Ambrosiaster was not the only Latin writer in the latter fourth century to contend against the Greek tradition. In his treatise On the Perpetual Virginity of Blessed Mary: Against Helvidius, Jerome faults Helvidius for contending that the Greek manuscripts are corrupt:
“you with marvellous effrontery contend that the reading of the Greek manuscripts is corrupt, although it is that which nearly all the Greek writers have left us in their books, and not only so, but several of the Latin writers have taken the words the same way” (Helv. 8; NPNF 2.6, 338)
Jerome later continues the accusation:
“Seeing that you have been foolish enough to persuade yourself that the Greek manuscripts are corrupt, you will perhaps plead the diversity of readings” (Helv. 16 (18); NPNF 2.6, 343)
But which verse and variant does Jerome have in mind?
Luke 2:33 in the contest over Mary’s perpetual virginity
In chapter 8 of Against Helvidius, Jerome argues against Helvidius that Joseph and Mary remained celibate after the birth of Jesus, noting that Joseph had been persuaded to do this after witnessing the miracles that attended Jesus’ birth. As proof Jerome cites Luke 2:33 before immediately digressing to discuss the textual variant:
“His father and mother were marvelling at the things which were spoken concerning Him.” (Luke 2:33) And yet you with marvellous effrontery contend that the reading of the Greek manuscripts is corrupt …
But what variant does Jerome have in mind? Luke 2:33 is known for one significant variant (“Joseph” for “his father”):
“Joseph and his mother were marvelling at the things which were spoken concerning Him.”
Because Jerome cites the form “his father and mother,” it seems Helvidius must have cited the variant, “Joseph and his mother.” The latter has been attributed to an impulse to protect the virginal conception, as observed by B. M. Metzger:
“In order to safeguard the doctrine of the virgin birth [sic] of Jesus, ὁ πατήρ was replaced by Ἰωσήφ in a variety of witnesses …” 
Of course, Helvidius did not oppose Mary’s conceiving as a virgin, but rather the notion that she maintained her virginity for life (i.e. perpetually). Why then would he care that the reading was “Joseph” rather than “his father”?
The reason becomes fully apparent only in chapter 16 (18 [NPNF]). Here Jerome addresses one of Helvidius’ main arguments, namely, his appeal to the New Testament references to the brothers and sisters of Jesus as evidence that Mary did not remain a virgin:
“It is clear that our Lord’s brethren bore the name in the same way that Joseph was called his father [i.e. not literally]” (Helv. 16 (18); NPNF 2.6, 343)
Jerome contends that terms such as “brother,” “sister,” and “father” are not literal but refer to members of the extended family, appealing to the Roman conception of paternitas, in Jerome’s own words, “many generations spreading from one root” (ex una radice multa generis turba diffunditur) (Helv. 14 (16); PL 23, 0197, my trans.).  So just as Joseph is not Jesus’ literal father, so the “brothers and sisters” mentioned in the New Testament are not literal brothers and sisters, but perhaps cousins.
According to Jerome, Helvidius prefers the reading “Joseph” over “his father” because his argument requires that the familial terms “brother and sister” imply a blood relation. This forces him into the difficulty of explaining why literal terms apply to the siblings but not to Joseph who is called “father,” but the difficulty is sidestepped by identifying Joseph by name:
“The Evangelist himself relates that His father [‘Joseph’ v.l.] and His mother were marvelling at the things which were spoken concerning Him, and there are similar passages which we have already quoted in which Joseph and Mary are called his parents. Seeing that you have been foolish enough to persuade yourself that the Greek manuscripts are corrupt, you will perhaps plead the diversity of readings.” (Helv. 16 (18); NPNF 2.6, 343)
So given Jerome’s charge, it is clear Helvidius had argued that the Greek tradition was corrupt in Luke 2:33, but the Latin was not. If this is so, we have another instance of a Latin writer — besides the anonymous Ambrosiaster (Comm. Rom 5:14) — who defends the Old Latin version ostensibly under the conviction that it preserved the old reading, where the Greek text had been corrupted. 
The Textual Evidence
Considering the textual evidence, it is clear that with the sole two exceptions of Codex Bezae (d) and Evangelium Gatianum (gat), the Old Latin gospels support Helvidius with iosef/ioseph: a aur b β c d e f ff2 l q r1. But the Latin column of Codex Bezae, like the Greek, sides with א B L W 1 131 700 1241 1582* and the versions Ss Sh mg Cs Cb (mss.) Gg (I.II) OS.
Here we have evidence that Bezae’s bilingual text may not be entirely unrelated to the Old Latin version. In the present case, it is clear that Bezae’s Greek column attests the reading of its Greek exemplar. But how do we explain the Latin column? Every indication suggests that Helvidius attests the predominant Old Latin reading, so it appears there are (at least) two possible explanations for Bezae’s Latin reading:
On the one hand, it is possible that D reflects the exemplar of an old layer of the Latin tradition preserved in d, that predates the introduction of the variant reading ioseph in the rest of the Old Latin version. But if D’s reading were truly the oldest Latin reading, it is striking that no other Latin witness follows this reading except the relatively late Gatianum. We might have expected at least one of the African-influenced MSS (c and e) or the early European MS (a) to have kept the old reading or at least to show signs of disturbance.
On the other hand, the alignment of the main Old Latin traditions and the isolation of d may indicate that its reading pater eius is fairly recent. It may even reflect a deliberate reaction to the Old Latin reading and selection of the Greek reading based on specific interests and concerns of Bezae’s producers. But what might these have been?
Bezae’s adoption of the Greek reading
Although Helvidius was never formally condemned, in subsequent decades, those with similar views did not fare so well, such as Jovinian, who was condemned by a Roman synod in 393. To the astute observer in the Latin West it would have been clear that any association with questionable proof texts of a proponent of low Mariology posed a certain liability. Indeed, we have reason to believe that Helvidius’ writings were initially well-known, because Jerome begrudgingly acknowledges the extent of the “scandal” they had instigated (Helv. 1). At the end of the fifth century, Gennadius still remembers Helvidius’ writings:
Helvidius … wrote, indeed, with zeal for religion but not according to knowledge, a book, polished neither in language nor in reasoning, a work in which he so attempted to twist the meaning of the Holy Scriptures to his own perversity, as to venture to assert on their testimony that Joseph and Mary, after the nativity of our Lord, had children who were called brothers of the Lord. In reply to his perverseness Jerome, published a book against him, well filled with scripture proofs (Vir. ill. 33; NPNF 2.3, 391-392)
But if d alone rejects an otherwise solid Old Latin reading that is nevertheless definitely attested as a key proof text of a writer who “attempted to twist the meaning of the Holy Scriptures to his own perversity” using a well-known variant of Luke 2:33, it seems not unlikely that this reflects a sense of caution in distancing Bezae’s text from potentially sensitive associations.
When we consider Bezae’s other well-known tendencies in the context of the latter fourth century (see this post for some examples), they tend to be consistent with a conservative, ecclesiastical, and even Roman perspective (e.g. promoting Peter, favorable to Roman officials) that reflects contemporary concerns of church order (e.g. the autonomy of female lay ascetics), discipline (e.g. Christian “Judaizing”), and doctrine (e.g. the consubstantiality of the Holy Spirit). If Bezae’s text developed in an ecclesiastical setting, as suggested by its initial corrector (G), we should expect a real hypersensitivity to heretical opinions and a self-conscious effort to distance the text from any party suspected of heresy. If this analysis is correct, the Greek reading was followed in Luke 2:33 because the Old Latin had strayed in a potentially dangerous direction.
What might this imply about Bezae?
If Bezae’s Latin reading in Luke 2:33 is not an old reading, but a recent correction to the Greek column, this might suggest that in its bilingual tradition, the Latin column was constructed with reference to the Greek column. (Of course the latter may have come about in a variety of ways, including the selective correction a Greek base to one or more Old Latin copies.) When the Greek reading was selected in Luke 2:33, the Latin column was translated and updated to reflect this reading, creating the (correct) impression that the Greek reading is the exemplar of the Latin.
It is possible to see now how a “layered” effect might have been achieved in Bezae’s Greek text through a process of selective correction of a Greek base to a reconstruction of one or more Old Latin versions. In such a model, the Greek layers reveal characteristics of the initial Greek exemplar of Bezae’s bilingual tradition, while the layers aligned with the versions reflect the well-known “free” characteristics. This two-layered model accounts for cases, as in Luke 2:33, where Bezae’s Latin column follows the Greek column in preference to the Old Latin version, while also allowing for the general resemblance of both columns to the Old Latin version.
Ironically, Helvidius’ sponsorship of the Old Latin reading in Luke 2:33 may have led to its ultimate rejection from a tradition otherwise well-inclined to preserve Latin readings. In another case, a writer who promoted Old Latin readings did not leave his name to posterity. But perhaps he knew that if he had, we would not have his writings at all.
 B. M. Metzger, A Textual Commentary on the Greek New Testament (2nd ed.; Stuttgart, 1994) 111. Presumably Metzger is not referring literally to a virgin birth (virginitas in partu), but to the virginal conception.
 He suggests they are most likely children of Mary’s sister (first cousins) (Helv. 14 [= NPNF 16]), but it seems this in-law relationship does not quite fit the paternitas scheme.
 This raises the question as to the identity of Ambrosiaster and Helvidius with the same individual, which as far as I know has not been proposed. This identification would face a few problems. First, Jerome refers to Helvidius disparagingly by name, but prefers to allude to Ambrosiaster indirectly without naming him. Second, it is generally thought that Helvidius was a lay person (see Helv. 1), but it seems Ambrosiaster a presbyter.