Five of the thirty letters may stand in front of a root letter as a prefix, the sngon jug: ga, da, ba, ma and the small a. No other letter of the thirty can occupy that slot, and the five are not interchangeable, since which prefix a given root takes is a property of the word and has to be learnt with it.
A prefix is written and is not pronounced. That statement is exact for the modern standard language and is the source of most of the apparent difficulty of Tibetan spelling: a syllable of two spoken sounds may be written with four or five letters, the first of which contributes nothing audible at all. What it does contribute is meaning, because prefixes distinguish words that would otherwise be spelled and spoken alike, and it contributes tone, because a prefix on a syllable can affect the pitch the syllable is given.
They were not always silent. The prefixes are generally understood to record initial consonant clusters that the language once had, and some of those clusters survive today in the dialects of Amdo, where a speaker may still produce a consonant that a Lhasa speaker has lost. Tibetan spelling is therefore conservative in the strict sense: it goes on recording a stage of the language that most speakers no longer pronounce.
The prefix slot is also the reason a Tibetan dictionary cannot be searched from the left. Words are ordered by their root letter, and the prefix is stepped over.