounts of CO2 and actually generate a mild respiratory alkalosis, with complete renal compensation and a normal pH- in fact this is a perfect model of an isolated, sustained respiratory alkalosis. This unexpected development is the result of progesterone. Progesterone seems to infiltrate the medullary respiratory control centres, adjusting the homeostatic setpoint for CO2. A brilliant study from 1978 led the way by feeding progesterone supplements to 11 healthy male volunteers, and observing how they hyperventilate. The net effect of this is to increase tidal volume and minute volume to a point beyond any maternofoetal oxygen demands or metabolic excretory requirements.
What is the role of this, one might ask? Surely, this physiological phenomenon must offer an evolutionary advantage beyond the preservation of maternal arterial oxygenation during high-altitude pregnancy. Or at least, there shouldnt be any disadvantage, right? However, early studies observing the effect of maternal hyperventilation have suggested that placental blood flow actually decreases in proportion to a decrease in CO2. Just as one might expect to happen in the dysregulated cerebral vessels of a person with a head injury, the immature foetal cerebral vessels also constrict, with frightening implications. Not only that, but the maternal pH is closely shared by the foetus; respiratory alkalosis would result in an increased affinity of maternal haemoglobin for oxygen, decreasing its availability to the foetus - and the foetal hemoglobin-oxygen affinity would increase, decreasing in its availability to the foetal tissues.
In short, it all seems very counterproductive. A review article by Huch (from 1986) summarises the contemporary findings, and is also at a loss for explanations. However, most of the data collected by this and other reviewers have been from healthy mothers who were compelled to hyperventilate by the experimentors. Obviously, normal pregnant hypocapnea has little effect on foetal cerebral vessels, and does not have any adverse effects on maternofoetal oxygen transport.
One might retreat to the simpicity of basic gas physics, such as Dalton's Law: if there is less CO2 in the alveolus, there is "more room" for O2, and therefore maternal arterial oxygenation increases (presumably to create a greater oxygen gradient driving it by diffusion into the placental vessels). Similarly, the slightly decreased maternal PaCO2 creates a gradient for CO2 removal from the foetus, facilitating its movement across the placenta.
A finding which lends some credibility to this theory is the discovery that maternal erythrocytes produce higher levels of 2,3-DPG during pregnancy, with a resulting decrease in the affinity of maternal hemoglobin for oxygen. One can expect this feature to enhance the delivery of oxygen to the foetus, particularly in the low oxygen tension environment of placental vessels. In the maternal blood stream, this 2,3-DPG accumulation would produce a right shift of the oxyhemoglobin dissociation curve if there was no alkalosis - but the borderline alkalaemic pH of the late-term gravida (7.40-7.47) pushes the curve in the opposite direction, and the next effect of these competing influences is actualy a preserved p50 value.
In summary, over the course of your pregnancy you an expect to develop a progressively lower CO2 and a progressively higher O2. In spite of this, because of several competing influences on the affinity of maternal haemoglobin for oxygen, the effectiveness of maternal oxygen transport seems to be largely unaffected.