$\newcommand{\nmrl}[1]{\overline{ #1 }}\newcommand{\godel}[1]{\ulcorner #1 \urcorner}\newcommand{\PR}{\mathrm{Pr}_T}\newcommand{\PROOF}{\mathrm{proof}_T}$
This seems to be just following definitions. I'll use the following notations to be somewhat precise:
- Given a formula
$\phi$, by $\godel{\phi}$ I'll denote the Gödel-number of $\phi$ with respect to some fixed arithmetization of the syntax.
- Given a natural number $n$, by $\nmrl{n}$ I mean the term $\overbrace{S \cdots S}^{n\text{ times}}0$.
Using the standard techniques $\PR ( x )$ is just $( \exists y ) ( \PROOF ( y , x ) )$, where $\PROOF ( y , x )$ means "$y$ is the encoding of a proof of (the formula coded by) $x$ from $T$" If $T$ is a recursive theory, then the standard techniques give that $\PROOF$ is $\Delta_0$, which means that $\PR$ is $\Sigma_1$. Therefore if $T \vdash \PR ( \nmrl{\godel{\phi}} )$ by $\Sigma_1$-soundness it follows that $\PR ( \nmrl{\godel{\phi}} )$ is true in the standard model, and so there is an $n \in \mathbb{N}$ such that $\mathbb{N} \models \PROOF ( n , \godel{\phi} )$. But then $n$ really codes a proof of $\phi$ from $T$, so we can decode $n$ to get a sequence $\phi_1 , \ldots , \phi_n$ of formulae which serves as a proof of $\phi$ from $T$: $T \vdash \phi$.
Note, in particular, that $\PR ( \nmrl{\godel{\phi}} )$ is $\Sigma_1$ regardless of what the formula $\phi$ is (since we just translate it into a term).